PATHS OF PRODUCTIVITY GROWTH IN POLAND A FIRM LEVEL PERSPECTIVE This project is carried out with funding by the European Union via the Structural Reform Support Programme and with the support and the partnership of the European Commission's DG REFORM PATHS OF PRODUCTIVITY GROWTH IN POLAND A FIRM LEVEL PERSPECTIVE November 2021 This project is carried out with funding by the European Union via the Structural Reform Support Programme and with the support and the partnership of the European Commission's DG REFORM © 2022 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denom- inations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowl- edge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. This document was produced with the financial assistance of the European Union. The views expressed herein can in no way be taken to reflect the official opinion of the European Union. Trust Fund (No. TF073493) — EC Contract No REFORM/ GA2020/007. Cover design: Wojciech Wołocznik, Cambridge, United Kingdom Interior design and typesetting: Piotr Ruczynski, London, United Kingdom CONTENTS Acknowledgments   7 Abbreviations and Acronyms   8 Executive Summary   9 I Introduction: Growth and Productivity — Where Does Poland Stand?    15 II Properties of the Firm-Level Sample and Productivity Growth in Poland    25 Characteristics of the Firm-Level Dataset    26 Productivity Growth Patterns at the Sectoral Level    30 Productivity Growth Patterns at the Industry Level    35 III What Drives Polish Firm-Level Productivity Growth?    41 Decomposition Methodology    42 Productivity Growth Decomposition in Manufacturing    44 Productivity Growth Decomposition in Construction    47 Productivity Growth Decomposition in Services    48 IV Heterogeneity in Productivity Performance and Its Determinants    53 Regression Approach    54 Firm and Sector Characteristics and Productivity Growth    59 R&D and Productivity Growth    62 V Areas for Policy Action    67 References    72 APPENDIX A.1  Endogenous and Explanatory Variables    75 A.2  Firm-Level Panel Dataset Characteristics    76 A.3  Production Function Estimation    80 A.4  R&D Dataset Description    83 A.5  Sector- or Industry-Specific Results Based on Firm-Level Data    85 A.6  Other Regression Results Based on the Firm-Level Sample    87 A.7  Selected Characteristics Based on the Aggregate Data    91 BOXES Box 1  Firm-Level Dataset — Sources and Preparation    26 Box 2  How to Measure Productivity    31 Box 3  What Does TFP Represent?    31 Box 4  How to Decompose Productivity    42 Box 5  Productivity-boosting policies    43 Box 6  Evidence from Poland’s Food Industry    46 Box 7  R&D Data Sources    63 FIGURES Figure 1  Growth of GDP Per Capita PPP (Constant 2017 International $), 1992 – 2020    15 Figure 2  GDP Per Capita, 1992 – 2020    16 Figure 3  Labor Productivity as % of Germany’s, 2018    17 Figure 4  Contributions to GDP Growth in Poland (1992 – 2019), Based on Macroeconomic Aggregates    18 Figure 5  Productivity Growth (2009 – 19)    30 Figure B3.1  What Influences Firm Performance?    32 Figure 6  Productivity Growth by Sectors (2009 – 19)    33 Figure 7  TFP Growth versus TFP Levels in the Manufacturing Sector, 2009 – 14 and 2015 – 19    35 Figure 8  TFP Growth versus TFP Levels in the Services Sector, 2009 – 14 and 2015 – 19    37 Figure 9  Firm-Level Productivity Dispersion by Sector, 2009 – 19    38 Figure 10  Growth in Market Size versus TFP Growth    39 Figure 11  Manufacturing Sector Productivity Growth Decomposition    45 Figure 12  Construction Sector Productivity Growth Decomposition    48 Figure 13  Services Sector Productivity Growth Decomposition    49 Figure A2.1  Levels of Productivity by Sector (2009 – 19)    78 Figure A2.2  Growth in Performance Indicators by Sectors (2009 – 19)    79 Figure A5.1  Disaggregation of Productivity Change on the Industrial Level — Construction Sector    85 Figure A5.2  Industry-Specific Time-Averaged Contributions to the Productivity: Melitz-Polanec Decomposition of the TFP Growth for the Entire Sample Period (2009 – 19).    86 Figure A7.1  Foreign Direct Investment to Poland    91 TABLES Table 1  Firm-Level Sample Description of the 2009 – 19 dataset.    27 Table 2  Sector-Specific Growth over the Entire Sample Period (2009 – 19)    28 Table 3  Firm Characteristics by Sector, Average for 2009 – 19    54 Table 4  Fixed-Effects Regression of TFP on Competition Indicators by Characteristics of Firms in the Manufacturing Industry, 2009 – 19    56 Table 5  Fixed-Effects Regression of TFP on Competition Indicators by Characteristics of Firms in the Construction and Services Industries, 2009 – 19     57 Table 6  Fixed-Effects Regression of TFP on R&D Variables, 2009 – 19    64 Table 7  Linear Regression of TFP Growth and Its Components on Selected Productivity Determinants, 2009 – 2019    65 Table A2.1  The Growth of Selected Industrial Characteristics in the Sample by Sectors — Manufacturing, Construction and Services. ∆ Represents the % Change of an Index Number between 2009 And 2019.    76 Table A3.1  Production Function Estimation Results    81 Table A4.1  The Descriptive Statistics, the Number of Observations, and Firms in the R&D Panel Dataset    83 Table A6.1  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Introduced Separately) in Manufacturing (2009 – 19)    87 Table A6.2  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Introduced Separately) in Construction and Services (2009 – 19)    88 Table A6.3  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Age and Ownership) in Construction and Services (2009 – 19)    90 ACKNOWLEDGMENTS The report was prepared by the Finance, Competitiveness, and Innovation Global Practice of the World Bank Group (WBG). The WBG team, led by Łukasz Marć (Economist), included Umut Kilinc (Economist), Magda Malec (Consultant) and Bartłomiej Skowron (Consultant). Mirosław Błażej and Mariusz Górajski from Statistics Poland cooperated with the WBG team on the preparation of the decom- position and regression results. We are grateful for the peer-review comments received from Marcio Cruz (Senior Economist) and Elwyn Davies (Economist). The report was prepared under the leadership and guidance of Gallina Andronova Vincelette, Marcus Heinz, Ilias Skamnelos, and Reena Badiani-Magnusson. We are also grateful to Barbara Skwarczyńska and Małgorzata Bargielewicz for their excellent organizational support, Natasha Kapil for leadership and support during the project setup, and Damian Iwanowski for useful comments and suggestions. The project was financed by the European Commission (EC) Directorate-General for Structural Reform Support (DG REFORM). Special thanks go to Kaspar Richter, Dobromiła Pałucha, Iulia-Mirela Serban, Valentin Ariton, Enrico Pesaresi, and Edward Tersmette (DG REFORM) for their cooperation, support, and feedback. The team would like to thank the government of Poland — particularly, Beata Lubos from the Ministry of Economic Development and Technology, Agata Wancio, Marta Mackiewicz, Marcin Łata, and Robert Błaszczykowski — for their support and feedback throughout the process. Special thanks go to Sta- tistics Poland — including Dominik Rozkrut, Katarzyna Szporek-Lutka, Joanna Dziekańska, Mirosław Błażej, Mariusz Górajski, Emilia Gosińska, Dariusz Kotlewski, and Magdalena Ulrichs — who partnered with the WBG on the data preparation, production functions estimation, and productivity growth decomposition. 7 Paths of productivity growth in Poland: a firm-level perspective ABBREVIATIONS AND ACRONYMS BPS Business Pulse Survey DESI Digital Economy and Society Index DG REFORM Directorate-General for Structural Reform Support EC European Commission EU European Union FDI foreign direct investment FOE foreign-owned enterprise GDP gross domestic product GVA gross value added GVC global value chain HHI Hirschman-Herfindahl index ICT information and communications technology LP labor productivity NACE Statistical Classification of Economic Activities in the European Union NIK Polish Supreme Audit Office OECD Organisation for Economic Co-operation and Development PDE private domestic enterprise PNT questionnaire on research and experimental development R&D research and development REGON National Business Registry Number SMEs small and medium enterprises SOE state-owned enterprise SP Annual Enterprise Survey TFP total factor productivity TFPQ quantity-based total factor productivity TFPR revenue-based total factor productivity UEFA Union of European Football Associations WBG World Bank Group 8 Executive Summary EXECUTIVE SUMMARY After a long period of economic transformation that included introducing a series of market-oriented reforms and joining the European Union (EU), Poland was one of the fastest-growing economies in the world by 2020. The Polish gross domestic product (GDP) per capita increased by 300 percent between 1992 and 2020, and the country reached high-income status in 2009. Despite this remarkable growth, Poland still lags many European comparator countries, with its income per capita currently at two-thirds of the EU15 average. Factors delay- ing the catch-up with advanced economies include weak innovation performance, insufficient technology adoption, and labor force digital skills that are below the EU average. Because the post-transition (capital-driven) development model might be reaching its limits, the policy focus needs to shift toward different growth engines, such as productivity. With low levels of investment and a shrinking labor force due to population aging, Poland will increasingly depend for long-term growth on productivity advances, likely more so than in other advanced economies. We employ firm-level data from Statistics Poland covering small, medium, and large enterprises between 2009 and 2019 to evaluate productivity per- formance in Polish manufacturing, construction, and non-financial ser- vices. Productivity as the technical efficiency in production can be quantified through two main indices — labor productivity and total factor productivity (TFP). Labor productivity indicates how much value each employee adds. Labor productivity growth has two sources — increase in the capital used per worker and TFP growth. TFP captures the efficiency of transforming inputs (such as cap- ital and labor) into outputs. We use both measures and employ a firm-level ap- proach, which allows identifying both the aggregate productivity drivers and the underlying heterogeneity. Aggregate productivity performance can grow through three main channels. First, productivity can rise due to efficiency im- provements within firms — by adopting better technology, increasing manageri- al skills, or innovation (the “within” component). Second, more productive firms can increase their market share within the industry, meaning that factors of pro- duction — workers and capital — are allocated to more efficient companies (“be- tween” component). Third, high-productivity firms can enter the market (“up- scaling” component), and less successful establishments can exit (“downscaling” component); together, these form the “net entry” component. 9 Paths of productivity growth in Poland: a firm-level perspective This report investigates differences in productivity dynamics across eco- nomic segments and attempts to derive policy recommendations to improve the Polish economy’s productivity performance. First, we estimate firm-level TFP, compute labor productivity indices, and analyze the main productivity pat- terns between 2009 and 2019. Second, we decompose aggregate productivity per- formance into the within, between, and net entry components using the Melitz- Polanec decomposition method to understand the underlying response behind the observed productivity growth in Polish sectors and industries. Even when there is no innovation or adoption of better technology that would increase in- dividual firm productivity, reallocating production factors such as capital and labor from less to more productive establishments increases economy-wide pro- ductivity. Therefore, barriers to this reallocation would suppress the produc- tivity performance of a given industry and hence, the aggregate productivity growth. However, significant productivity improvements require progress on every front. Even if the business environment is crystalline, there will be no growth if entrepreneurs do not have the necessary human capital to take advan- tage of it. To support productivity, Poland needs to design and adopt an effective mix of policies to improve market functioning, create an efficient business en- vironment, and provide incentives for entrepreneurship and firm upgrading. Despite Poland’s remarkable economic growth, productivity growth has stag- nated in the Polish manufacturing sector since 2012 and is significantly lower than in services and construction. The empirical analysis based on small, me- dium, and large Polish enterprises indicates that economy-wide TFP grew on av- erage by 3 percent between 2009 and 2019. However, manufacturing, construc- tion, and services follow distinctively different productivity trends. There are no significant TFP improvements in manufacturing after 2012 (1 percent growth between 2012 and 2019). At the same time, the construction and service sectors demonstrate continuous modest TFP growth of 3 percent per year. Except for 2012, labor productivity follows an overall increasing trend in all sectors over the en- tire 11-year sample period. Faster labor productivity growth compared to TFP suggests increasing capital intensity of production methods between 2009 and 2019. In other words, to a large extent, firms expanded their production by using more machines per employee rather than by improving production efficiency. In Poland, firms within narrowly defined industries are highly heteroge- nous in their productivity performance. This heterogeneity is partially driv- en by observable industry and firm characteristics such as firm size, age, and ownership status but also depends on firm-level decisions such as research and development (R&D) investments. Industries lagging the aggregate productivity 10 Executive Summary growth include manufacturing of papers, food and beverages, metals, and util- ity services. The best productivity growth performers are the construction sec- tor, manufacturing of computers and electronics, telecommunications, and ac- commodation. Because some industries and groups of firms perform distinctly worse than others, we highlight selected areas for further policy consideration.1 However, addressing poor productivity performance in specific industries is not straightforward. It might result from natural market failures or inefficien- cies that require policy attention. It might also be a consequence of poorly de- signed policies or excessive regulations that generate market imperfections. Furthermore, the economic performance of an industry is subject to changing outside conditions due to, for instance, the evolution of global value chains (GVCs) or production techniques that may provide new advantages to specific industries and disadvantages to others. Policies aiming to accelerate creative destruction can facilitate the reallocation of production factors toward more efficient uses and, in turn, lead to aggregate productivity improvements. The efficiency of resource allocation (measured by the between effect) wors- ened over time in manufacturing and was responsible for the sector’s pro- ductivity slowdown while allocative efficiency gains improved productivity performance in construction and services. Large low-productivity produc- ers in the two biggest manufacturing industries (food and beverages and met- als) increase their market share over time at the cost of more productive firms within their industries, reducing the manufacturing sector’s aggregate pro- ductivity performance. Simultaneously, allocative efficiency improved in some manufacturing industries, demonstrating significant differences in productiv- ity patterns across industries within the same sector. The deterioration in alloc- ative efficiency calls for policy attention especially because the worsening alloc- ative efficiency in manufacturing exemplifies a break in the long-lasting trend of between component driving the aggregate productivity growth in Poland. It points to the importance of removing barriers to the undisturbed flow of pro- duction factors and removing regulatory restrictions on competition. Econom- ic policy in Poland would benefit from supporting companies with high poten- tial to innovate or grow rather than helping inefficient establishments survive. 1.  The Technology Adoption Survey (Phase 2 of the project “Technological Readiness and Manage- ment Skill — Productivity Growth Drivers”) will further study one of the low productivity growth industries, namely the manufacturing of food. The Technology Adoption Survey will also investi- gate, among other industries, manufacturing of textiles, vehicles, wholesale and retail trade, and accommodation. 11 Paths of productivity growth in Poland: a firm-level perspective Productivity growth accelerated in 2017 in all sectors, mainly driven by with- in-firm productivity improvements. The Polish service sector exhibited con- tinuous growth in TFP during the sample period that was on average higher than in the manufacturing sector. Starting from 2017, productivity growth acceler- ated in all sectors simultaneously, such that the within component drove most of the recent productivity improvements. A more disaggregated sector classifica- tion shows that not all industries had positive within-firm productivity growth throughout the sample period. In industries such as the manufacturing of paper and paper products, chemicals, civil engineering, and construction of buildings, the within component was negative, which underlines the need for strengthen- ing firm-specific productivity performance in these industries. Small and medium-sized companies are the engines of productivity growth in Poland. The empirical evidence shows that small and medium firms are more likely to exhibit significantly higher productivity growth than large Polish firms. Large firms, however, do not lose their market shares in some sectors, indicat- ing inefficient reallocation of market shares. These results have three main pol- icy implications. First, empirical findings suggest that there is a need for policy intervention to intensify the competition in Polish industries. Second, remov- ing barriers to growth for smaller firms, especially in manufacturing, seems to be key to accelerating aggregate productivity growth. Enhancing the growth of high-productivity smaller firms can be achieved, for instance, through facil- itating their access to finance, promoting financial market deepening, and sup- porting the development of the innovation supply side (for instance, dedicated software) for small and medium enterprises (SMEs). Third, large firms should be incentivized (however, not financially) to improve their within-firm productiv- ity performance and further investigation is needed to understand their barri- ers to productivity enhancement. The results on the relationship between ownership status and productivity growth are inconclusive. There is some weak empirical evidence that in terms of productivity improvements, the most successful ownership type is foreign ownership. This result, however, should not be interpreted directly as the for- eign-owned firms are better and more efficiently managed. It may also be the case that the most productive firms may be the ones that are taken over by mul- tinational enterprises. The empirical analysis also suggests that firms in expand- ing industries exhibit better productivity performance than establishments in other industries. Because rising demand can motivate firms to be more pro- ductive, facilitating access to global markets might increase the demand for an industry’s products and hence the productivity growth. The empirical findings 12 Executive Summary justify policies to strengthen the linkages between Polish firms (especially SMEs) and foreign firms. Programs supporting export promotion should also be con- sidered. R&D incentives are also justified because, in manufacturing, we find evi- dence that R&D has positive effects on productivity performance. To boost Polish productivity, the empirical evidence provided in the report indicates certain areas for policy actions as well as a few directions for neces- sary further investigation. First, Polish firms need incentives to improve their capabilities through digitization, building innovation capacity, and increasing managerial skills. Second, the productivity-enhancing programs should focus on small and medium firms because they are the engines of productivity growth and exhibit higher productivity potential than large firms. Third, manufactur- ing calls for policy attention because the sector’s productivity has stagnated since 2012. The reason behind poor productivity performance in manufactur- ing is worsening allocative efficiency in the largest industries within manufac- turing, namely food and beverages and metals. The productivity in those indus- tries requires further investigation. Moreover, because delivering the most effective policy recommendations requires an evidence-based approach, it is necessary to improve firm-level data accessibility. 13 I INTRODUCTION: GROWTH AND PRODUCTIVITY  — WHERE DOES POLAND STAND? Poland was one of the fastest-growing economies in the world in recent de- cades, reaching high-income status in 2009. With GDP per capita tripling be- tween 1992 and 2020, Polish economic growth surpassed that of all peer coun- tries (Figure 1). Moreover, during that time, Poland was developing twice as fast as the EU average. Certainly, part of Poland’s remarkable growth stems from its very low initial income level. In 1992, Poland’s GDP per capita in purchasing pow- er parity (PPP) was 52 percent of the Czech Republic’s, 67 percent of Hungary’s, and 28 percent of Germany’s. Nonetheless, Poland (like other countries in Cen- tral and Eastern Europe) offered attractive conditions for foreign investors due to its geographical and cultural proximity to Western economies. These condi- tions drove foreign direct investment (FDI) and hence boosted economic growth FIGURE 1  Growth of GDP Per Capita PPP (Constant 2017 International $), 1992 – 2020 Poland Republic of Korea Romania Slovakia Central Europe and the Baltics Hungary Czech Republic EU (27) Germany Eurozone (19) GDP growth per capita, 1992=100 Source: Elaboration based on World Development Indicators. 15 Paths of productivity growth in Poland: a firm-level perspective (Grela et al., 2017). The process was further accelerated by Poland’s accession to the EU in 2004, which increased trade openness, linked Poland to GVCs, and sup- ported pro-growth reforms. Despite Poland’s outstanding growth performance for about three decades, its economy still lags those of many European countries. As depicted in Fig- ure 2, Poland reached the same GDP per capita in PPP as Hungary in 2011, and Po- land’s GDP per capita in PPP surpassed Slovakia’s in 2018. Nevertheless, in 2020, Poland’s GDP per capita PPP was still equivalent to only 63 percent of Germany’s and 84 percent of the Czech Republic’s, and it was 28 percent below the EU aver- age. Moreover, Poland’s convergence has slowed in recent years. Between 1992 and 2008, Poland went from 28 percent to 47 percent of Germany’s GDP per cap- ita in PPP. In this period, average annual growth in Poland was 4.5 times faster than growth in Germany. After the global financial crisis, however, Poland’s GDP per capita in PPP increased by only 12 percentage points. In this period, average annual growth in Poland was 3 times faster than growth in Germany. FIGURE 2  GDP Per Capita, 1992 – 2020 60,000 50,000 constant 2007 international $ GDP per capita PPP, 40,000 30,000 20,000 10,000 0 Poland Germany EU (27) Eurozone Czech Republic Slovakia Hungary Central Europe and the Baltics Republic of Korea Source: Elaboration based on World Development Indicators. A manufacturer in Poland needs almost three times as many employees to produce the same output as an average German firm. In 2018, gross value-add- ed per person employed in a manufacturing firm in Poland, a measure of labor productivity, was only 35 percent of the German average and slightly lower than 16 I. Introduction: Growth and Productivity  — Where Does Poland Stand? in Hungary or the Czech Republic (Figure 3). It was 37 percent for construction and 46 percent for services. The labor productivity across all sectors in Poland is substantially lower than in the EU, meaning that an average firm in Poland needs more than twice as many workers to produce the same output as an av- erage EU firm. The gross value-added generated by each sector does not signifi- cantly vary across countries, but the structure of firm sizes is distinctly differ- ent (Eurostat 2021). For instance, in Poland, most employees work in the micro companies (34% of the labor force), while in Germany, only 19% of the workers are employed in the micro firms. Most of the German labor force (41%) is engaged in the largest companies (with more than 250 employees). However, the aggregated value added by firms in each size class is similar between Poland and Germany. FIGURE 3  Labor Productivity as % of Germany’s, 2018 Germany EU (27) Manufacturing Hungary Czech Republic Slovakia Poland 35% Romania Germany EU (27) Construction Hungary Czech Republic Poland 37% Romania Slovakia Germany EU (27) Czech Republic Services Poland 46% Slovakia Hungary Romania 0 10 20 30 40 50 60 70 80 90 100 Labor productivity (gross value added per person employed), Germany = 100 Source: Elaboration based on Structural Business Survey (Eurostat 2021). Note: Latest available data is for 2018. Poland’s post-transition growth model based mainly on capital inflows will not be sufficient to secure convergence to its main aspirational peers from Western Europe. The growth of Poland’s economy after economic transforma- tion was driven by capital accumulation (49 percent contribution to GDP growth) 17 Paths of productivity growth in Poland: a firm-level perspective and improving production efficiency (37 percent contribution to GDP growth), with the remaining 14 percent associated with an increase in the size of the labor force (Figure 4). It is an upcoming challenge to maintain a strong pace of development while the capital-driven growth model is reaching its limits (Grela et al., 2017). Due to diminishing returns (for example, from education and capital), it will not be possible to maintain equally strong physical and human capital growth. Grela et al. (2017) argue that the new factors contributing the most to growth are related to structural competitiveness, innovation activity, and institutional environment. These are productivity components. Additionally, it is expected that exports will remain a significant growth driver for direct value added as they were in the past (Szpunar and Hagemejer, 2018). Building on these engines of growth will be cru- cial to counteract a substantial demographic challenge in the years to come caused by low fertility rates. The European Commission (2020) projects that Poland will reach the EU27 average old-age dependency ratio (52 people aged over 64 years for each 100 aged 15 to 64) in 2050 and will have the worst effective economic old- age dependency ratio in the EU in 2070 (92 versus 70 in the EU27). FIGURE 4  Contributions to GDP Growth in Poland (1992 – 2019), Based on Macroeconomic Aggregates 8 Percentage point contribution to GDP growth 7 6 5 4 3 2 1 0 -1 -2 -3 -4 92 93 94 95 96 19 7 98 20 9 00 20 1 02 20 3 20 4 20 5 06 20 7 08 09 10 11 12 13 14 15 16 20 7 18 19 0 9 0 1 0 0 9 0 20 20 20 20 20 20 20 20 19 20 20 20 19 19 19 19 19 19 20 20 Labor Capital Productivity (residual) GDP growth Source: Elaboration based on the Conference Board Total Economy Database (Conference Board 2021). Note: The presented values are a rough estimation of actual GDP growth constituent parts. Contributions to GDP growth are calculated by weighting the growth of the input by their respective share in income, the labor income share for labor in- puts, and (1 − labor income share) for capital inputs. The productivity is a residual of GDP growth minus the input contri- butions (labor and capital) and not the production function estimation result. The numbers might not correspond to the firm-level productivity estimation presented in the report. 18 I. Introduction: Growth and Productivity  — Where Does Poland Stand? With its income at two-thirds of the EU15 level and the capital-driven devel- opment model weakening, Poland needs to realize substantial economic growth based on productivity improvements. A country’s ability to improve its standard of living over time depends almost entirely on its ability to raise its productivity (Krugman, 1994). A global review by ILO (2013) finds that an in- crease in individual firm productivity (the within component) is the most criti- cal factor contributing to economic growth. Such within-firm performance can be complemented by the reallocation of resources between firms and sectors (the between component), the entrance of high-productivity firms, and the exit of low-productivity firms. As a matter of fact, the advances in the between com- ponent were the main driver of the TFP growth in Poland between 1997 and 2013 (World Bank, 2017). Putting productivity at the center of Poland’s growth agen- da means focusing on monitoring and developing programs that support pro- ductivity growth (while not disturbing unnecessarily the functioning of the market), with particular attention to strengthening firms’ productivity — inno- vation or adoption of better technologies, digitization, managerial and organi- zational talent, and human capital skills. Innovation and R&D positively influence technological progress, contribut- ing to productivity increases and, hence, economic growth. Because produc- tivity is becoming the essence of economic growth, its determinants are also the center of research attention (Doraszelski and Jaumandreu, 2013). Nekrep et al. (2018) argue that there is a causal link between R&D expenditures and subse- quent productivity growth in the EU countries between 1995 and 2013. Pop Silaghi et al. (2014) differentiate between the impact of private and public R&D spend- ing. They find that only market R&D expenditures positively influenced growth in ten new member states between 1998 and 2008, whereas public R&D spending had a negligible effect. Additionally, Verspagen (1995) suggests that finding a sig- nificant link between R&D and productivity strongly depends on the empirical methodology and that R&D has a positive impact on productivity only in high- tech sectors. These contradictory findings imply that simple implementation of the best policy practices is not enough to enhance growth. A custom productiv- ity-enhancing policy formula must be based on country-specific factors, like a country’s economic structure and its institutional and regulatory environment. Poland’s innovation system is one of the least developed in Europe. Poland lags most countries on the European innovation scoreboard (European Commission 2021), surpassing only Romania, Bulgaria, and Latvia. The poor innovation eco- system in Poland results from a low number of SMEs introducing product and business process innovations, small employment in knowledge-intensive sectors 19 Paths of productivity growth in Poland: a firm-level perspective and innovative enterprises, and a maladjusted academic research system. The Community Innovation Survey 20182 found that only 22 percent of non-micro companies in Poland were innovative, the second-lowest result in the EU. The share of the population with tertiary education, the number of design applica- tions, and environment-related technologies are the most positive aspects of Poland’s innovation ecosystem. Overall innovation performance is also related to R&D expenditures (Mohnen and Hall, 2013), which amounted to 1.32 percent of Poland’s GDP in 2019 compared to the European average of 2.19 percent. The weak innovation performance of Poland’s businesses is partially caused by insufficient technology adoption coupled with inadequate digital skills among the labor force. The Digital Economy and Society Index (DESI) is a tool used by the European Commission to monitor the digital progress of member states in five categories: broadband connectivity, human capital, use of internet services, integration of digital technology, and digital public services. In 2020, Poland ranked 23rd of 28 EU member states in DESI. The category that drags down Poland’s rank the most is the integration of digital technologies, which considers the use of technologies such as electronic information sharing, big data, the cloud, and the degree of e-commerce utilization. Moreover, Poland needs to increase the digital skills of its labor force, which are currently below the EU average. Im- proving digital skills requires not only providing basic (or better) digital skills to the whole population but also training information and telecommunications technology (ICT) specialists and equipping them with the most advanced skills. The general attitude of Poland’s companies to training and digitization and the slow evolution of managerial practices indicates that state intervention can significantly accelerate the digital transformation of the economy. The World Bank’s (2021) Business Pulse Survey (BPS) 4 shows that 50 percent of en- terprises in Poland are convinced they do not need further digitization. What is more, over half of the firms did not train their employees in 2020, and among those that did not, as many as two-thirds think their employees’ skills are ade- quate.3 BPS 4 findings also point out the slow adoption of good managerial practic- es. The lack of knowledge is often a greater barrier to technology adoption than access to the technology itself (see Arendt, 2008; Bartel et al., 2007; Martin et al., 2013). Not knowing one’s shortcomings, accompanied by a visible technological 2. https://ec.europa.eu/eurostat/web/science-technology-innovation/data/ database?node_code=inn_cis11 3.  Likely, the stand applies not only to the skills but to the productivity as well. In the UK, 79 percent of leaders believe that they are at least as productive as their peers (Wu and Broughton, 2019). 20 I. Introduction: Growth and Productivity  — Where Does Poland Stand? gap between Poland and peer countries, creates a rationale for public interven- tion. However, interventions that focus on just financing (with loans or subsidies for capital equipment) and ignore managerial capabilities and the human fac- tor do not yield the full benefits of digital technology adoption (Matteucci et al., 2005). Embracing the full benefits of digital technology requires complementary firm-level capabilities — technical, managerial, and organizational (OECD, 2019). Effective policy design aiming to boost productivity requires not only identify- ing aggregate productivity trends but also determining micro productivity driv- ers. To investigate productivity dynamics across economic segments and to derive policy recommendations to improve the productivity performance of Poland’s econ- omy, one needs to build the analysis on firm-level data. Each industry faces unique market conditions and is endowed with distinct technologies and skills. Without comprehensive knowledge about the nature of this heterogeneity, it is difficult to de- sign a productivity-enhancing system of incentives successfully and cost-effective- ly. Employing firm-level data enables determining both the aggregate productivity trends and the underlying heterogeneity across sectors and industries. Moreover, it allows decomposing the productivity growth using the Melitz-Polanec method into four components — within, between, upscaling, and downscaling — that represent different drivers of productivity growth (Melitz and Polanec, 2015). Furthermore, it helps to determine whether the potential drivers of productivity growth that are well-known in the literature (investing in R&D and strengthening competition) are also effective in influencing the performance of firms in a particular country. After economic transformation, the productivity growth in Poland was his- torically driven by improvements in allocative efficiency and growth of pro- ductive firms. The most recent study decomposing productivity growth using the Melitz-Polanec method investigated firm-level data in Poland between 1997 and 2013 (Albinowski et al., 2015; World Bank, 2017) focused on the manufactur- ing sector. The TFP growth in manufacturing between 1997 and 2013 was impres- sively fast, mainly driven by a reallocation of resources from less to more pro- ductive firms (the between component accounted for three-fourths of aggregate TFP growth). In fact, a further empirical investigation indicated that many small firms were forced to exit or downsize at that time, but a few grew fast and created enough new jobs to accommodate laid-off workers. Moreover, the report found that foreign-owned firms were on levels more productive than domestic firms, but domestic firms experienced faster TFP growth. To the best of the authors’ knowledge, the decomposition was never applied to post-2013 data (most likely due to restricted access to firm-level data in Poland) and construction and ser- vice sectors. The following study aims to fill this research gap. 21 Paths of productivity growth in Poland: a firm-level perspective This report investigates the aggregate patterns and underlying heterogene- ity in productivity growth in Poland across three sectors — manufacturing, construction, and services — and industries within these sectors between 2009 and 2019 and formulates policy recommendations aiming to enhance the economy’s productivity performance. The remainder of the report is struc- tured as follows. Chapter 2 provides a detailed description of the firm-level data- set used in the study and reports productivity growth patterns at the aggregate, sectoral (manufacturing, construction, non-financial services), and further dis- aggregated industrial levels. The chapter also describes the methodology of labor productivity computation and TFP estimation. Chapter 3 documents the productivity growth channels by presenting the sectoral results of a dynamic Melitz-Polanec productivity growth decomposition. Chapter 4 provides in-depth empirical analysis by identifying the underlying productivity heterogeneity across different firm groups. It also investigates the well-established produc- tivity growth determinants (R&D investments and competition). The report con- cludes with a policy recommendations chapter. This study is part of a project “Technological readiness and management skills — Productivity growth drivers in Poland” conducted in collaboration with DG REFORM. The project aims to support the Ministry of Economic Devel- opment and Technology in enhancing the effectiveness of firms’ support sys- tems in Poland by providing evidence based on firms’ capabilities, context, and barriers to productivity growth. The project consists of three phases. Phase 1 focuses on understanding firm-level productivity dynamics and analyzing in- struments supporting managerial skills and technology adoption. Phase 2 pro- vides evidence based on Polish firms’ capabilities by implementing and analyz- ing a Technology Adoption Survey (following Cirera et al., 2021). Phase 3 aims to build capacity and support in the redesign of instruments to build firms’ capa- bilities. The following report is the main output of Phase 1 and provides infor- mation necessary for successfully conducting Phases 2 and 3. 22 II PROPERTIES OF THE FIRM-LEVEL SAMPLE AND PRODUCTIVITY GROWTH IN POLAND 1 We employ a firm-level panel dataset covering small, medium, and large compa- nies between 2009 and 2019 to evaluate the productivity performance in Polish manufacturing, construction, and private non-financial services. The analysis of the productivity dynamics of Polish enterprises relies on two measures: gross value added per worker (labor productivity) and TFP. 2 Labor productivity and TFP grew during the analysis period on average by 4% and 3% per year respectively. However, the manufacturing, construction, and ser- vices sectors followed distinctively different productivity trends: growing pro- ductivity in construction and services and stagnating productivity in manufactur- ing. Moreover, there was significant heterogeneity in firm performances not only across the main three sectors, but also within the sectors, across industries. 3 The manufacturing sector calls for particular attention because TFP in the sec- tor has stagnated since 2012. The largest industries of the sector, namely the manufacturing of food, metals, and chemicals, are dragging down overall pro- ductivity growth. Moreover, labor productivity growth that significantly outpaces TFP growth indicates that the expansion of the manufacturing industry comes primarily from increasing capital intensity rather than improvement in techni- cal efficiency. 25 Paths of productivity growth in Poland: a firm-level perspective Characteristics of the Firm-Level Dataset The analysis uses a firm-level panel dataset covering small, medium, and large Polish companies in the manufacturing, construction, and service sec- tors from 2009 to 2019.4 The data originates from the Statistics Poland annual economic activity surveys of non-financial enterprises. All firms with more than nine employees operating in Poland are obliged to fill out the survey. To create a consistent firm-level dataset for the study, one needs to combine micro- data surveys, remove outliers, and deal with missing values (Box 1). One-third of all observations were dropped in the procedure. The final dataset comprises almost 140,000 firms and 700,000 observations reported during an 11-year-pe- riod (Table 1).5 Unfortunately, the universal reporting obligation does not apply to micro firms6, which play a substantial role in the Polish business environment, employing 40 percent of the labor force and generating 30 percent of total value added. Five percent of micro firms are appointed to fill out the economic activ- ity survey every year, and only a small portion answer the survey in consecu- tive years. Because the database for micro firms does not have a sufficient panel structure, the study excludes micro firms. BOX 1  Firm-Level Dataset — Sources and Preparation The 2009 – 19 firm-level panel dataset used in this study is based on the Annual Enterprise Survey (SP), maintained by Statistics Poland. The firm-level sample covers all non-financial enterprises operating in Poland with more than nine employees. First, yearly surveys were merged to form a panel dataset based on the National Business Registry Number (REGON) identifier. Second, the dataset was cleared of outlier observations. Third, the dataset was further edited to remove all observations with zero or negative gross value added (the difference between deflated output and intermediate consumption) and capital. Lastly, every observation reporting non-positive interme- diate inputs (sum of energy and material costs) was dropped. The production function estimation method employs intermediate inputs as proxies for unobserved productivity. The missing data on gross value added, capital and costs sum to 33 percent of all observations. The detailed dataset preparation and a description of endogenous and explanatory variables are reported in Appendix A.1 The nominal variables in the dataset (gross value added, capital, and investments) were transformed into real values using industry-level price indices (published yearly by Statistics Poland as Prices 4.  Throughout the report, the word “sector” applies to the three major economy parts — manu- facturing, construction, and services. The division follows the statistical classification of eco- nomic activities used in the European Union (NACE Rev. 2). We disaggregate the sectors further into industries (level 2 of NACE), such as the manufacture of textiles, construction of buildings, or telecommunications. 5.  Because the firm-level data is strictly confidential, Statistics Poland performed the data cleaning process, calculated the variables, and aggregated them to the NACE 2 level. 6.  Micro firm employs up to 9 full-time workers. 26 II. Properties of the Firm-Level Sample and Productivity Growth in Poland in the national economy) with the base year of 2010 (constant prices). The industrial deflators are defined at the level of the 2-digit NACE Rev. 2 (The Statistical Classification of Economic Activities in the European Union). As explained in Box 3, because firm-level prices or quantities are not availa- ble, we do not use a quantity-based productivity indicator. The study analyses the economic activity of enterprises classified at the 2-digit NACE level from divi- sions 10 to 88. It covers a wide range of industries — manufacturing, construction, and services (for example, trade, transport, and ICT). Due to the highly regulated business environment in divisions 12 (tobacco) and 19 (coke and refined petroleum products), we excluded those industries from the sample. Following the literature, we also left out the financial and insurance companies (divisions 65 – 66) because, among other reasons, their balance sheets differ significantly from the non-finan- cial firms. Because the number of observations in some divisions (for example, manufacturing of beverages or veterinary activities), is too low for the production function estimation, we grouped them following guidelines from the statistical office of the EU (Eurostat). The details of merging selected industries and their economic properties are given in Appendix A.2. The sample comprises about 140,000 small, medium, and large companies in an 11-year-period. As expected, the panel is unbalanced, with an average of about five observations per firm. Table 1 reports the number of firms and their market share in terms of gross value added and employ- ment by sector. In the firm-level dataset used in this study, TABLE 1  Firm-Level Sample Description of the real gross value added and total employ- the 2009 – 19 dataset. ment in the Polish manufacturing and ser- Labor GVA vice sectors experienced almost uninterrupt- Firms share share ed growth between 2009 and 2019. The real Sector (no.) (%) (%) gross value added of manufacturing and ser- Manufacturing 39,341 40 37 vices companies increased by 51 percent and Construction 19,807 7 7 53 percent, respectively, accompanied by a 12 percent and 16 percent rise in the total employ- Services 77,505 53 56 ment (Table 2). The only exception was the af- Source: Elaboration based on Statistics Poland data. termath of the global financial crisis, which re- Note: GVA = gross value added. sulted in a mild drop in values of both indices until 2013. (See Appendix Figure A2.2) The growth patterns of the performance indicators in the manufacturing and services followed each other closely over the sample period except for a slight fall in the total employment within ser- vices in 2013. What is more, there was a universal increase in real wages across sectors: 60 percent on average in the firm-level dataset between 2009 and 2019. The real average wage growth is higher than observed in the national econo- my — which equaled 35 percent between 2009 and 2019 according to Statistics Poland — because micro firms, the agriculture sector, and civil services are ex- cluded from the sample. 27 Paths of productivity growth in Poland: a firm-level perspective TABLE 2  Sector-Specific Growth over the Entire Sample Period (2009 – 19) Sector ∆ TFP ∆ LP ∆ GVA ∆ Employment ∆ Wages ∆ Revenues (%) (%) (%) (%) (%) (%) Manufacturing 22 56 51 12 68 60 Construction 39 44 11 −27 59 23 Services 38 35 53 16 55 56 Sample 31 44 49 11 60 54 Source: Elaboration based on Statistics Poland calculations. Note: Detailed sector-specific growth, divided into industries, is given in Appendix A.2. ∆ = delta (change), GVA = gross value added, LP = labor productivity, TFP = total factor productivity. The volatility of total employment in the construction sector is rather high, mostly because micro companies were not analyzed in this study. The total employment dropped by 27 percent between 2009 and 2019 among small, medi- um and large firms in the dataset and by 25 percent in the same group accord- ing to Poland’s labor force survey, which covers the whole economy (Statistics Poland, 2021). At the same time, employment in micro firms rose by 29 per- cent according to the labor force survey. In other words, the labor force out- flowed to the micro firms that are unobserved in the firm-level dataset, which explains the detected volatility. This feature corresponds with the change in the number of construction companies in the economy over the sample peri- od. According to the Activity of non-financial enterprises reports published by Statistics Poland, the number of micro firms in the economy rose by 47 percent between 2009 and 2019, accompanied by a drop in the number of larger firms. The growth of the number of micro firms within the sector is likely a result of construction business practices in Poland. Namely, as a bankruptcy precau- tion, real estate developers tend to establish a new entity for every construc- tion project (Szreder, 2018; Office of Competition and Consumer Protection, 2021). The decrease in employment in the construction sector was accompanied by a far slower growth in the real gross value added and revenues — of 11 per- cent and 23 percent, respectively — than for the other two sectors. Moreover, the construction of buildings was the only industry in any sector that expe- rienced a fall in the real gross value-added: 3 percent between 2009 and 2019. (See Appendix A.2.) The construction sector experienced a massive disruption in 2011 – 2013 due to a demand boom following public investments for the UEFA Euro 2012 and EU structural funds co-financed projects. Between 2011 and 2012, con- struction companies experienced a drop of more than 10 percent in real gross 28 II. Properties of the Firm-Level Sample and Productivity Growth in Poland value-added. (See Appendix Figure A2.2) The highly volatile performance of firms within the sector corresponds to the wave of bankruptcies and liquidity problems of the construction companies in that period that is being attributed to the issues in the payment system developed by the General Directorate for National Roads and Motorways for railways and football infrastructure build- ings (Supreme Audit Office, 2018).7 During 2012, about 10 percent of all construc- tion companies declared bankruptcy, which accounted for over a quarter of all collapsed firms, 80 percent more than a year earlier and four times more than in 2008. The road system in Poland had never been worse than before the EU accession, and from 2004 onward, the country landscape changed into a con- struction site. The bankrupted companies engaged in the most significant in- frastructure projects financed by the EU funds. The construction sector also experienced an enormous drop in real investments,8 namely 80 percent in the firm-level data, between 2011 and 2012. First, the sharp fall in the investment rate was primarily caused by a decline in public investment, as the large construc- tion projects connected to preparations for the Euro 2012 were ending. Second, following the global trend, due to policy uncertainty for investors and elevat- ed geopolitical risks, there was a slowdown of the net inflow of FDI to Poland. (See Appendix Figure A7.1). Moreover, the scale of capital in transit (meaning the transactions of special purpose entities that transfer funds at the mother company’s request for the purposes of tax optimization) severely influenced the value of global FDI in Poland.9 7.  Poland co-hosted the UEFA Euro 2012 with Ukraine. Hosting the event set in motion a massive stimulus for public infrastructure investments. Due to increased demand for building materi- als, their price increased substantially; for example, the price of tarmac rose by 25 percent in 2011 (Supreme Audit Office, 2018). However, the public tenders didn’t contain an indexation clause. The construction companies didn’t forecast high material costs and were forced to operate with nega- tive returns or declare bankruptcy. Moreover, the large construction consortia were not paying their subcontractors on time (or at all), which disturbed their financial liquidity and brought them to default. According to the report prepared by the Polish Supreme Audit Office (Supreme Audit Office, 2018), the public tenders were settled mainly based on the lowest price criterion. There were several problems. First, severe competition on the market led to tendering below the real construc- tion costs. Second, the General Directorate did not always verify the contractor’s financial condi- tion and signed contracts with unreliable construction companies. Third, the public tenders did not give the General Directorate the authority to monitor the payments to subcontractors. 8.  Real investments are calculated as a log of outlays on tangible and intangible assets (reported in SP surveys), deflated with the investment price deflators at 2-digit NACE Rev. 2 (published yearly by Statistics Poland as Prices in the national economy). 9.  More on the global FDI transfers can be found in World Investment Reports (UNCTAD 2013, 2014, 2015). 29 Paths of productivity growth in Poland: a firm-level perspective Productivity Growth Patterns at the Sectoral Level Establishments in Poland’s manufacturing, construction, and services had predominantly positive productivity growth between 2009 and 2019, but the growth rates fluctuated considerably over time. The evaluation of the efficien- cy performances of Poland’s enterprises relies on two measures of productiv- ity: gross value added per worker (labor productivity) and TFP. Labor produc- tivity growth has two sources: capital deepening (increases in the capital used by workers) and change in TFP. The latter is the traditional measure of efficien- cy that captures output not explained by intermediate inputs, labor, and capi- tal. (See Box 2 and Box 3 for details.) Over a decade, labor productivity — result- ing from the growth of capital used by each worker and TFP growth — improved by 44 percent, while TFP increased by 31 percent, translating to average annu- al growth of 4 and 3 percent respectively. Labor productivity and TFP followed a similar trend over the sample period (Figure 5). In 2009 – 11, companies dis- played on average fast productivity growth that can be considered the econom- ic rebound from the global financial crisis. From 2012 until 2015, productivity growth fluctuated around zero for TFP and around 2 percent for labor produc- tivity. Starting in 2015, productivity growth recovered, reaching its peak in 2017. There is an apparent slowdown in productivity for the last two years in the sam- ple, with TFP even slightly decreasing between 2018 and 2019. Faster labor pro- ductivity growth compared to TFP suggests that firms were using increasingly more capital relative to labor in production between 2009 and 2019. FIGURE 5  Productivity Growth (2009 – 19) 10 8 Productivity change (%) 6 4 2 0 -2 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Labor productivity TFP Source: Elaboration based on Statistics Poland calculations. 30 II. Properties of the Firm-Level Sample and Productivity Growth in Poland BOX 2  How to Measure Productivity Productivity as the technical efficiency in production can be quantified through two main indi- ces — labor productivity and TFP. Labor productivity indicates how much value added is produced per employee. Thus, it indicates how efficiently labor is employed in production, which also depends on the intensity of capital in the production process. Consider two hypothetical textile manufacturers. The first one produces handcrafted shirts. The second one uses high-tech equipment to produce shirts of similar quality. The equipment only requires one machine operator rather than several sewers. We can measure labor productivity in this case as the ratio of one shirt to the number of workers employed to pro- duce it. It would be lower for the producer that handcrafts shirts, even though both producers make shirts of the same quality. Labor productivity is easily measured and comparable between coun- tries, keeping in mind the limitation that it depends on capital intensity in the production process. TFP captures how efficiently firms transform inputs (in our case, capital and labor) into outputs. Thus, it captures the increase in output that is not attributed to a change in the quantity of factors of production. The higher the TFP, the less input is needed for a given output. TFP can depend on a range of factors, such as skills, organizational structure, managerial talent, and adaptation or inno- vation of new or better technologies and processes to produce larger amounts or higher-quality products or services with fewer resources. TFP is not observable from the data directly but can be obtained as a residual of the production function. TFP can be computed using both macro and micro data. Micro-level TFP, however, has some advan- tages. First, it enables a more in-depth analysis by identifying the underlying productivity heter- ogeneity across different firm groups and industries. Second, it allows employing decomposition methods to indicate the channels of productivity change. Third, it addresses the endogeneity issues that can contaminate the causal link between TFP and usage of factors of production. The method- ological details of calculating TFP are described in Appendix A.3. Labor productivity is computed as a log of the ratio of gross value added to the log number of full- time employees. In this study, we use firms’ employment shares as the weights in aggregating firm- level productivity to reach the economy-wide averages. The complete definitions of the variables are provided in Appendix A.1. This study focuses on the dynamic aspects of productivity growth, namely how productivity evolves over time, rather than the levels. Applying different weights to calculate the economy average or sectoral averages affects the productivity value, but the dynamic analysis is free of such influence. BOX 3  What Does TFP Represent? TFP can be interpreted as the part of firm-level value added that cannot be explained by the quan- tity of traditional inputs (in our case, capital and labor) used in production (Figure 8). To draw a comparable TFP index, one needs to deflate nominal firm-level variables to adjust, for example, firm sales for inflation and express time-series data in one comparable measure (“constant” prices com- parable across years). We deflated nominal firm-level variables using industry-level price indices. Using industry-level prices in the computation of firm-level TFP causes the productivity to reflect firm-specific variation in the prices, so the TFP index used in this study embodies demand-side effects to some degree. Quantity-based TFP would be ideal to avoid this shortcoming, but its con- sistent computation requires observing quantities of outputs and inputs or their prices at the firm 31 Paths of productivity growth in Poland: a firm-level perspective level, and this information is not available in our dataset. Nevertheless, in this study, a large por- tion of the results are reported at an industry level that is either the same as or more aggregated than the level of price indices, which mitigates the bias due to unobserved firm-specific price issues. FIGURE B3.1  What Influences Firm Performance? Firm performance TFP measure used in the report Factors TFPR Deflated with industry-level price index of production (Revenue TFP) Captures not only productivity, but also whatever drives price variation Deflated with firm-level price index TFPQ Price Captures technical e ciency only (Physical TFP) Unavailable due to data limitation Market power Marginal costs Rents Quality Demand Given the substantial differences across sectors due to characteristics of pro- duction, market structure or surrounding business environment, the produc- tivity patterns were analyzed separately for manufacturing, construction, and services. This was done using disaggregated data and by breaking down the eco- nomic performance indicators at a level of aggregation that would not hide the het- erogeneity across different segments of the economy. Furthermore, given the dif- ferences in the data structure and coverage across sectors (significant variation of the data characteristics for construction services as compared to manufacturing and services), it was necessary to go beyond sector-level aggregation and analyze productivity at a more disaggregated level, for instance, for each 2-digit industry. Except for the initial two years in the sample, the TFP in Polish manufac- turing stagnated, while labor productivity followed an overall increasing trend. (See Figure 6) Even though the difference in the TFP levels between 2009 and 2019 amounted to 22 percent in total, as much as 20 percentage points of the 32 II. Properties of the Firm-Level Sample and Productivity Growth in Poland growth was realized between 2009 and 2011. Afterwards, there were no signif- icant productivity gains in Polish manufacturing — the difference between TFP levels in 2012 and 2019 amounts to only 1 percent. Labor productivity increased by 56 percent between 2009 and 2019, and again, the fastest pace was between 2009 and 2010, with a growth of 26 percent. In the next nine years, labor produc- tivity increased by 30 percent. While the labor productivity growth in manu- facturing and services followed similar upward trends, the TFP growth path in manufacturing coincided neither with that in construction nor that in services and fluctuated around zero most of the time (Figure 6). One possible reason for the discrepancy in the growth rates between the labor productivity and TFP in- dices is that the manufacturing sector adopted more capital- and technology-in- tensive production methods over time. Labor productivity may have increased as a result of capital deepening, meaning using capital more intensively in pro- duction. As a result, the amount of labor used in the production of one addition- al unit would decrease as time passes. However, TFP levels are not necessarily affected by this transformation in the production method, because TFP accounts for the increase in the intensity of capital used in production. FIGURE 6  Productivity Growth by Sectors (2009 – 19) a. Labor productivity b. TFP 20 20 15 15 Productivity change (%) Productivity change (%) 10 10 5 5 0 0 -5 -5 -10 -10 -15 -15 -20 -20 10 10 18 18 14 16 14 16 19 19 13 15 13 15 12 12 17 17 11 11 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 Manufacturing Construction Services Source: Elaboration based on Statistics Poland calculations. In Polish services, both labor productivity and TFP improved between 2009 and 2019 at a modest but positive rate of 4 percent per year — significantly fast- er than in manufacturing. TFP was more variable than labor productivity, ris- ing nearly 10 percent between 2016 and 2017 after five years of modest 2 percent 33 Paths of productivity growth in Poland: a firm-level perspective annual growth. Overall, there was a noticeable acceleration in the TFP growth in all sectors between 2015 and 2018. Yet, compared to manufacturing and con- struction, services experienced the most stable upward trend over the entire sample period. Nevertheless, the evolution of other sectoral outcomes — such as the growth paths of gross value added and labor (except for exports) — was comparable between services and manufacturing. (See Appendix Figure A2.2) The construction sector displayed the highest productivity variability, result- ing in a net moderate increase of both productivity indices between 2009 and 2019. The construction sector consists of only three industries: construction of buildings, civil engineering, and specialized construction activities. They are of roughly similar size. (See Appendix Table A2.1) Due to the previously described sectoral disruption, there was a significant drop in labor productivity and TFP between 2011 and 2012 that amounted to 16 percent and 8 percent respectively (Figure 6). This decline was led by the construction of buildings, which experi- enced more than a 40 percent decrease in labor productivity between 2011 and 2012. (See Appendix Figure A5.1) However, the labor productivity change is likely not a reflection of the actual sectoral situation but rather a statistical artifact because the exit rates of enterprises in the dataset during 2011 – 13 are unprec- edently high.10 There was a significant drop in the FDI inflows at the same time, which negatively affected the investments and gross value added. Both industrial and sectoral labor productivity grew considerably in 2013. The labor productiv- ity in construction caught up with the values before the downturn in just two years. The volatility in TFP originated mainly from the civil engineering indus- try, which consists of firms involved in building roads, railways, bridges, tun- nels, or city infrastructure. The civil engineering industry was predominantly affected by the market disruption following the increased construction demand due to the UEFA Euro 2012 infrastructure investments, as previously mentioned. In civil engineering, there was an almost 25 percent increase in TFP followed by a more than 30 percent drop between 2010 and 2013. (See Appendix Figure A5.1) Despite the volatility and possible sectoral disruption during 2011 – 13, the pro- ductivity in the entire construction sector, in terms of both gross value added per worker and TFP, began increasing again starting in 2013. The time paths of the two productivity indices coincide closely with each other, with the construc- tion sector having total labor productivity growth of 44 percent and total TFP growth of 39 percent between 2009 and 2019, both corresponding to an annual average 4 percent increase. 10.  The mean exit rate measured as an exit of an individual firm amounts to as high as 20 percent between 2011 and 2013. 34 II. Properties of the Firm-Level Sample and Productivity Growth in Poland Productivity Growth Patterns at the Industry Level The largest Polish manufacturing industries — manufacturing of food and bev- erages and manufacturing of basic metals and fabricated metal products — ex- perienced a drop in TFP between 2009 and 2019. The poor productivity perfor- mance of manufacturing was driven by the largest industries dragging down the sectoral TFP. As depicted in Figure 7, the manufacturing of food and beverages in- dustry and the manufacturing of basic and fabricated metals industry, which joint- ly produce 34 percent of gross value added and employ 33 percent of the sectoral la- bor force, experienced negative TFP growth. Two other manufacturing industries also had negative productivity performance: manufacturing of paper and paper products and manufacturing of chemicals and pharmaceuticals. These two indus- tries account for an additional 11 percent of gross value added in manufacturing. FIGURE 7  TFP Growth versus TFP Levels in the Manufacturing Sector, 2009 – 14 and 2015 – 19 a. Yearly average 2009−2014 b. Yearly average 2015−2019 14 14 12 12 Electronics 10 10 Apparel Productivity change (%) Productivity change (%) 8 8 Non−metallics Textiles Electronics 6 Print Other 6 Machinery Wood Furniture Other Repair 4 4 Furniture Plastics Machinery Repair Plastics Apparel 2 Vehicles 2 Vehicles Wood Paper Non− Metals Chemicals Textiles metallics 0 0 Food Print Metals Food -2 Chemicals -2 -4 -4 Paper -6 -6 2.0 2.5 3.0 3.5 4.0 4.5 2.0 2.5 3.0 3.5 4.0 4.5 Weighted TFP Weighted TFP Source: Elaboration based on Statistics Poland calculations. Note: An industry’s relative size represents the percentage of gross value added generated in that industry. In manufacturing industries, productivity trends differed considerably be- tween 2009 – 14 and 2015 – 19. As shown on the left-hand side of Figure 7, in the earlier period, some industries displayed outstanding productivity performance with an average yearly productivity growth close to 10 percent: manufacturing of computers and electronics and manufacturing of wearing apparel and leather. 35 Paths of productivity growth in Poland: a firm-level perspective The latter industry was undergoing microeconomic restructuring,11 also asso- ciated with a drop in employment amounting to 42 percent over the entire sam- ple period. The TFP increase in manufacturing of computers and electronics re- flects the fast-developing global IT technology sector. This industry experienced a remarkable real gross value added growth, amounting to more than 200 per- cent during the 11 years from 2009 to 2019, probably driven by the upsurge in demand for electronics (and the productivity advantages within companies; see Appendix Figure A5.2).12 However, between 2015 and 2019, high-performing in- dustries converged to the manufacturing average. Productivity stagnated in some service industries (such as IT13 and consulting services14) and in utilities,15 where state ownership is prevalent. Those stag- nant industries (Figure 8) represent 25 percent of sectoral gross value added and accommodate one-fifth of the services labor force. Among all the service indus- tries, only scientific research, advertising, market research, and veterinary16 had negative TFP growth between 2009 and 2019, with a striking 39 percent drop in TFP between 2015 and 2016. The drop in the productivity of research, advertising, and veterinary was most likely caused by the unprecedently high rate of low-productiv- ity firms entering the sample. (See Appendix Figure A5.2) The number of observa- tions increased by 37 percent, and the industrial labor force doubled. One possible reason for the expanded coverage of the firm-level data in the research industry is the introduction of R&D tax relief from 2016 onwards (Laplante et al., 2019). Almost 40 percent of companies in the research industry report R&D expenditures, and the number of companies performing R&D doubled between 2015 and 2016. While a nearly all service industries exhibited similar productivity growth, telecommunications was different. In telecommunications — a highly regu- lated industry including, for instance, Internet providers and mobile network operators — TFP increased by 86 percent in 11 years, a far higher growth than the sectoral average. (See Table 2 and Appendix Table A2.1) In addition, labor 11.  Since the economic transformation, there has been a systematic decline of employment in the Polish wearing apparel industry (Lewandowski and Magda, 2014). Firms transformed their pro- duction process toward less labor-intensive methods, and the high-productivity establishments in- creased their market share. 12.  For more on the relations between demand shocks and productivity, see Mayer, Melitz, and Ottaviano (2014, 2016) and Cusolito, Fernandes, and Maemir (2018). 13.  Computer programming, information service activities, consultancy, and related activities 14.  Legal and accounting activities, management, architectural and engineering 15.  Electricity, gas, water, sewerage, waste, etc. 16.  Section M of NACE Rev. 2 36 II. Properties of the Firm-Level Sample and Productivity Growth in Poland FIGURE 8  TFP Growth versus TFP Levels in the Services Sector, 2009 – 14 and 2015 – 19 a. Yearly average 2009−2014 b. Yearly average 2015−2019 16 16 14 Telecommunications 14 12 12 Productivity change (%) Productivity change (%) 10 10 8 8 Support 6 Accomodation 6 Media Food Vehicles Retail Real estate 4 Wholesale 4 Vehicles Real estate Telecom. Utilities Transport Retail Support Accomodation IT Wholesale 2 2 Water Consulting Transport Research Food 0 0 Water Consulting Utilities -2 IT -2 Media Research -4 -4 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Weighted TFP Weighted TFP Source: Elaboration based on Statistics Poland calculations. Note: An industry’s relative size represents the percentage of gross value added generated in that industry. productivity growth was 91 percent in the telecommunications industry. These productivity performance increases were associated with a 24 percent drop in employment and an 81 percent increase in gross value added. A further empiri- cal investigation of the industry (details provided in Chapter 3) reveals that most of the productivity growth in telecommunications came from firms improving their capabilities, for instance through successful digitalization and innovation. (See Appendix Figure A5.2, Panel c.) At the same time, more productive telecom- munications companies increased their market share within the industry. This provides some evidence that less productive establishments in telecommunica- tions decreased their number of employees (possibly, for instance, due to lack of digitalization in some companies) so that high-productivity growth occurred simultaneously with shrinking industry employment. Manufacturing had higher TFP dispersion than construction or services, and industries with high dispersion tended to have low TFP. An industry may ex- hibit high productivity dispersion when, for instance, low-productivity compa- nies (firms that lag the industry and sectoral TFP) have a considerable share in the industry. Such dominance of inefficient firms might be a result of barriers to firm entry and exit or a rigid labor market, either of which would hinder re- source allocation toward more productive firms. Firms being unable to adjust 37 Paths of productivity growth in Poland: a firm-level perspective their workforces easily (due to regulations or labor market conditions) will in- evitably lead to higher dispersion. Another cause of high TFP dispersion can be highly productive firms that are in the process of technology adoption or are un- dergoing risky R&D activities and innovations. These firms would contribute to the growth in aggregate outcomes (Doraszelski and Jaumandreu, 2013). Moreover, due to data limitations, the TFP measure employed in the study captures not only efficiency in production but also price and quality effect. (See Box 3 for more de- tails.) High dispersion might signify substantial price and quality differences between firms. Figure 9 exhibits the productivity dispersion (calculated as the ratio of the 80th percentile of productivity to the 20th percentile of productivity) for industries in the manufacturing, services, and construction sectors during 2009 – 19. The figure demonstrates that there was a negative link between pro- ductivity dispersion and TFP in manufacturing. (That is, the higher the produc- tivity dispersion within a manufacturing industry, the lower the TFP levels in that industry tended to be.) This association provides some evidence for the pres- ence of a significant number of inefficient firms in the high-dispersion indus- tries. Therefore, the following sections pay particular attention to how market shares are allocated across producers with different productivity levels and fur- ther analyze the patterns in resource allocation by decomposing productivity. FIGURE 9  Firm-Level Productivity Dispersion by Sector, 2009 – 19 a. Labor productivity b. TFP 1.9 1.9 1.8 1.8 Productivity dispersion Productivity dispersion 1.7 1.7 1.6 1.6 1.5 1.5 1.4 1.4 1.3 1.3 1.2 1.2 1.1 1.1 2 3 4 5 6 2 3 4 5 6 Manufacturing Manufacturing (fitted values) Construction Construction (fitted values) Market services Market services (fitted values) Source: Elaboration based on Statistics Poland calculations. Note: Dispersion is calculated as the ratio of weighted productivity in the 80th percentile to weighted productivity in the 20th percentile within an industry (as in the TFP estimation). The dots represent the result for each industry in each year. 38 II. Properties of the Firm-Level Sample and Productivity Growth in Poland Sectors that have higher revenue growth also had higher TFP growth on aver- age. Figure 10 compares the TFP of industries in a given year and the annual real growth rates of their revenues in that year. For each sector, there was a positive correlation between real revenue growth and TFP growth. This positive correla- tion implies that industries that expanded (in terms of total revenues) tended to have higher TFP growth. Conversely, industries with decreasing total revenues tended to experience productivity declines. Figure 10 provides a dynamic per- spective, meaning that it depicts the relationship between growth in market size and TFP growth levels across years. Some large industries had negative TFP growth rates in earlier years, and their size contracted over subsequent years. Such industries underly the positive correlation between the growth rates of TFP and market size. In Poland, some of the large traditional sectors (for exam- ple, manufacturing of food and beverages or utility services) were shrinking while more productive industries were growing in size, indicating that micro- economic restructuring has been occurring in a productivity-enhancing way. FIGURE 10  Growth in Market Size versus TFP Growth 40 30 20 Growth in market size (%) 10 0 -10 -20 -30 -40 -50 -40 -30 -20 -10 0 10 20 30 40 50 Productivity (TFP) growth (%) Manufacturing Manufacturing (fitted values) Construction Construction (fitted values) Market services Market services (fitted values) Source: Elaboration based on Statistics Poland calculations. Note: Each dot represents the combination of TFP and annual real revenue growth rates for a specific industry and year. 39 III WHAT DRIVES POLISH FIRM-LEVEL PRODUCTIVITY GROWTH? 1 To understand the underlying response behind the observed productivity growth in Polish sectors and industries, we decompose aggregate productivity growth using the Melitz-Polanec decomposition method into four components: within (innovation or adaptation), between (more efficient resource allocation), upscal- ing (entry of more productive producers), and downscaling (exit of less successful establishments). 2 The manufacturing sector’s productivity stagnation between 2012 and 2019 resulted mainly from a deterioration in the allocative efficiency in the largest manufacturing industries: metals and food and beverages. During that period, the low-productivity firms in these industries grew and expanded their market share at the cost of more productive firms. Consequently, the between component neg- atively contributed to productivity performance and pulled down the aggregate productivity of the entire manufacturing sector. This negative pattern changed the long-observed (since 1997) trend of the between component driving the TFP growth in Poland’s manufacturing. 3 Poland’s private non-financial service sector exhibited continuous growth in TFP during the sample period that was on average higher than that in manufacturing. In services, resource allocation also improved continuously and dominated the within contribution during 2012 – 16. In the later period of the sample, the within compo- nent rose above the between component and became the main driver of productiv- ity growth in services, much like in manufacturing and construction after 2017. 4 Only two industries substantially contributed to sectoral productivity growth: telecommunications and manufacturing of computers and electronics. These two industries can attribute their outstanding productivity performance to the within component: advances in firm capabilities, such as through innovation, technology adoption, increasing managerial and organizational practices, or improvements in human capital skills. 41 Paths of productivity growth in Poland: a firm-level perspective Decomposition Methodology Aggregate productivity can grow as a result of firms increasing their capabil- ities (within-firm productivity growth), allocating resources in more produc- tive firms and facilitating high-productivity firms to increase their market share (between-firm productivity growth) as well as productive firm entry and exit. Employing firm-level panel data enables breaking down the productiv- ity growth into four components, each representing a different source of produc- tivity. This method is called Melitz-Polanec decomposition (Melitz and Polanec, 2015). Tracking the performance of an individual firm over time allows analyz- ing not only the within-firm productivity improvements but also the pattern of market share reallocations across firms and its consequences for aggregate pro- ductivity. Box 4 and Box 5 provide methodological details and linkages between decomposition components and relevant policy responses. BOX 4  How to Decompose Productivity Productivity measured with firm-level panel data can be broken down into four components, each representing a different source of productivity. To distinguish sources of Polish productivity growth during 2009 – 2019, we employ the dynamic Olley-Pakes decomposition, also known as Melitz-Polanec decomposition (Melitz and Polanec, 2015). The method relies on tracking the performance of indi- vidual firms over time to analyse the pattern of market share reallocations across firms and its con- sequences for aggregate productivity. Let’s define the aggregate productivity (θ) in two successive periods (θ1 and θ2) with: θ1 = SS1θS1 + SX1θX1 = θS1 + SX1 (θX1 – θS1) θ2 = SS2θS2 + SE2θE2 = θS2 + SE2 (θE2 – θS2), where S is the market share of: firms surviving between periods (survivors, SS1 SS2), firms exiting the sample or downscaling (exiters, SX1), and firms entering the sample or upscaling (entrants, SE2). The difference in productivity between two periods is then: ∆θ = (θS2 – θS1) + SE2 (θE2 – θS2) + SX1 (θS1 – θX1) survivors entrants exiters Hence, the productivity is decomposed into contributions from three groups of firms: survivors, entrants and exiters. The survivor’s contribution is further separated using Olley-Pakes (Olley and Pakes, 1996) into two: 1) the unweighted mean change in the productivity of survivors and 2) the 42 III. What Drives Polish Firm-Level Productivity Growth? covariance change between market share and productivity for survivors. Finally, the productivity change can be expressed as: ∆θ = ∆θS + ∆cov(θ,s) + SE2 (θE2, – θS2 ) + SX1 (θS1 – θX1) within covariance entry exit The aggregate productivity is decomposed into four components: (1) the change in the mean (within component), (2) the change in the covariance (between component), (3) entry or upscaling, and (4) exit or downscaling. The within component measures the gains from firms’ own productivity performance. It represents a shift in the distribution of firm productivity. The between component represents the productivity growth coming from the reallocation of resources across producers. The upscaling and downscaling components measure productivity gains from entering and exiting firms. If a firm enters a market any time in the period between t and t+1 and displays a productivity performance above the industry average as of time t, this is reflected as a positive contribution to the aggregate pro- ductivity growth. Similarly, if a firm whose productivity is lower than the industry average at time t exits a market until time t+1, this is reflected as a positive productivity contribution. However, cap- turing the contributions of firm entry and exit robustly depends heavily on the quality of the data, particularly the reliability of the panel structure. In the Melitz-Polanec decomposition, the entry and exit components must be computed based on the absence or presence of data rather than the actual information on firm entry and exit. Moreover, given that we observe only companies with more than nine employees, if a firm in a specific year drops to fewer than ten employees, that com- pany will be detected as an exiter, while in fact it still operates in the market. That is why we call the dynamic components upscaling and downscaling. Consequently, if micro firms are not included, the significant entry/exit productivity contribution reflects to a larger extent changes of firm size around the threshold rather than actual establishment of new firms and closures of existing companies. BOX 5  Productivity-boosting policies Boosting productivity requires policy actions addressing all components of productivity growth. The figure below presents an exemplary set of policy interventions and programs broken down into the three productivity growth components with which they are typically associated. (For these pur- poses, the upscaling and downscaling components are combined into a single “dynamic” compo- nent.) It is important to recognize that, although conceptually distinct and driven by different types of firm dynamics, these three components are closely interlinked. For instance, barriers to reallo- cation that dampen the between component (market distortions such as poorly designed institu- tions, excessive regulation, or the disproportionate presence of state-owned enterprises) can also discourage investment in innovation (within component) or disincentivize low-productivity firms from exiting (dynamic component). At the same time, without innovation and technology adoption (within component), there would be limited space for reallocation (between component) or selec- tion (dynamic component). 43 Paths of productivity growth in Poland: a firm-level perspective TFP growth Within growth Between growth Dynamic growth Increasing firm capabilities Allocating resources to Entry of high-productivity more productive firms firms and exit of low-pro- ductity firms Promoting innovation through incentivizing R&D investments Allowing markets to work e - (matching grants or fiscal incen- ciently, reducing distortions and Removing barriers to entry and tives such as loans). disincentives to grow for more experimentation, reducing the productive firms, targeting the risk of failure by making exit and elimination of labor market fric- re-entry low cost (digitalization Enhancing technology adop- tions, removing industrial pro- and simplification of require- tion not only by financial incen- tection, addressing market fail- ments for starting up business- tives but also by strengthening ures such as information asym- es, simplifing licensing, imposing managerial capabilities and the metries or moral hazard (finan- limited bureaucratic burden on human factor (filling the knowl- cial market deepening). low-risk activities, simplifing re- edge gap, working on peer net- gimes of bankruptcy or restruc- works, outsourcing technology turing. adoption consultants, account- Strengthening competition, im- ing for loss aversion associated proving investment climate and with new technology adoption). business environment. Lowering risks of innovative products or the market risk as- sociated with novel business Upgrading organizational and Strengthening linkages with for- models (including fiscal or mon- management practices in firms eign markets, reducing transac- etary policies, financial sector by improving education quality, tion costs for firms to integrate policies, protection of property supplying digital and technical with larger markets (trade facil- rights, etc.). skills (hiring professional man- itation, certifying products for agers). sophisticated export markets, matchmaking). Encouraging entrepreneurship (entrepreneurial training). Sources: Cirera and Maloney (2017); Cusolito and Maloney (2018). Productivity Growth Decomposition in Manufacturing In manufacturing, the efficiency of resource allocation across firms worsened over time, which slowed the sector’s productivity growth. The between com- ponent that represents the productivity growth from the reallocation of produc- tion factors across manufacturers within industries was negative for most years, especially from 2014 until 2019 (Figure 11). Decomposition at the industry level reveals that large low-productivity manufacturers of food, beverages, metals, and rubber increased their market share over time at the cost of more produc- tive companies, negatively contributing to the manufacturing sector’s aggregate productivity performance. (See Appendix Figure A5.2 and Box 6) At the same time, manufacturing displayed growth in terms of gross value added as well as labor. (See Appendix Figure A2.2) The fact that this high-growth period of value added and labor corresponds to substantial drops in the between component 44 III. What Drives Polish Firm-Level Productivity Growth? provides some evidence that fast-growing manufacturing firms had low TFP lev- els initially. This stems from the fact that the between component reflects the part of the increase in the sector’s size that comes from the expansion of incum- bents but not the part that comes from the entry of new firms (which is reflected in the upscaling component). FIGURE 11  Manufacturing Sector Productivity Growth Decomposition 7 6 5 4 3 TFP growth (%) 2 1 0 -1 -2 -3 -4 2012 2013 2014 2015 2016 2017 2018 2019 Within Between Upscaling Downscaling Net Source: Elaboration based on Statistics Poland calculations. Note: The figure shows the results of decomposing 3-year productivity growth rates using the Melitz-Polanec method, smoothed to represent an annual change. Worsening efficiency of resource allocation in manufacturing means break- ing the long-observed trend of between component driving the TFP improve- ments in Poland. The fast TFP growth between 1997 and 2013 was a result of more productive firms gaining market share at the expense of less productive firms shrinking (World Bank, 2017). However, as noted in the World Bank’s report, in 2012 the selection mechanism was interrupted and the between component turned negative. This was primarily driven by relatively more productive firms in the electronics, metals and beverages industries losing their market share. The decomposition results applied to 2009 – 19 data confirmed that worrying shift in trend. Moreover, our study also detected the efficiency deterioration in resource allocation in the biggest manufacturing industries in Poland — metals, food, and beverages. (See Box 6.) It means that as the years go by, the situation in those industries hasn’t improved for almost a decade, and still, more produc- tive firms are losing their market share. 45 Paths of productivity growth in Poland: a firm-level perspective BOX 6  Evidence from Poland’s Food Industry Since 2012 the efficiency of resource allocation in Poland’s food industry worsened over time, which significantly slowed the manufacturing’s productivity growth. (See Appendix Figure 18 and Appendix Table 8.) The food industry is one of the biggest parts of Poland’s economy, employing in 2019 18% of the manufacturing’s labor force and generating 16% of the sectoral gross value added. Worsening resource allocation in the industry means that since 2012 more productive food producers have been losing their market share and at the same time, less productive establishments were growing. World Bank’s report Poland Catching-Up Regions 2: Safer Food, Better Business in Podkarpackie and Lubelskie (World Bank, 2018) indicates that one of the potential reasons behind that is the enforce- ment of food-related requirements that could be excessively burdensome to firms, hamper entry, limit competition and stall growth (typical factors negatively contributing to the between compo- nent of productivity growth, see Box 5). Firms in the food industry need to comply with demanding EU- and nation-wide requirements. These requirements help to minimize negative externalities, such as the spread of food-borne illnesses. They are enforced mainly through business inspections carried out by five inspectorates respon- sible for food control in Poland. However, firms struggle to understand requirements and perceive inspections mainly as a burden. Poland’s authorities have many institutional practices in place to reduce the inspection burden and improve compliance with regulations in the country’s southeast. The fragmentation of inspections is partly addressed through multi-annual control plans and inter- institutional cooperation agreements. Risk assessment methodologies are used to determine how frequently businesses should be inspected. Inspectorates use checklists to harmonize inspections and occasionally provide guidance to businesses. However, Poland’s authorities can still learn from international best practices in regulatory enforcement to reduce administrative burden and pro- vide firms with more information about requirements. Improvements should focus on strengthen- ing coordination between inspectorates, refining and fully implementing risk assessment, and scal- ing up compliance promotion. The manufacturing industries with the fastest TFP growth can credit their pro- ductivity performance to the within component, reflecting firms’ own produc- tivity improvements, for instance, through innovation, better managerial prac- tices, or adopting new technologies. Manufacturing of machinery and equipment, computers and electronics, wearing apparel and leather, and furniture exhibited fast TFP growth in Poland. (See Appendix Table A2.1) The decomposition indicates that this growth performance was primarily driven by the within component re- flecting firms’ increasing performance capabilities (Figure A5.2, panel a). This com- ponent was on average positive in all but one industry: paper and paper products. However, the productivity gains in the other industries were insufficient to offset the misallocation of productive resources (the between component — that is, increas- ing market share of low-productive firms in the industry) that caused the produc- tivity in manufacturing to stagnate from 2013 to 2017. Starting in 2017, TFP growth was positive due to the upscaling and within components. This means not only that existing firms were getting more productive but also that new firms with high pro- ductivity were entering the dataset (upscaling from micro to at least small-sized). 46 III. What Drives Polish Firm-Level Productivity Growth? There was no significant productivity contribution from the net of upscal- ing and downscaling components in the manufacturing sector.17 As shown in Figure 11, the upscaling contribution was positive and slightly larger than the downscaling component, which was negative. One possible reason for the small contribution of upscaling is that manufacturing entrants tend to have larger sunk investments and other fixed start-up costs, which causes their post-entry performance to be slower than that of the firms in other sectors (Brouthers and Brouthers, 2003). Given our sample, we cannot determine whether the net entry effect resulted from the exit of unproductive firms from the market or firms lowering headcount and dropping out of the SP survey. (As mentioned earlier, the decomposition method relies on the absence of the data to detect the exit of a firm, and micro firms are not obliged to fill out the SP survey.) Therefore, the available evidence is insufficient to determine whether the market selection pro- cess has been functioning effectively in manufacturing. Productivity Growth Decomposition in Construction Until 2018, Poland’s construction industry profited from more efficient alloca- tion of resources between firms, but afterward, its TFP growth was primarily driven by productivity improvements within firms. The within and between productivity components in construction followed divergent paths (Figure 12). The between component was positive and relatively high until 2018, while the within was generally negative. Starting from 2017, the situation reversed: the within component remained positive, while the between component was nega- tive for the last two years. This suggests that Polish construction received a pos- itive productivity shock in 2017 that increased the within component and the aggregate productivity growth performance of the sector. Firms that benefited from this positive shock, however, did not necessarily have large market shares, so the between component decreased by the end of the sample. 17.  This study implements Melitz-Polanec decomposition from a 3-year window, but the decompo- sition can be applied to annual productivity growth rates, which may result in negligibly small or negative entry contributions. Even equipped with the latest technology and skills, entrant firms generally need a start-up period to learn the market demand, advertise their products, and exploit their productivity advantage. This start-up period is usually longer than one year, so decomposing annual growth may not capture the real contribution of entrants in the longer term. Therefore, in the study, the Melitz-Polanec decomposition is applied to 3-year productivity growth rates (to cap- ture entrants’ productivity contribution after their first year in the market) and then divided by three (to represent the annual dynamics). Melitz-Polanec decomposition was also applied to 5-year productivity growth rates. The results based on the 5-year differencing do not differ significantly from those based on 3-year differencing. 47 Paths of productivity growth in Poland: a firm-level perspective FIGURE 12  Construction Sector Productivity Growth Decomposition 12 10 8 6 TFP growth (%) 4 2 0 -2 -4 -6 -8 2012 2013 2014 2015 2016 2017 2018 2019 Within Between Upscaling Downscaling Net Source: Elaboration based on Statistics Poland calculations. Note: The figure shows the results of decomposing 3-year productivity growth rates using the Melitz-Polanec method, smoothed to represent an annual change. The construction industry in Poland has a very dynamic structure with high rates of firm entry and exit from the dataset. As mentioned previously, these entry and exit rates do not necessarily reflect newborn or dying firms. Rather, they may reflect project-specific reorganization of the companies that generates gaps in the panel structure of the dataset, which does not capture micro firms. Accounting for allocative efficiency gains, however, is not straightforward with an unbalanced panel structure, and unrealistically large fluctuations may oc- cur as a result. Nevertheless, as mentioned previously, the construction sector experienced a disruption in 2010 – 13 due to an unprecedented demand boom re- lated to the UEFA Euro 2012 and projects co-financed by EU structural funds. The large negative downscaling contribution in 2014 that is depicted in Figure A2.1 might reflect this negative shock and the bankruptcy wave that followed. It is also likely connected to Poland’s lower net inflow of FDI. (See Appendix Figure A7.1) Productivity Growth Decomposition in Services As in manufacturing and construction, the within component in services contributed substantially to aggregate productivity growth, especially for the later period in the sample (2016 – 2019), but unlike in other sectors, the effi- ciency of resource allocation improved continuously in services. The service sector in Poland had uninterrupted positive productivity growth throughout 48 III. What Drives Polish Firm-Level Productivity Growth? the sample period, with an increase in the growth rate after 2016 (Figure 13). In the earlier years, much of the growth came from the entry of new producers to the dataset (upscaling component) and efficiency in resource allocation (between component). After 2016, however, firms’ own productivity performance (within component) was the main driver of the accelerated productivity performance. The only negative productivity contribution came from the downscaling com- ponent, indicating that some high-productivity firms exited the dataset, if not the market. The negative exit contribution, however, may be due to gaps in the data for some high-productivity firms. When a data point is missing for such a high-productivity firm, there are two consequences. First, its absence in a given year will be reflected as a negative contribution to the downscaling component for that year. Second, the upscaling component will rise in the next year, when the missing firm is observed again in the sample. Thus, the simultaneously nega- tive downscaling and positive upscaling components in the service sector’s pro- ductivity decomposition may be mainly due to the unbalanced structure of the sample rather than to actual firm entries and exits. FIGURE 13  Services Sector Productivity Growth Decomposition 7 6 5 TFP growth (%) 4 3 2 1 0 -1 -2 2012 2013 2014 2015 2016 2017 2018 2019 Within Between Upscaling Downscaling Net Source: Elaboration based on Statistics Poland calculations. Note: The figure shows the results of decomposing 3-year productivity growth rates using the Melitz-Polanec method, smoothed to represent an annual change. The Polish services industries with the fastest-growing TFP owed their pro- ductivity performance to the within component. Among services indus- tries in Poland, telecommunications had by far the fastest-growing TFP. (See Appendix Table A2.1) Although its outstanding growth performance was pri- marily driven by the within component, the allocative efficiency also improved considerably. (See Annex Figure A5.2). The within component was on average positive in almost all services industries. The best-performing industries in the 49 Paths of productivity growth in Poland: a firm-level perspective sector — accommodation, trade, real estate, and administration18 — can credit their productivity performance to within-firm improvements and more effi- cient factor allocation. Among all the service industries in the firm-level sam- ple, only professional, scientific, and technical activities19 had, on average, neg- ative TFP growth between 2009 and 2019. The services industries with the largest shares of state-owned enterprises had below-average between components. That is, they experienced worse than sectoral average allocation of the factors of production. Utilities20 and publish- ing21 had the largest shares of firms in the dataset with a majority of public cap- ital: over 60 percent in utilities such as electricity, gas, and steam, 55 percent in water utilities, and 8 percent in publishing. The efficiency in the allocation of resources in those industries was the worst among all services. Moreover, the productivity performances of utilities and publishing did not change signifi- cantly over the entire 11-year period. It is worth mentioning that, apart from util- ities and publishing, the share of state-owned enterprises in most industries in the service sector is less than 2 percent. In the entire firm-level dataset, the firms controlled by the state amounted to 5 percent. 18.  Rental, leasing, employment activities, travel agencies, office administration, office supporting activities, and security. 19.  Section M of NACE Rev. 2 20.  Electricity, gas, water, sewerage, waste, etc. 21.  Publishing activities, television production, programming, and broadcasting activities. 50 IV HETEROGENEITY IN PRODUCTIVITY PERFORMANCE AND ITS DETERMINANTS 1 In the last decade, small and medium Polish enterprises were the engines of pro- ductivity growth in Poland, displaying faster productivity performance than large firms. Moreover, there is some weak evidence that companies with foreign owner- ship status were more successful in terms of productivity growth than establish- ments with domestic or state ownership status. 2 Firms operating in expanding industries, such as the manufacturing of computers and electronics, vehicles and equipment, as well as IT services, had better produc- tivity performance than other establishments over time. While this can be inter- preted as better-performing firms driving size growth in the industry, increasing demand for an industry’s products can also motivate firms to be more productive. 3 Performing R&D activities improved firms’ market share within the industry (between component), but there is no evidence that it also led to firms’ own pro- ductivity improvements (within component). This may imply that R&D invest- ments help more productive firms to capture larger market shares, which in turn improves allocative efficiency. Also, R&D’s explanatory power with respect to pro- ductivity growth differs between sectors: R&D had a positive impact on TFP in the manufacturing sector but was insignificant in services and construction. 53 Paths of productivity growth in Poland: a firm-level perspective Regression Approach Because different sectors have different characteristics, analyzing the het- erogeneity of productivity growth is needed to understand the paths it can take. As shown in Table 3, the firm-level sample used in this analysis includes firms from every size (except for micro), owner, and age group. Even though large firms constitute only 5 percent of the companies in the sample, they hire more than half of the entire labor force (which corresponds to the shares observed in the aggregate economy). Small, medium, and large enterprises account for an equal share of the labor force in the construction sector, while large compa- nies account for over 50 percent of employment in the manufacturing and ser- vices sectors. Even though small firms are universally prevalent in each sector, they employ less than one-fifth of the labor force. Sixty-one percent of the con- struction companies are owned by Polish entities, while foreign ownership is most common in manufacturing, having 38 percent of all manufacturing work- ers. Over two-thirds of companies in the dataset are well-established entities, operating for more than a decade on the market and firms’ maturity patterns are similar across sectors. TABLE 3  Firm Characteristics by Sector, Average for 2009 – 19   Firm share by sector (%) Labor share by sector (%)   Manufacturing Manufacturing Construction Construction Services Services Sample Sample Size classes  a Small (10 – 49) 74 65 81 78 19 14 36 21 Medium (50 – 249) 21 28 17 18 29 32 37 26 Large (250+) 5 7 2 4 51 54 28 53 Ownership classes  b Private domestic (PDE) 44 41 48 44 41 44 56 37 State-owned (SOE)  c 4 1 2 5 10 2 5 16 Foreign-owned (FOE)  d 10 13 3 10 31 38 14 28 Dispersed capital  e <1 <1 <1 <1 2 2 1 2 Undefined  f 42 45 47 40 17 14 24 18 54 IV. Heterogeneity in Productivity Performance and Its Determinants   Firm share by sector (%) Labor share by sector (%)   Manufacturing Manufacturing Construction Construction Services Services Sample Sample Age classes Less than 3 years 9 8 10 9 9 8 9 9 From 4 to 9 years 22 21 22 23 24 23 25 25 Older than 10 years 69 71 68 68 67 69 66 66 Source: Elaboration based on Statistics Poland data. Note: Percentages sum to 100 within firm classes and sectors. FOE = foreign-owned enterprise; PDE = private domestic en- terprise; SOE = state-owned enterprise. a. Three categories based on the annual average number of employees in full-time equivalents. b. Five categories based on the source of the largest share in core capital. The classification is contingent on data avail- ability in Statistics Poland but largely follows Eurostat’s guidelines.22 c. State-owned enterprise, SOE = enterprise with a state as the largest shareholder.23 d. Foreign-owned enterprise, FOE = enterprise with a foreign entity as the majority shareholder. e. Enterprise in which dispersed capital constitutes the majority. f. Enterprise not reporting ownership.24 Productivity levels and their dynamics are expected to differ across firm siz- es, ownership status, and years on the market. We employ fixed-effects regres- sions to derive implications for the relationship between firm groups, competition, and firm productivity performance. In the fixed-effects model, every variable is transformed by subtracting the variable’s mean, and the estimation is done on so-called time-demeaned data. In our case, we investigate the effect of firms’ fea- tures (size, age, ownership) and competition on TFP growth rates. The first lags of the size dummies are introduced into the firm performance regressions along with 2-digit industry, year, and region fixed effects. In this setup, controlling for 22. https://ec.europa.eu/eurostat/web/products-manuals-and-guidelines/-/ks-ra-12-016. 23.  The state-owned enterprises constitute 4 percent of firms in our sample, and they are mostly concentrated in utilities (60 percent of firms in the industry are state-owned) and real-estate activi- ties (14 percent of firms in the industry are state-owned). They generate 6 percent of aggregate reve- nue and employ 10 percent of the labor force (16 percent in services). 24.  Due to poor reporting, more than 40 percent of companies in the dataset prepared by Statistics Poland do not disclose their ownership. Those companies are mostly small entities (76 percent of those firms hire between 10 and 49 employees, 13 percent firms are micro firms, and the remaining 11 percent are medium firms) engaged in manufacturing (34 percent of firms with undefined owner- ship are manufacturers), construction (14 percent), and retail and trade activities (32 percent). Even though firms with undefined ownership constitute more than 42 percent of all firms in the sample, they hire only 17 percent of the labor force and generate 12 percent of aggregate revenue. 55 Paths of productivity growth in Poland: a firm-level perspective industry fixed effects is particularly important due to the unobserved differenc- es in industry-level conditions (e.g., Cohen and Klepper, 1996). The small, young, and domestic companies are set as the benchmarks so that the corresponding size dummies are omitted in the regressions. However, due to poor ownership report- ing (40 percent of companies do not disclose their ownership status), the analy- sis of the relationship between firms’ ownership and TFP growth is not straight- forward, so we employ two regression strategies. First, we divide all companies into two categories: foreign (enterprise with a foreign entity as the majority share- holder) and the rest. The results are given separately for manufacturing (Table 4) and construction and services (Table 5). The second approach assumes running regressions with ownership dummies, as they were defined in the firm-level dataset: private-owned domestic, state-owned, foreign-owned, dispersed capi- tal and undefined (see Table 3 for details on size classes and Appendix Table A6.1, Panel b and Table A6.2, Panel b for regression results). Moreover, since the size dummy might capture all the productivity variation of the state-owned and for- eign-owned companies, we present additional two sets of the results with own- ership and size dummies introduced one at a time.25 Those estimation results are given in the Appendix. (See Table A6.1 for manufacturing and Table A6.2 for con- struction and services.) The regression results investigating the link between firm’s age and productivity performance are also given separately and for manu- facturing (Table A6.3, Panel a) and construction and services (TableA6.3, Panel b). TABLE 4  Fixed-Effects Regression of TFP on Competition Indicators by Characteristics of Firms in the Manufacturing Industry, 2009 – 19 Dependent variables TFP Specifications (1) (2) (3) Competition 0.0615*** 0.0718*** Number of firms (log) (0.0222) (0.0217) 0.657*** Industry-level profit margin (0.0886) 2.930*** 3.071*** 3.152*** Concentration Index (0.287) (0.289) (0.281) 25.  Because, for instance, state-owned and foreign-owned enterprises are usually also large, intro- ducing size and ownership dummies at the same time might make the results hard to interpret. Even though the majority of results do not considerably differ when the dummies are introduced simul- taneously or one at a time, the link between foreign-owned firms and productivity growth is uni- versally positive and statistically significant across sectors when the ownership dummies are intro- duced separately and for more detailed size classes’ definition (See Appendix Table A6.1 and Table A6.2). Hence, we find a weak positive relationship between foreign ownership status and TFP growth. 56 IV. Heterogeneity in Productivity Performance and Its Determinants Dependent variables TFP Specifications (1) (2) (3) Firm size (reference group: small firms) 0.0635*** 0.0638*** 0.0628*** Micro-small  a (0.009) (0.009) (0.009) -0.0934*** -0.0941*** -0.0935*** Medium [50 – 250) (0.008) (0.008) (0.008) -0.134*** -0.135*** -0.135*** Large [250+) (0.0135) (0.0136) (0.0133) Ownership (reference group: all except foreign-owned)  b -0.0171 -0.0172 -0.00774 Foreign-owned (FOE) (0.0248) (0.0248) (0.0229) 2.547*** 2.033*** 1.933*** Constant (0.154) (0.239) (0.234) Observations 168,162 168,162 168,162 R-squared 0.059 0.059 0.107 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes Source: Statistics Poland calculations. Note: Small firms and private domestic firms are considered the reference groups in their categories. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. a. Recall that micro firms are excluded from the sample. However, the dataset prepared by Statistics Poland (based on the SP surveys) includes several companies employing more than nine people not working full-time. Firm sizes are defined by full-time employment. Hence, when recalculating for full-time engagement, there appears a small group of micro firms in the dataset, namely 7 percent of all observations (1 percent of the labor force), employing most often 7–9 full-time workers. In the economy, the percentage of micro firms approaches 97 percent. In the production function estimation and produc- tivity analysis, those special-case micro firms were classified as small firms. Those companies appear to be the most pro- ductive in the sample (universally in all sectors), which is likely a statistical artifact. b. The regression’s results with ownership dummies as defined in the SP survey are given in Appendix Table 11, Panel b. TABLE 5  Fixed-Effects Regression of TFP on Competition Indicators by Characteristics of Firms in the Construction and Services Industries, 2009 – 19 Dependent variables TFP Specifications (1) (2) (3) Competition -0.0230** -0.0231** Number of firms (log) (0.0104) (0.0104) 0.000807 Industry-level profit margin (0.00114) -0.0776* -0.0814* -0.0813* Concentration Index (0.0456) (0.0456) (0.0456) 57 Paths of productivity growth in Poland: a firm-level perspective Dependent variables TFP Specifications (1) (2) (3) Firm size (reference group: small firms) a 0.0807*** 0.0807*** 0.0807*** Micro-small  (0.00648) (0.00648) (0.00648) -0.0740*** -0.0737*** -0.0737*** Medium [50 – 250) (0.00572) (0.00572) (0.00572) -0.126*** -0.126*** -0.126*** Large [250+) (0.0135) (0.0135) (0.0135) Ownership (reference group: all except foreign-owned)  b 0.0239 0.0237 0.0237 Foreign-owned (FOE) (0.0179) (0.0179) (0.0179) 2.323*** 2.479*** 2.479*** Constant (0.0719) (0.100) (0.100) Observations 353,276 353,276 353,276 R-squared 0.069 0.069 0.069 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes Source: Statistics Poland calculations. Note: Small firms and private domestic firms are considered the reference groups in their categories. Robust standard er- rors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. a. Recall that micro firms are excluded from the sample. However, the dataset prepared by Statistics Poland (based on the SP surveys) includes several companies employing more than nine people not working full-time. Firm sizes are defined by full-time employment. Hence, when recalculating for full-time engagement, there appears a small group of micro firms in the dataset, namely 7 percent of all observations (1 percent of the labor force), employing most often 7–9 full-time workers. In the economy, the percentage of micro firms approaches 97 percent. In the production function estimation and produc- tivity analysis, those special-case micro firms were classified as small firms. Those companies appear to be the most pro- ductive in the sample (universally in all sectors), which is likely a statistical artifact. b. The regression’s results with ownership dummies as defined in the SP survey are given in Appendix Table 12, Panel b. There is a strong body of empirical evidence that competition drives produc- tivity — by motivating firms to innovate and develop better products, ensur- ing that more productive firms increase their market share, and by placing pressure on the managers to become more efficient.26 To capture the link be- tween competition and TFP for Polish enterprises, we introduce an index of mar- ket concentration into the analysis. The Hirschman-Herfindahl index (HHI) is a standard measure of market concentration. The index ranges from 0 to 1, so that the higher the market’s concentration (HHI closer to 1), the lower the competition within the industry. The index is calculated for each year and industry (grouped 26. On the link between competition and productivity, see Ahn (2002), Holmes and Schmitz. (2010), and Buccirossi et al. (2013). 58 IV. Heterogeneity in Productivity Performance and Its Determinants as in the production function estimations). The coefficient on the HHI variable is expected to be negative, indicating that more intense competition is positive- ly associated with firms’ TFP performances. Furthermore, we employ different variables that can be correlated with competition, such as the industry-level av- erage profit margin, the log number of firms in an industry, as well as variables representing aggregate demand conditions, such as the log industry size calcu- lated based on total sales. Moreover, to derive implications for the relation be- tween competition (as well as R&D expenditures) and separate components of firm performance, we regressed the productivity indicators (growth, within and be- tween components) on the concentration index, profit margin, and R&D intensi- ty (calculated as R&D expenditures to sales ratio). The second lag of each regres- sor is used because the dependent variable covers the years t and t−1 (we employ 1-year average growth). We control for industry and year fixed effects in each re- gression. The industries are grouped as for the production function estimation. In addition, to identify the link between R&D and productivity performance, we employed two regression approaches: on the firm-level sample and industry-lev- el data. First, we utilized the firm fixed-effects approach controlling for 2-digit industry, year, and region effects. We used the first lag of each regressor in two separate specifications: taking the log of total R&D expenditures and the log of each R&D expenditure separately (internal, external, and equipment; see Box 7 and Table A4.1 for details on the data description). The results are given in Table 6. Firm and Sector Characteristics and Productivity Growth Small and medium-sized firms constitute the engine of productivity growth in Poland. The results in Table 4 and Table 5 show that small and medium com- panies in each sector have significantly higher productivity growth than large companies in Poland. Moreover, a peculiar subset of micro firms in the firm- level sample (henceforth called “micro-small”) — having fewer than 10 full-time employees, but at least 10 people engaged and accounting for 7 percent of the sam- ple — are improving their productivity the fastest over time across all firms and sectors. The large companies have the least dynamic productivity growth. SMEs exhibit the highest productivity performance potential, which implies that tar- geting smaller firms with public incentive programs will more likely increase economy-wide productivity further and, in turn, lead to higher growth. The characteristics of the industry that a company operates in matter for its individual productivity performance — in expanding industries, firms tend to improve productivity performance over time. The change in the relative 59 Paths of productivity growth in Poland: a firm-level perspective size of the industry, measured based on the industry’s total sales, has significant explanatory power on firm-level productivity performance. This result is valid for all three sectors and has several implications (Appendix Table A6.1 and Table A6.2). First, higher sales indicate the relative size of the sector and potentially higher competition. Second, productivity often rises as a result of positive de- mand shocks (Mayer, Melitz, Ottaviano, 2014 and 2016). Increased demand leads producers to shift their production toward their best-performing products or raise prices, both of which lead to increases in labor productivity and TFP. Third, more sizable sectors can generate higher demand for productivity-enhancing technologies because their market is larger (from the perspective of technolo- gy providers). We found a positive link between number of competitors in a giv- en industry as well as an industry-level profit margin and productivity growth for manufacturing (see Table 4). What is more, the concentration index for the construction and services industries indicates that high competition in expand- ing sectors leads to higher TFP. While competition positively influences aggregate economic outcomes, more intense competition is not always associated with better firm performance at the micro level. The relationship between competition and productivity at the firm level depends on various factors related to the industry’s structure and the initial level of competition. Contrary to what might be expected, the link is not always positive. Furthermore, innovation — the major component of firm-spe- cific productivity performance, has an inverted-U shape relationship with prod- uct market competition (Aghion et al., 2005). Thus, when competition is already intense and firms differ significantly in performance, more intense competition may discourage laggard firms from innovating. In this case, one can find a neg- ative link between productivity and competition. Conversely, when the level of competition is moderate initially and there is neck-to-neck competition, one can expect that firms tend to be more productive to escape from competitors. In or- der to assess the link between competition and firm performance, this study in- troduces four indicators that can be correlated with the product market com- petition: the number of competitors, the industry size calculated based on total sales, the industry-level profit margin, and market concentration (HHI Index). However, having a relatively large number of competitors in a very large mar- ket may not necessarily imply more intense competition. The changes in the mar- ket size, therefore, are introduced as a control variable into the estimations (re- sults with log industry size calculated based on total sales dummy are given in Appendix Table A6.1 and Table A6.2). 60 IV. Heterogeneity in Productivity Performance and Its Determinants In manufacturing, intense competition measured by the number of compet- itors positively influences productivity performance, while in services and construction, decreases in the market concentration are associated with high- er productivity. Table 4 displays that the number of competitors in manufac- turing has a significantly positive coefficient estimate regardless of the industry size introduced as a control (Appendix TableA6.1). This indicates that manufac- turing firms tend to exhibit faster productivity performance in sectors where there is an increase in the number of competitors. Moreover, product market concentration measured by the Herfindahl-Hirschman index has a positive co- efficient estimate indicating a positive association between productivity growth and increases in the market concentration. Also, firms operating in industries with an increasing average profit margin perform better in terms of productiv- ity. One possible explanation for this is that in manufacturing, firms that exhib- it improvements in their productivity performance tend to occupy more of the market, which leads to a joint increase in productivity and concentration in the industry. This, however, more likely occurs in the industries where concentra- tion is initially low, since in the latter part of this study, we also find evidence to a negative link between the level of concentration and productivity obtained at the industry-level (Table 7). Unlike in manufacturing, decreases in the market concentration are associated with higher productivity growth performance in services and construction (see Table 5). Moreover, the industry-level profit mar- gin does not have any significant influence on firm performance. These results jointly indicate that higher competition leads to better productivity performance in construction and services. This variation in the relationship between compe- tition and productivity in the two main sector groups is most likely due to the differences in the initial level of competition and concentration observed in the market. Further investigation is needed to understand so distinctively opposite results on the link between competition and productivity performance between manufacturing and construction and services. Ownership status matters for productivity growth, but the empirical results on the relationship are mixed. Due to the structure of the SP survey regarding ownership, we had to employ two regression strategies to inves- tigate the relationship between ownership status and TFP growth. When we introduce the foreign ownership dummy alone, the coefficient turns out to be insignificant, indicating no distinctive productivity performance for for- eign-owned establishments (Table 4 for manufacturing and Table 5 for con- struction and services). In our sample, however, there is a large group of firms 61 Paths of productivity growth in Poland: a firm-level perspective whose ownership status is unknown (42 percent of all companies), but they have distinctively better productivity performance. When we control for this undefined ownership category, the coefficient of foreign-owned firms turns out to be significantly positive (Appendix Table A6.1 for manufacturing and Table A6.2 for construction and services). This implies that apart from the undefined category, foreign-owned firms perform much better than others that reported domestic ownership in Poland. The positive relationship between foreign own- ership and productivity performance is likely driven by the transfer of tech- nologies and best management practices from abroad and easier access to for- eign markets (export-led productivity growth27). There is empirical evidence that the transfer of knowledge occurs only if the gap between host (Polish firms) and source (foreign entities) is sufficiently pronounced (Benfratello and Sembenelli, 2006). However, it is challenging to determine the causal effect of ownership on productivity performance explicitly. With takeovers being a predominant mode of entry by foreign entities, higher productivity at for- eign-owned establishments might reflect the selection (so-called “cherry-pick- ing”) of high-productivity firms for takeover (Griffith et al., 2004; Benfratello and Sembenelli, 2006). Young firms in Poland exhibit faster productivity growth than older estab- lishments. The relationship between the ages that a firm operates on the mar- ket and productivity growth suggests that younger firms experience the fast- est productivity growth among all companies (see Appendix Table A6.3). Even though new firms entering the market in manufacturing might begin with low levels of productivity (see Figure 11 and Chapter III for details), they grow much faster than older establishments in the sector. Also, the micro firms (described as micro-small) with an exceptional productivity performance that we observe in the dataset might also be the youngest among all companies, which makes the result consistent. In this regard, facilitating market entry conditions might fos- ter aggregate productivity growth. R&D and Productivity Growth Firms of various characteristics invest in R&D differently. R&D expenditures vary significantly across firm size, ownership status, and leading industry of activity. (See Box 7 for details on the R&D dataset.) First, large firms are more 27  See, for instance, Lederman et al. (2018), Atkin et al. (2017), and De Loecker (2013). 62 IV. Heterogeneity in Productivity Performance and Its Determinants likely to incur R&D expenditures than other establishments — 20 percent of large firms in the sample perform R&D, compared to only 7 percent of medium firms and 1 percent of small firms. Second, foreign-owned companies report invest- ments in R&D (7 percent of foreign-owned firms in the sample) more often than state-owned (5 percent), sole domestic (4 percent), and foreign-domestic estab- lishments (4 percent). Third, the older the company, the more likely it is to incur R&D expenditures. Seven percent of companies in the sample operating on the market longer than 20 years perform R&D, compared to 1 percent of young firms (less than 3 years on the market) and 3 percent of established firms (between 4 and 19 years on the market). Lastly, firms operating in medium-high–technology manufacturing industries and knowledge-intensive service industries are not only more likely to report R&D activities but also spend more on those invest- ments. (See Appendix Table A4.1 for details on R&D expenditures across indus- tries and sectors.) However, across firm groups, R&D expenditures are usually performed in the reporting unit (around 83 percent) rather than purchased from other contractors and subcontractors. BOX 7  R&D Data Sources To evaluate the link between R&D and Polish enterprises’ performance, we use firm-level data cov- ering R&D expenditures incurred by the companies from 2009 to 2019. The data on R&D activities comes from the annual Questionnaire on research and experimental development (PNT-01), main- tained by Statistics Poland. All firms (regardless of size) operating in Poland, conducting or commis- sioning R&D activities (continuous or ad hoc), and allocating funds for such works are legally obliged to submit the form. Even though the firms are additionally incentivized to fill out the questionnaire (including the requirement of providing proof of R&D activities in the form of PNT-01 when apply- ing for publicly-operated funds), the number of observations in the dataset is very low. Data confi- dentiality decreases the number of observations further. When there are fewer than three compa- nies or the share of one firm is greater than three-quarters of the total in a given aggregation (year and industry in the study), the data cannot be shared (the Law of 29 June 1995 on official statistics, Journal of Laws No 88, item 439 as amended). There are 8,000 enterprises of all sizes (micro firms included) reporting R&D expenditures in the manufacturing, construction and service sectors between 2009 and 2019. Furthermore, after exclud- ing micro firms and merging the R&D dataset with the estimated TFP index that is based on the enterprise survey productivity dataset (we need to employ the same companies as for the produc- tion function estimations), we are left with 17,000 observations and about 5,000 companies during an 11-year-period. Thus, only 3.5 percent of the firms observed in the productivity dataset are dis- closing R&D expenditures, with an average of about three observations per firm. The details of the dataset are given in Appendix A.4. Due to the low number of observations and the highly unbalanced structure of the R&D dataset, the analysis results must be interpreted with caution. The R&D analy- sis investigates a correlation with different firm groups and productivity rather than causal link and capturing the causal relationship between R&D and productivity requires further empirical work. 63 Paths of productivity growth in Poland: a firm-level perspective There is a positive link between R&D expenditures and productivity perfor- mance in the manufacturing sector, but not in the construction and services sectors. The regression results on R&D and firm performance are mixed (Table 6). We do not find any significant effect of firm-level expenditures on R&D on TFP growth in the services and construction industries. However, R&D is significantly and positively associated with firms’ productivity growth in manufacturing.28 The fact that R&D expenditures and productivity are positively correlated in man- ufacturing may imply potential positive returns from R&D incentives. One pos- sible reason for inconclusive results regarding the link between R&D expendi- tures and productivity growth might be the low number of companies reporting R&D expenditures in the panel dataset, especially until 2016. Second, productiv- ity advances in services often happen through investments in intangible assets, however, their reliable measurement is still a challenge (Demmou et al. 2019). TABLE 6  Fixed-Effects Regression of TFP on R&D Variables, 2009 – 19 Dependent variable TFP Specifications (1) (2) (1) (2) Manufacturing Construction & services 0.0024***   0.0002 R&D internal (log) (0.0005)   (0.0007) -0.0006   -0.0003 R&D external (log) (0.0005)   (0.0009) 0.0001   -0.0001 R&D equipment (log) (0.000567)   (0.001) 0.0019*** 0.0001 R&D total (log) (0.0004) (0.0006) 3.309*** 3.309*** 3.776*** 3.774*** Constant (0.004) (0.0076) (0.007) (0.014) Observations 168,162 168,162 353,276 353,276 R-squared 0.000 0.000 0.000 0.000 2-Digit industry FE yes yes yes yes Year FE yes yes yes yes Region FE yes yes yes yes Source: Statistics Poland calculations. Note: Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. 28.  The literature finds similar results, see Doraszelski and Jaumandreau (2013), Hall et al. (2010), Bloom et al. (2017). 64 IV. Heterogeneity in Productivity Performance and Its Determinants R&D expenditures lead to more efficient resource allocation (impact on the between component) but not to improved firm productivity performance (the within component). Firms can make R&D investments to create innovative new products, production methods, or organizational structures or adopt them from others, which would increase the R&D-making firms’ individual productiv- ity performance. In some cases, however, firms can do R&D to understand market demand and trends to better advertise products, learn consumer preferences, and test demand conditions (for instance, on foreign markets), in which case the firm benefits from the R&D investment by expanding its market share. As Table 7 shows, there is a positive link between R&D expenditures and the between com- ponent. This implies that firms that are more productive invest in R&D and ex- pand their market shares rather than improving their innovation performance. The R&D effect on the industry-specific within component measuring overall growth from innovation or adaptation performance, however, is not significant. TABLE 7  Linear Regression of TFP Growth and Its Components on Selected Productivity Determinants, 2009 – 2019   (1) (2) (3) Dependent Variables Productivity Growth Within Between Concentration Index -1.823*** -0.303 -1.52*** (0.596) (0.459) (0.478) Profit Margin -0.0601** 0.0006 -0.0607*** (0.0299) (0.0115) (0.0217) R&D Intensity 1.048 -0.0426 1.091** (0.688) (0.354) (0.517) Constant 0.0931*** 0.0506* 0.0425* (0.027) (0.0289) (0.0221) Observations 242 242 242 R-squared 0.319 0.405 0.224 2-Digit Industry FE yes yes yes Year FE yes yes yes Source: Elaboration based on Statistics Poland calculations. Note: Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. 65 V AREAS FOR POLICY ACTION Based on the key findings from the analysis, this section aims to provide insights on where and how regulations and support mechanisms could be adjusted to improve the productivity performance of firms. It also offers some consider- ations for areas for further research. Enhancing firms’ capabilities. Poland’s aggregate productivity growth accel- erated in all sectors from 2017 onwards and was mainly driven by within-firm productivity improvements. However, not all industries saw their firms’ capabil- ities increase. Moreover, Polish companies lag in terms of digitalization and inno- vation capabilities. (Poland ranks 23rd in the Digital Economy and Society Index and 24th on the Innovation Scoreboard out of 27 EU member states.) About half of Polish firms do not use even basic techniques that improve management (World Bank, 2021), and relatively few current instruments are dedicated to improving management practices (World Bank, 2019). Strengthening within-firm productiv- ity growth is critical for sustaining long-term economic growth, and it remains a priority on the productivity-enhancing policy agenda. Improving within-firm performance means increasing the amount of output firms produce with a con- stant quantity of inputs (such as labor, capital, and intermediate inputs) by, for instance, strengthening managerial skills, workforce skills, innovation capac- ity, and technology absorption capability. In this respect, human capital–related abilities such as digital literacy and leadership skills are as important as tech- nology itself. Policy interventions aiming to improve within-firm performance include, for instance, providing business advisory (outsourcing consultants) and technology extension services, facilitating entrepreneurial networks and clusters, offering vouchers for training, and improving employees’ digital skills. Supporting small and medium firms because they are the engine of pro- ductivity growth in Poland. The empirical results indicate a need to intensi- fy competition in Polish industries and allow smaller establishments to exert 67 Paths of productivity growth in Poland: a firm-level perspective competitive pressure on their larger counterparts. Besides reducing barriers to competition by detecting and removing regulations that provide advantag- es asymmetrically to large firms, eliminating barriers to growth for smaller firms, especially in manufacturing, is critical for economic growth. Improv- ing SMEs’ ability to grow means, for instance, facilitating their access to finance, promoting financial market deepening, and supporting the development of their innovation supply side. Besides financial constraints, barriers to adop- tion may also include information gaps (such as the mistaken belief that dig- italization is not suitable for SMEs), lack of awareness of technology benefits, managers’ loss aversion, and overestimation of firm’s productivity (Martin et al., 2013; Bloom et al., 2012; Dincă et al., 2019). There is evidence that digital adop- tion more often takes place in larger establishments (World Bank, 2021). Poli- cy interventions need to address potential barriers to SMEs’ technology adop- tion because their gains from adoption are possibly more significant than for larger establishments. Improving allocative efficiency in manufacturing. Deterioration in allocative efficiency in manufacturing calls for attention in the structure and targeting of existing incentive programs in the form of tax reliefs, subsidized credits, grants, and other types of firm-specific interventions. For instance, taxes that depend on the size of a company may incentivize firms to remain below a certain threshold and limit their growth. Poland’s growth would benefit from supporting compa- nies with high potential to innovate or grow rather than help inefficient establish- ments survive in the market. Supporting potentially high-productivity produc- ers should not be limited to subsidizing selected growth-enhancing investments, but also facilitating their conditions to do business and access finance. For exam- ple, an economic policy that would mitigate market imperfections such as lack of developed second-hand markets for capital goods or improving information asymmetries by developing efficient employment agencies would facilitate labor flow to more productive firms. These types of supportive interventions will work in favor of more efficient factor allocation and accelerate the creative destruc- tion necessary for restructuring of poor-performing industries. Investigating barriers for growth for large Polish firms. Even though large establishments constitute only 5 percent of all firms in Poland, they employ more than half of the labor force. Their productivity performance is essential for the aggregate outcomes. In terms of productivity-enhancing investments, compared to smaller entities, large enterprises are usually not financially con- strained. However, the empirical evidence suggests that their productivity per- formance is substantially worse than the performance of smaller establishments. 68 V. Areas for Policy Action Since the financial nudges to improve productivity are not the most effective policy option for large enterprises, the policy makers could further investigate what barriers firms face when innovating or adopting new technologies. One possible reason for large firms’ poor performance is that they are frequently exposed to many policy interventions, such as industrial regulations or taxes. Revising the existing large firms’ business environment, especially in the food, beverages, metals and rubber industries and targeting elimination of indus- trial protection might be the best policy option to improve the performance of larger establishments.  Strengthening linkages between Polish and foreign firms. The empirical ev- idence indicates that firms in expanding industries exhibit better productivity performance than establishments in other industries. In addition, there is some weak empirical evidence, that firms that reported foreign ownership experience faster productivity growth than those that reported domestic and state owner- ship. Also, following the results from “Return on Investment of Public Support to SMEs and Innovation in Poland”, these findings can lead to the conclusion that firms in the Polish economy generally benefit from lower barriers to interna- tional trade (World Bank, 2019). Policy makers could consider interventions to establish or strengthen the linkages between Polish firms (especially SMEs) and foreign firms through, for instance, supplier development programs and match- making that would facilitate the integration of domestic producers with GVCs. Promoting exports also involves facilitating domestic infrastructure for test- ing and certification, without which the sunk costs of exporting are high. How- ever, a tailored set of skills (linkage capabilities) and managerial processes are required to fully benefit from exposure to international markets. Tacit knowl- edge about foreign markets is best transferred by interacting with the business community of the export market. Enhancing the industry-specific and tailored approach to policy design. Polish firms exhibit a high degree of heterogeneity in their productivity performance within narrowly defined industries and across sectors. Consequently, an indus- try-specific perspective is a necessity in policy design. Carefully targeted pro- grams lower the financial costs of policy implementation. Even though the driv- ers of productivity are distinct, they are also interlinked, and a well-designed system of incentives accounts for complementarity in policy interventions. Po- land needs a mix of policies that include improving the functioning of markets and business operating environments (between component), as well as policies that support firm upgrading (within component). While firms differentiate by their capabilities, especially across sizes, and face various market conditions, 69 Paths of productivity growth in Poland: a firm-level perspective effective policy design accounts for those differences. Given the high produc- tivity growth potential in Polish SMEs, it is crucial to revise and adjust existing policies to increase the share of smaller establishments investing in digitaliza- tion and technology adoption. Employing the new wave of productivity diagnostics and analytics. Due to the lack of variables reflecting firm-level prices and quantities of outputs and in- puts in the dataset prepared by Statistics Poland, the computation of TFP in the study employs industry-level price indices (Box 3). Using industry-level pric- es in the estimation of firm-level TFP causes the estimated productivity of indi- vidual firms to reflect not only efficiency in production but also factors driving firm-level price variation (general firm performance). In turn, the TFP value grows not only as a result of production efficiency improvements but also due to higher prices following better product quality, advertising, marketing, or product market competition. Distinguishing production efficiency and firm performance is crucial because otherwise improvements in productivity driv- en by market power and markups are confused with improvements in real effi- ciency. Even though both efficiency and firm performance are important for an overall growth strategy, providing precise policy recommendations requires a comprehensive understanding of drivers and barriers to productivity growth. Therefore, employing the new wave of productivity diagnostics is necessary to quantify the importance of different market failures or distortions and define ap- propriate priorities. These approaches require detailed firm-level data on prices, marginal costs, intangible assets, quality, and management. Consequently, the institutional capacity building of national statistical offices is essential for evi- dence-based policy design. Improving data accessibility.  Due to statistical confidentiality law, firm-lev- el data on business activities in Poland are restricted and not easily accessible to researchers. Providing the results presented in this report was only pos- sible through the World Bank’s close cooperation with an experienced team at Statistics Poland. However, this type of cooperation is often time-consum- ing to establish and may result in delays and inefficiencies that would be re- solved if the analysts were to have direct access to the firm-level data. In some countries, researchers can directly access data (for instance, using comput- ers on the statistics agency’s premises), with appropriate privacy controls pre- venting firm-identifying information from being made public (World Bank, 2019). Delivering high-quality research depends on government openness to share data. Given the importance of evidence-based research in effective policy design, one recommendation would be to establish an “Innovation and 70 V. Areas for Policy Action Productivity Excellence Center” or similar entity (see also World Bank, 2019) to provide analytical inputs to instrument design and facilitate a knowledge platform with a repository of best practices related to policy design, imple- mentation, and evaluation. 71 Paths of productivity growth in Poland: a firm-level perspective REFERENCES Aghion, P., Bloom, N., Blundell, R., Griffith, R., & Cirera, X., & Maloney, W. F. (2017). The innovation Howitt, P. (2005). Competition and innovation: paradox: Developing-country capabilities and An inverted-U relationship. The Quarterly the unrealized promise of technological catch- Journal of Economics, 120(2), 701 – 728. up. Washington, DC: World Bank. Ahn, S. (2002). Competition, innovation and pro- Cirera, X., Comin, D., Cruz, M., Lee, K. M., & Soares ductivity growth: a review of theory and evi- Martins-Neto, A. (2021). Firm-level technol- dence. Available at SSRN 318059. ogy adoption in the state of Ceará in Brazil. Albinowski, M., Hagemejer, J., Lovo, S., & Varela, G. Washington, DC: World Bank. (2015). Sustaining micro competitiveness to en- Cohen, W. M., & Klepper, S. (1996). A reprise of size sure convergence and macro resilience of the and R&D. The Economic Journal, 106(437), 925 – 951. Polish economy (No. 23). Ministry of Finance Cusolito, A. P., Fernandes, A. M. & Maemir, H. in Poland. (2018). Cross-country evidence on the ef- Arendt, L. (2008), Barriers to ICT adoption in SMEs: fects of demand shocks on the product mix How to bridge the digital divide? Journal of Sys- [Background paper for Productivity Revisited]. tems and Information Technology, 10(2), 93 – 108. Washington, DC: World Bank. Atkin, D., Khandelwal, A. K., & Osman, A. (2017). Cusolito, A. P., & Maloney, W. F. (2018). Productivity Exporting and firm performance: Evidence revisited: Shifting paradigms in analysis and from a randomized experiment. The Quarterly policy. Washington, DC: World Bank. Journal of Economics, 132(2), 551 – 615. De Loecker, J. (2013). Detecting learning by ex- Bartel, A., Ichniowski, C., & Shaw, K. (2007). How porting. American Economic Journal: does information technology affect produc- Microeconomics, 5(3), 1 – 21. tivity? Plant-level comparisons of product in- Demmou, L., I. Stefanescu & A. Arquie (2019). Pro- novation, process improvement, and work- ductivity growth and finance: The role of intan- er skills. The Quarterly Journal of Economics, gible assets — a sector level analysis. OECD Eco- 122(4), 1721 – 1758. nomics Department Working Papers, No. 1547. Benfratello, L., & Sembenelli, A. (2006). Foreign Dincă, V. M., Dima, A. M., & Rozsa, Z. (2019). ownership and productivity: Is the direction Determinants of cloud computing adoption by of causality so obvious? International Journal Romanian SMEs in the digital economy. Journal of Industrial Organization, 24(4), 733 – 751. of Business Economics and Management, 20(4), Bloom, N., Jones, C. I., Van Reenen, J. & Webb 798 – 820. M. (2017). Are ideas getting harder to find? Doraszelski, U., & Jaumandreu, J. (2013). R&D and NBER Working Paper 23782, Cambridge, MA: productivity: Estimating endogenous pro- National Bureau of Economic Research. ductivity. Review of Economic Studies, 80(4), Brouthers, K. D., & Brouthers, L. E. (2003). Why 1338 – 1383. service and manufacturing entry mode choic- European Commission (2020). Country report es differ: The influence of transaction cost fac- Poland 2020 [Commission staff working tors, risk and trust. Journal of Management document]. Studies, 40(5), 1179 – 1204. European Commission (2021). European Buccirossi, P., Ciari, L., Duso, T., Spagnolo, G., & Innovation Scoreboard 2021. Luxembourg: Vitale, C. (2013). Competition policy and pro- European Union. ductivity growth: An empirical assessment. Eurostat. (2021). EUROSTAT: Structural Business Review of Economics and Statistics, 95(4), Survey, Statistical Office of the European 1324 – 1336. Communities (Eurostat), Luxembourg. 72 References Grela, M., Majchrowska, A., Michałek, T., Mućk, J., Mayer, T., Melitz, M. J., & Ottaviano, G. I. (2016). Stążka-Gawrysiak, A., Tchorek, G., & Wagner, Product mix and firm productivity responses M. (2017). Is Central and Eastern Europe con- to trade competition. The Review of Economics verging towards the EU-15? Narodowy Bank and Statistics, 1 – 59. Polski, Education & Publishing Department. Melitz, M. J., & Polanec, S. (2015). Dynamic Olley‐ Griffith, R., Redding, S., & Simpson, H. (2004). For- Pakes productivity decomposition with entry eign ownership and productivity: New evi- and exit. The Rand journal of economics, 46(2), dence from the service sector and the R&D lab. 362 – 375. Oxford Review of Economic Policy, 20(3), 440 – 456. Mohnen, P., & Hall, B. H. (2013). Innovation and Hall, B. H., Mairesse, J., & Mohnen, P. (2010). productivity: An update. Eurasian Business Measuring the Returns to R&D. In Handbook of Review, 3(1), 47 – 65. the Economics of Innovation, Vol. 2, 1033 – 1082. Office of Competition and Consumer Protection North-Holland. (2021 May 25). [Press release] Holmes, T. J., & Schmitz Jr, J. A. (2010). Competition Organisation for Economic Co-operation and and productivity: a review of evidence. Annu. Development (2019). OECD Economic Outlook Rev. Econ., 2(1), 619 – 642. 2019, Chapter 2. International Labour Organization (2013). Global Pop Silaghi, M. I., Alexa, D., Jude, C., & Litan, C. Employment Trends 2013. (2014). Do business and public sector research Laplante, S. K., Skaife, H. A., Swenson, L. A., & and development expenditures contribute Wangerin, D. D. (2019). Limits of tax regu- to economic growth in Central and Eastern lation: Evidence from strategic R&D classi- European countries? A dynamic panel estima- fication and the R&D tax credit. Journal of tion. Economic Modelling, 36(C), 108 – 119. Accounting and Public Policy, 38(2), 89 – 105. Nekrep, A., Strašek, S., & Boršič, D. (2018). Lederman, D., Miion, G., Prado-Provenca R., & Silva Productivity and economic growth in the J. (2018). Trade, skills, and productivity [Work- European Union: Impact of investment in re- ing paper]. Washington, DC: World Bank. search and development. Naše gospodarstvo/ Lewandowski, P., & Magda, I. (2014). Zatrudnienie Our Economy, 64(1), 18 – 27. w Polsce 2013. Praca w dobie przemian Statistics Poland (2020). Activity of non-financial strukturalnych. enterprises in 2019. Levinsohn, J., & Petrin, A. (2003). Estimating pro- Statistics Poland (2021). Labour force survey in duction functions using inputs to control Poland — quarter 1/2021. for unobservables. The Review of Economic Supreme Audit Office. (2018). Inspection report: Studies, 70(2), 317 – 341. “Zabezpieczenie interesów Skarbu Państwa i Krugman, P. R. (1994). The Age of Diminished podwykonawcy w umowach na realizację in- Expectations: U.S. Economic Policy in the 1990s. westycji drogowych prowadzonych przez Gen- Cambridge, MA: MIT Press. eralną Dyrekcję Dróg Krajowych i Autostrad” Martin, F. M., Ciovica, L., & Cristescu, M. P. (2013). (in Polish). https://www.nik.gov.pl/aktualnos- Implication of human capital in the devel- ci/inwestycje-drogowe-zabezpieczenie-intere- opment of SMEs through the ICT Adoption. su-skarbu-panstwa-i-podwykonawcow.html Procedia Economics and Finance, 6, 748 – 753. Szpunar, P., & Hagemejer, J. (2018). Globalisation Matteucci, N., O’Mahony, M., Robinson, C., & and the Polish economy: macro and micro Zwick, T. (2005). Productivity, workplace per- growth effects. BIS Paper No. 100r. formance and ICT: Industry and firm‐level ev- Szreder, J. (2018). Nowoczesne metody finansowa- idence for Europe and the US. Scottish Journal nia przedsięwzięć deweloperskich. Studia i of Political Economy, 52(3), 359 – 386. Prace WNEiZ US, (54/1), 33 – 47. Mayer, T., Melitz, M. J., & Ottaviano, G. I. (2014). United Nations Conference on Trade and Develop- Market size, competition, and the product ment. (2013). World investment report 2013. UN. mix of exporters. American Economic Review, United Nations Conference on Trade and Develop- 104(2), 495 – 536. ment. (2014). World investment report 2014. UN. 73 Paths of productivity growth in Poland: a firm-level perspective United Nations Conference on Trade and Develop- World Bank. (2018). Poland Catching-Up Regions 2 ment. (2015). World investment report 2015. UN. — Safer Food, Better Business in Podkarpackie Van Beveren, I. (2012). Total factor productivi- and Lubelskie. ty estimation: A practical review. Journal of World Bank. (2019). Return on investment of pub- Economic Surveys, 26(1), 98 – 128. lic support to SMEs and innovation in Poland. Verspagen, B. (1995). R&D and productivity: A World Bank. (2021). Business Pulse Survey: Poland, broad cross-section cross-country look. 4th edition. https://www.parp.gov.pl/storage/ Journal of Productivity Analysis, 6(2), 117 – 135. publications/pdf/Prezentacja-v7---FINAL.pdf World Bank. (2017). Lessons from Poland, Insights Wu, H., & Broughton, N. (2019). Business Basics: for Poland: A Sustainable and Inclusive Nudging firms to improve productivity. BEIS Transition to High Income Status. Research Paper Number 2019/17. 74 APPENDIX A.1  Endogenous and Explanatory Variables To calculate productivity, one must construct variables not explicitly reported in the Annual Enterprise Survey (SP). We express firm labor costs, intermediate consumption, and output as: costsLabor = wages + insurFees + costsBusTrip, intermConsump = costsMatEnerg + costsExternalServ + costsOtherIC – costsBusTrip, output = revNet – goodsChange – tax – valCM + goodsU2 – goodsU1 + goodsF2 – goodsF1, revNet = revNetP + goodsChange + costOwn + revNetCM with the SP variables summarized: wages remunerations (from the income statement) insurFees social security contribution costsBusTrip business travel costs costsMatEnerg use of raw materials and energy costs costsExternalServ outside services costs costsOtherIC other operating costs (includes business travel costs) revNet net revenues revNetP net revenues from the sale of products (goods and services) costOwn cost of producing products for own use revNetCM net revenues from sales of commodities and raw materials goodsChange change in stocks of finished goods and work in progress tax excise tax valCM the value of sold commodities and raw materials goodsU(1/2) semi-finished products and production in progress (the beginning and end of the year) goodsF(1/2) finished goods (the beginning and end of the year) tangFixAss(1/2) tangible fixed assets (the beginning and end of the year) intangFixAss(1/2) intangible fixed assets (the beginning and end of the year) 75 Paths of productivity growth in Poland: a firm-level perspective The define firm gross value-added and capital as: gva  = output – intermConsump, tangFixAss1 + tangFixAss2 intangFixAss1 + intangFixAss2 capital = + 2 2 The relatively short sample period makes the Perpetual Inventory Method to construct capi- tal unreliable. The final measurement of variables is calculated with the enterprise’s real gross value added and real capital at constant average prices from 2010. For this purpose, we employ capital and gross value added price deflators at 2-digit NACE Rev. 2 (published yearly by Statistics Poland as Prices in the national economy). A.2  Firm-Level Panel Dataset Characteristics TABLE A2.1  The Growth of Selected Industrial Characteristics in the Sample by Sectors — Manufacturing, Construction and Services. ∆ Represents the % Change of an Index Number between 2009 And 2019. a. manufacturing Sectoral Sectoral Sectoral Sectoral GVA GVA labor labor Divisions share in share in share in share in (NACE Rev. 2) code ∆ TFP ∆ LP ∆ GVA 2009 2019 ∆ Employ. 2009 2019 ∆ Rev. Low technology Food and 10 – 11 -1% 34% 21% 21% 15% -4% 18% 16% 57% beverages Textiles 13 40% 59% 63% 1% 1% -4% 2% 2% 105% Wearing apparel & 14 – 15 56% 67% 13% 2% 1% -42% 5% 3% 22% leather Wood, cork 16 32% 56% 57% 4% 3% -4% 4% 4% 58% & straw Paper 17 -11% 30% 47% 3% 3% 21% 2% 3% 100% Printing 18 29% 42% 81% 1% 1% 25% 1% 2% 68% Furniture 31 41% 62% 100% 4% 5% 14% 7% 7% 86% Others 32 55% 86% 164% 1% 2% 22% 1% 2% 136% Medium-low technology Rubber & 22 24% 47% 109% 7% 9% 40% 7% 9% 115% plastics Non-metallic 23 39% 77% 104% 6% 7% 8% 5% 5% 77% minerals 76 Appendix Sectoral Sectoral Sectoral Sectoral GVA GVA labor labor Divisions share in share in share in share in (NACE Rev. 2) code ∆ TFP ∆ LP ∆ GVA 2009 2019 ∆ Employ. 2009 2019 ∆ Rev. Basic & fabricated 24 – 25 8% 52% 65% 15% 15% 16% 15% 15% 93% metals Machinery repair & 33 25% 34% 61% 3% 3% 17% 3% 4% 116% installation Medium-high technology Chemicals & pharmaceu- 20 – 21 -5% 19% 28% 8% 6% 10% 5% 5% 58% ticals Computers & 26 – 27 84% 111% 205% 5% 10% 15% 7% 8% 73% electronics Machinery & 28 41% 70% 79% 6% 7% -3% 6% 5% 68% equipment Vehicles & 29 – 30 19% 54% 93% 10% 11% 28% 10% 12% 85% transport b. construction Sectoral Sectoral Sectoral Sectoral GVA GVA labor labor Divisions share in share in share in share in (NACE Rev. 2) code ∆ TFP ∆ LP ∆ GVA 2009 2019 ∆ Employ. 2009 2019 ∆Rev. Buildings 41 45% 45% -3% 41% 36% -39% 39% 33% 32% Civil 42 27% 43% 16% 33% 34% -21% 33% 36% 29% engineering Specialized 43 45% 45% 29% 26% 30% -18% 28% 31% 28% activities c. services Sectoral Sectoral Sectoral Sectoral GVA GVA labor labor Divisions share in share in share in share in (NACE Rev. 2) code ∆ TFP ∆ LP ∆ GVA 2009 2019 ∆ Employ. 2009 2019 ∆ Rev. Electricity, 35 7% 60% 18% 14% 10% -27% 6% 4% -5% gas & steam Water 36 – 39 0% 10% 36% 4% 3% 24% 4% 4% 69% utilities Vehicles: wholesale, 45 37% 43% 81% 3% 4% 13% 3% 3% 107% retail and repair Wholesale 46 31% 46% 61% 20% 19% 6% 18% 16% 65% trade Retail trade 47 35% 42% 76% 13% 14% 18% 21% 21% 79% Transport & 49 – 53 18% 25% 63% 13% 13% 13% 18% 17% 103% storage 77 Paths of productivity growth in Poland: a firm-level perspective Sectoral Sectoral Sectoral Sectoral GVA GVA labor labor Divisions share in share in share in share in (NACE Rev. 2) code ∆ TFP ∆ LP ∆ GVA 2009 2019 ∆ Employ. 2009 2019 ∆ Rev. Accommoda- 55 40% 39% 45% 1% 1% -2% 2% 1% 72% tion Food & beverage 56 29% 38% 68% 1% 1% 15% 2% 2% 109% activities Real estate 68 33% 38% 33% 4% 3% -8% 4% 3% 32% Knowledge-intensive Publishing & 58 – 60 17% 43% 27% 3% 2% -6% 2% 1% 21% broadcasting Telecommu- 61 86% 91% 81% 7% 8% -24% 2% 2% 40% nications IT 62 – 63 9% 18% 219% 3% 6% 180% 2% 5% 225% Consulting 69 – 71 7% 24% 114% 4% 5% 82% 3% 5% 99% Research 72 – 75 -8% -25% 66% 3% 3% 95% 2% 3% 49% Adminis- tration & 77 – 82 46% 52% 125% 5% 7% 36% 11% 12% 220% support Source: Elaboration based on Statistics Poland calculations. Note: Darker shades of green indicate stronger development characteristics relative to industries with lighter blue within sectors. FIGURE A2.1  Levels of Productivity by Sector (2009 – 19) 5.0 4.5 Productivity 4.0 3.5 3.0 TFP: Manufacturing Construction Services Labor Productivity: Manufacturing Construction Services Source: Elaboration based on Statistics Poland calculations. 78 Appendix FIGURE A2.2  Growth in Performance Indicators by Sectors (2009 – 19) a. Gross value added b. Labor 20 10 15 5 10 5 Growth (%) Growth (%) 0 0 -5 -5 -10 -10 -15 -20 -15 c. Capital d. Investment 40 30 20 20 10 20 0 -10 10 Growth (%) Growth (%) -20 0 -30 -40 -10 -50 -20 -60 -70 -20 -80 -40 -90 e. Export f. Import 30 40 25 30 20 20 10 15 Growth (%) Growth (%) 0 10 -10 5 -20 0 -30 -5 -40 -10 -50 Manufacturing Construction Services Source: Elaboration based on Statistics Poland calculations. 79 Paths of productivity growth in Poland: a firm-level perspective A.3  Production Function Estimation TFP is calculated using the structural production function estimation approach by Ackerberg et al. (2015) that is implemented separately for every 2-digit NACE manufacturing, construction, and services industry. Due to lack of sufficient observations or data confidentiality, some of the 2-digit NACE industries were grouped in the production function estimations (e.g., Food&Beverages). The estimation sample covers the period from 2009 to 2019. Estimation results for the indus- try-level production functions and grouping of industries are given in Table A3.1. The TFP estimation is based on the Cobb-Douglas production function in the following form: βk βl Yit = Ait Kit Lit , (1) In the above equation Yit, Lit, K it is, respectively, the real gross value-added, labor, and capital inputs of enterprise i in period t. Ait is an idiosyncratic Hicks technological level used in the pro- duction process. Yit, Kit were measured by deflating gross value-added and physical capital to 2010 prices. The technology level used in the production process can be decomposed: β0+dummies Ait = TFPit Uit = e Vit Uit. (2) β+dummies Ait is the unobservable variable that can be expressed as the product of the constant term e0 , the volatility of individual productivity Vit and idiosyncratic white noise Uit = euit. If yit, lit, kit, vit are logarithms of Yit, Lit, Kit, Vit, then ωit = β0 + dummies + vit (3) represents the logarithm of productivity of enterprise i. Production function can be presented in a log-linear form: yit = ωit + βk kit + βl lit + uit (4) Coefficient ωit is often interpreted as a state variable in the enterprise decision problem of select- ing factor inputs, with the error term uit correlated with measurement errors and represents so-called unpredictable productivity shock. Equation (4) is estimated to determine individual TFP. As a result, we obtain estimator of logTFP: ωit = yit – βk kit + βl lit. (5) It follows that individual TFP is given by: ω TFPit = e it (6) 80 Appendix To estimate the production function, we use the control function method (Levinsohn and Petrin, 2003). It helps address endogeneity problems (van Beveren, 2012) by employing variables that proxy for unobservable productivity shocks. It assumes that productivity can be proxied by outlays on materials and energy. The model is estimated in a 2-step procedure. In the first step, the estimates of unobserved productivity are calculated. In the second step, using generalized method of moments (GMM) and estimates from the previous stage, non- linear regression of gross value added of surviving firms is estimated to determine capital and labor elasticities. TABLE A3.1  Production Function Estimation Results a. Manufacturing Labor Capital #obs #firms 0.764*** 0.296*** Food & beverages 25,172 4,153 -0.00233 -0.0018 0.780*** 0.180*** Textiles 3,561 645 -0.00311 -0.00535 0.829*** 0.164*** Wearing apparel & leather 6,033 1,090 -0.0043 -0.00584 0.827*** 0.228*** Wood, cork & straw 7,711 1,418 -0.00497 -0.00551 0.619*** 0.399*** Paper 4,564 753 -0.00646 -0.00533 0.894*** 0.171*** Printing 3,913 737 -0.00867 -0.00612 0.709*** 0.319*** Chemicals & pharmaceuticals 5,988 994 -0.00801 -0.00674 0.810*** 0.241*** Rubber & plastics 14,226 2,366 -0.00261 -0.00192 0.840*** 0.255*** Non-metallic minerals 8,338 1,483 -0.00713 -0.00786 0.860*** 0.160*** Basic & fabricated 27,911 5,031 -0.00187 -0.000865 0.815*** 0.177*** Computers & electronics 8,220 1,369 -0.00579 -0.00561 0.835*** 0.175*** Machinery & equipment 10,241 1,800 -0.00296 -0.00149 0.811*** 0.241*** Vehicles & transport 6,367 1,074 -0.00932 -0.0124 0.922*** 0.140*** Furniture 7,181 1,288 -0.00479 -0.00347 0.776*** 0.215*** Others 2,943 570 -0.0111 -0.00852 0.865*** 0.115*** Machinery repair & installation 8,229 1,819 -0.00535 -0.00629 81 Paths of productivity growth in Poland: a firm-level perspective b. Construction Labor Capital #obs #firms 0.803*** 0.200*** Buildings 20,755 4,997 -0.00559 -0.00478 0.852*** 0.179*** Civil engineering 17,155 3,511 -0.00296 -0.0053 0.931*** 0.130*** Specialized activities 19,354 4,686 -0.00296 -0.00374 c. Services Labor Capital #obs #firms 0.758*** 0.332*** Electricity, gas & steam 4,866 747 -0.00698 -0.00713 0.839*** 0.198*** Water utilities 13,916 2,055 -0.00109 -0.00179 0.904*** 0.176*** Vehicles - wholesale, retail, repair 15,781 2,825 -0.000966 -0.00154 0.846*** 0.168*** Wholesale trade 95,548 18,980 -0.000001 -0.000001 0.796*** 0.147*** Retail trade 64,801 13,285 -0.000001 -0.000001 0.822*** 0.148*** Transport & storage 30,817 6,433 -0.00159 -0.00169 0.935*** 0.144*** Accommodation 6,149 1,289 -0.00458 -0.00656 0.848*** 0.130*** Food & beverage activities 5,359 1,441 -0.00685 -0.00742 0.848*** 0.191*** Publishing & broadcasting 4,397 849 -0.00658 -0.0112 0.803*** 0.270*** Telecommunications 1,942 484 -0.0111 -0.0182 0.941*** 0.119*** IT 10,638 2,510 -0.00383 -0.00591 0.794*** 0.156*** Real estate 21,918 3,579 -0.00319 -0.0039 0.948*** 0.105*** Consulting 17,861 3,992 -0.00307 -0.00288 0.911*** 0.0595*** Research 7,551 1,882 -0.00635 -0.00809 0.706*** 0.167*** Administration & support 18,090 4,362 -0.00211 -0.00246 Source: Elaboration based on Statistics Poland calculations. Note: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All equations include time dummies and 2-digit industry dummies if a group contains more than one 2-digit industry. 82 Appendix A.4  R&D Dataset Description TABLE A4.1  The Descriptive Statistics, the Number of Observations, and Firms in the R&D Panel Dataset a. Manufacturing (by an employee) R&D equipment R&D external R&D internal (thous. PLN) No. observ. No. Firms Divisions code R&D R&D (NACE 2.) Low technology Food and beverages 10 – 11 1,068 337 2,113 0.6 1,960 154 723 Textiles 13 259 71 2,443 1.6 2,305 138 453 Wearing apparel & leather 14 – 15 119 48 936 0.2 772 164 75 Wood, cork & straw 16 289 105 1,698 0.6 1,479 219 390 Paper 17 276 100 802 0.4 584 218 196 Printing 18 181 79 1,481 0.8 1,447 34 455 Furniture 31 358 117 1,968 0.5 1,803 165 466 Others 32 405 119 1,313 1.7 1,272 41 292 Medium-low technology Rubber & plastics 22 1,415 450 1,132 0.9 873 259 365 Non-metallic minerals 23 706 222 1,356 0.9 1,129 226 504 Basic & fabricated metals 24 – 25 2,533 775 1,999 1.6 1,579 421 423 Machinery repair & installation 33 408 160 2,054 1.1 1,805 249 151 Medium-high technology Chemicals & pharmaceuticals 20 – 21 1,876 401 2,729 5.4 2,206 523 378 Computers & electronics 26 – 27 2,224 563 3,863 5.6 3,563 300 708 Machinery & equipment 28 1,926 521 2,030 3.4 1,629 401 377 Vehicles & transport 29 – 30 1,185 284 10,545 5.5 6,932 3,613 1,169 b. Construction (by an employee) R&D equipment R&D external R&D internal (thous. PLN) No. observ. No. Firms Divisions code R&D R&D (NACE 2.) Buildings 41 141 67 1,281 0.1 1,156 125 74 Civil engineering 42 241 96 1,205 0.2 1,086 120 144 Specialized activities 43 319 126 1,378 0.4 904 474 198 83 Paths of productivity growth in Poland: a firm-level perspective c. Services (by an employee) R&D equipment R&D external R&D internal (thous. PLN) No. observ. No. Firms Divisions code R&D R&D (NACE 2.) Electricity, gas & steam 35 275 74 3,244 0.6 1,672 1,571 748 Water utilities 36 – 39 346 132 732 0.2 659 73 256 Vehicles — wholesale, retail, repair 45 102 39 3,632 0.4 3,485 147 393 Wholesale trade 46 1,944 780 4,097 1.7 3,032 1,066 319 Retail trade 47 289 157 1,006 0.0 920 87 118 Transport & storage 49 – 53 186 68 1,146 0.0 600 546 147 Accommodation 55 10 7 238 0.0 228 9 100 Food and beverage activities 56 10 6 474 0.0 311 164 178 Real estate 68 93 38 3,133 0.3 3,065 68 1,133 Knowledge-intensive Publishing & broadcasting 58 – 60 202 86 5,787 3.3 5,694 93 399 Telecommunications 61 161 67 62,285 20.8 59,124 3,161 5,950 IT 62 – 63 1,847 735 6,236 11.7 5,983 253 461 Consulting 69 – 71 626 250 4,702 2.5 4,256 446 754 Research 72 – 75 2,287 933 6,031 20.1 5,513 518 458 Administration & support 77 – 82 153 75 1,765 0.1 1,259 506 124 Source: Elaboration based on Statistics Poland calculations. Note: The table reports the real mean firm-level expenditures of R&D (total, total divided by the number of employees in the industry, internal, external, and on equipment). The mean values are deflated with gross value-added deflators and ex- pressed at 2010 constant prices, in thousands of Polish zloty (national currency). Variables’ description: 1) R&D internal: R&D performed in the reporting unit, regardless of the source of funds; 2) R&D external: R&D works purchased from other contractors (subcontractors) domestic and foreign; 3) R&D equipment: machinery and technical equipment, software. 84 Appendix A.5  Sector- or Industry-Specific Results Based on Firm-Level Data FIGURE A5.1  Disaggregation of Productivity Change on the Industrial Level — Construction Sector a. Buildings 30 Productivity change (%) 20 10 0 -10 -20 -30 -40 -50 b. Civil engineering 30 Productivity change (%) 20 10 0 -10 -20 -30 -40 -50 c. Specialised construction 30 Productivity change (%) 20 10 0 -10 -20 -30 -40 -50 Labor productivity TFP Source: Elaboration based on Statistics Poland calculations. Note: There are only three industries that constitute the whole construction sector: construction of buildings, civil engi- neering and special construction activities. 85 Paths of productivity growth in Poland: a firm-level perspective FIGURE A5.2  Industry-Specific Time-Averaged Contributions to the Productivity: Melitz- Polanec Decomposition of the TFP Growth for the Entire Sample Period (2009 – 19). a. manufacturing Computers & electronics Wearing apparel & leather Others Machinery & equipment Furniture Textiles Non-metallic minerals Wood, cork & straw Printing Machinery repair & installation Rubber & plastics Vehicles & transport Basic & fabricated metals Food and beverages Chemicals & pharmaceuticals Paper -80 -60 -40 -20 0 20 40 60 80 100 120 Percent b. construction Specialised activities Civil engineering Buildings -80 -60 -40 -20 0 20 40 60 80 100 120 Percent c. services Telecommunications Administration & support Accommodation Vehicles - wholesale, retail, repair Retail trade Real estate Wholesale trade Food and beverage activities Publishing & broadcasting Transport & storage IT Electricity, gas & steam Consulting Water utilities Research -80 -60 -40 -20 0 20 40 60 80 100 120 Percent Within Between Upscaling Downscaling Source: Elaboration based on Statistics Poland calculations. 86 Appendix A.6  Other Regression Results Based on the Firm-Level Sample TABLE A6.1  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Introduced Separately) in Manufacturing (2009 – 19) a. Firm Size Dependent variables TFP Specifications (1) (2) (3) Competition 0.0614*** 0.0437* Number of firms (log) -0.0222 -0.0229 0.0384* Industry size (log) -0.0204 2.929*** 3.070*** 2.962*** Concentration Index -0.287 -0.289 -0.297 Firm size (reference group: small firms) 0.0635*** 0.0638*** 0.0638*** Micro-small -0.00879 -0.00878 -0.00878 Small [10–50) - ref. -0.0934*** -0.0941*** -0.0943*** Medium [50–250) -0.0077 -0.0077 -0.0077 -0.134*** -0.135*** -0.135*** Large [250+) -0.0135 -0.0136 -0.0136 2.543*** 2.029*** 1.460*** Constant -0.154 -0.239 -0.407 Observations 168162 168162 168162 R-squared 0.059 0.059 0.059 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes b. Ownership Dependent variables TFP Specifications (1) (2) (3) Competition 0.00735 0.0782*** Number of firms (log) -0.0202 -0.0215 -0.137*** Industry size (log) -0.019 1.938*** 1.954*** 2.367*** Concentration Index -0.255 -0.257 -0.268 87 Paths of productivity growth in Poland: a firm-level perspective Dependent variables TFP Specifications (1) (2) (3) Ownership (reference group: private domestic) -0.0113 -0.0112 -0.0094 State-owned (SOE) -0.0259 -0.0259 -0.0258 0.0831*** 0.0831*** 0.0829*** Foreign-owned (FOE) -0.0234 -0.0234 -0.0235 -0.0215 -0.0214 -0.0231 Dispersed capital -0.0487 -0.0487 -0.0489 a 0.733*** 0.733*** 0.733*** Undefined  -0.0229 -0.0229 -0.0229 2.170*** 2.108*** 4.078*** Constant -0.13 -0.211 -0.358 Observations 216,780 216,780 216,780 R-squared 0.093 0.093 0.094 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes Source: Statistics Poland calculations. Note: Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. a. Less than 1 percent of firms with undefined ownership status are used for identification in the fixed-effects regressions, meaning that they are changing ownership status between years which is captured by the regressions. Those companies appear to be the most productive in the sample, but it is likely a statistical artifact. (See Table 3 for details). TABLE A6.2  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Introduced Separately) in Construction and Services (2009 – 19) a. Firm Size Dependent variables TFP Specifications (1) (2) (3) Competition -0.0231** -0.0378*** Number of firms (log) (0.0104) (0.0106) 0.0563*** Industry size (log) (0.0107) -0.0777* -0.0815* -0.238*** Concentration Index (0.0457) (0.0457) (0.0541) Firm size (reference group: small firms) 0.0807*** 0.0807*** 0.0806*** Micro-small (0.0065) (0.0065) (0.0065) -0.0739*** -0.0737*** -0.0747*** Medium [50 – 250) (0.0057) (0.0057) (0.0057) -0.126*** -0.125*** -0.128*** Large [250+) (0.0135) (0.0135) (0.0135) 2.326*** 2.482*** 1.527*** Constant (0.0719) (0.100) (0.211) 88 Appendix Dependent variables TFP Specifications (1) (2) (3) Observations 353,276 353,276 353,276 R-squared 0.069 0.069 0.069 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes b. Ownership Dependent variables TFP Specifications (1) (2) (3) Competition -0.0067 -0.0182* Number of firms (log) (0.0098) (0.01) 0.0327*** Industry size (log) (0.0098) -0.133*** -0.134*** -0.226*** Concentration Index (0.0439) (0.0439) (0.0517) Ownership (reference group: private domestic) 0.0268* 0.0267* 0.0264 State-owned (SOE) (0.0162) (0.0162) (0.0162) 0.0708*** 0.0708*** 0.0704*** Foreign-owned (FOE) (0.0166) (0.0166) (0.0166) Dispersed capital 0.0311 0.0311 0.0310 (0.0399) (0.0399) (0.0399) a 0.715*** 0.715*** 0.715*** Undefined  (0.0178) (0.0178) (0.0178) 2.132*** 2.177*** 1.643*** Constant (0.0658) (0.0927) (0.190) Observations 471,644 471,644 471,644 R-squared 0.077 0.077 0.077 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes Source: Statistics Poland calculations. Note: Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. a. Less than 1 percent of firms with undefined ownership status are used for identification in the fixed-effects regressions, meaning that they are changing ownership status between years which is captured by the regressions. Those companies appear to be the most productive in the sample, but it is likely a statistical artifact. (See Table 3 for details). 89 Paths of productivity growth in Poland: a firm-level perspective TABLE A6.3  Fixed-Effects Regression of TFP on Competition Indicators by Firms’ Characteristics (Age and Ownership) in Construction and Services (2009 – 19) a. Manufacturing Dependent variables TFP Specifications (1) (2) (3) Competition 0.00396 0.0102 Number of firms (log) (0.0206) (0.0203) 0.443*** Industry-level profit margin (0.0956) 1.895*** 1.904*** 1.982*** Concentration Index (0.264) (0.266) (0.261) Ownership (reference group: all except foreign-owned) a -0.0007 -0.0007 0.0025 Foreign-owned (FOE)  (0.0223) (0.0223) (0.0214) Firm age (reference group: young firms) -0.0482*** -0.0482*** -0.0527*** Age (4 – 9 years) (0.0084) (0.0084) (0.0082) -0.0624*** -0.0624*** -0.0661*** Age (10 – 19 years) (0.0097) (0.0097) (0.0095) -0.0614*** -0.0614*** -0.0641*** Age (20+ years) (0.0106) (0.0106) (0.0105) 2.557*** 2.524*** 2.383*** Constant (0.126) (0.211) (0.233) Observations 216,780 216,780 216,780 R-squared 0.060 0.060 0.091 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes b. Construction and Services Dependent variables TFP Specifications (1) (2) (3) Competition -0.0083 -0.0082 Number of firms (log) (0.0098) (0.0098) -0.0003*** Industry-level profit margin (0.0001) -0.120*** -0.121*** -0.121*** Concentration Index (0.0434) (0.0434) (0.0434) Ownership (reference group: all except foreign-owned) b -0.0087 -0.0087 -0.0087 Foreign-owned (FOE)  (0.016) (0.016) (0.016) 90 Appendix Dependent variables TFP Specifications (1) (2) (3) Firm age (reference group: young firms) -0.0393*** -0.0393*** -0.0391*** Age (4 – 9 years) (0.006) (0.006) (0.006) -0.0578*** -0.0577*** -0.0575*** Age (10 – 19 years) (0.0069) (0.0069) (0.0069) -0.0544*** -0.0544*** -0.0542*** Age (20+ years) (0.008) (0.008) (0.008) 2.474*** 2.530*** 2.529*** Constant (0.0663) (0.0932) (0.0932) Observations 471,644 471,644 471,644 R-squared 0.052 0.052 0.052 2-Digit industry FE yes yes yes Year FE yes yes yes Region FE yes yes yes Source: Statistics Poland calculations. Note: Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.10. a. Introducing ownership and size dummies one at a time makes the relationship between foreign ownership and pro- ductivity growth statistically significant. (See Appendix Table A6.2, Panel b.) b. Introducing ownership and size dummies one at a time makes the relationship between foreign ownership and pro- ductivity growth statistically significant. (See Appendix Table A6.2, Panel b.) A.7  Selected Characteristics Based on the Aggregate Data FIGURE A7.1  Foreign Direct Investment to Poland 7 Foreign direct investment, net inflows 6 5 (% of GDP) 4 3 2 1 0 Source: Elaboration based on World Development Indicators data. 91