68742 INGENS - Income Generation through Energy and Complementary Services STATUS REPORT June 2008 A joint study by Gesellschaft für Technische Zusammenarbeit (GTZ) and Energy Sector Management Assistance Program (ESMAP) Copyright © 2008 All rights reserved Requests for permission to reproduce portions of this internal status report to GTZ and ESMAP should be sent to both Task Team Leaders and to the authors of the respective chapters. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s). CONTENTS 1. Introduction ........................................................................................................... 1 2. Methodology .......................................................................................................... 8 3. Status and First Findings from Country Studies ............................................. 31 A. BENIN: The Impact of Electrification on Micro-Enterprises in Rural Benin32 B. GHANA: Summary of Key Findings from MSE Survey in the Brong Ahafo Region ............................................................................................... 39 C. NIGERIA: Summary of the Key Findings from the MSE Survey in Lagos State .................................................................................................. 46 D. UGANDA: The Impact of Electrification on Micro-Enterprises in Rural Uganda .............................................................................................. 52 E. SOUTH AFRICA ....................................................................................... 59 4. Emerging Overall Results and Lessons from Country Studies ...................... 60 5. Outlook: Next Steps and Funding Request ...................................................... 64 Draft Bibliography ................................................................................................... 67 ANNEX: THE INGENS QUESTIONNAIRE .............................................................. 75 ANNEX: INGENS Enumerator Training ................................................................. 76 ANNEX: Lessons Learned from Ghana Fieldwork ............................................... 77 ANNEX: Productive Uses in Project Practice – Case Examples ........................ 84 ANNEX: Some Pictures from INGENS Cases ....................................................... 85 ANNEX: Budget Estimates for Outlook............................................................... 111 iv INGENS STATUS REPORT JUNE 2008 Preface The present document is an internal status report to ESMAP and GTZ which summarizes the ongoing work under a joint study, which is funded by BMZ and ESMAP. Acknowledgments The INGENS study is implemented by a joint Task Team under the supervision of Lucius Mayer-Tasch (GTZ), Mohua Mukherjee (WB) and Kilian Reiche (GTZ lead consultant). The study’s main authors are listed at the start of each chapter. In addition, we would like to thank the following contributors, reviewers and interviewed experts: tbd for final report. vi INGENS STATUS REPORT JUNE 2008 Abbreviations and Acronyms TBD for final report Units of Measure Executive Summary TBD for final report 1 1. Introduction By Kilian Reiche “Productivity isn't everything, but in the long run it is almost everything.� - P. Krugman 1.1 Although modern energy is an important input factor for growth, poverty reduction and achievement of the Millennium Development Goals (MDG),1 the energy sectors in least developed countries (LDC) are in need of improvements. Generation capacities and service quality are often unsatisfactory; tariffs and fuel prices are often heavily subsidized; and about 1.6 billion people around the world lack access to electricity. This limits GDP growth and has direct effects for household welfare.2 The World Bank PRSP Sourcebook identifies five goals for energy development with positive effects on poverty: (i) expanding access; (ii) improving supply reliability; (iii) ensuring fiscal sustainability; (iv) improving sector governance and regulation; and (v) reducing health and environmental costs. 1.2 High hopes during the early stages of power sector reform have been overly optimistic regarding simple solutions: today we know that private sector (funding) alone will not suffice to improve LDC energy sectors.3 Improved access to affordable energy to new users has proven especially unattractive to private sector, as the most profitable customers usually already have service, leaving mainly those who are unable to pay for connection and use. Rural areas are the 1 GTZ 2007, UNDP 2005, World Bank 2004 2 ESMAP 2003, Estache 2006 3 Besant-Jones 2006, World Bank 2007, AFDB 2008 2 INGENS STATUS REPORT JUNE 2008 most unattractive market segment, due to low demand densities at high costs. The development community has recognized the crucial role of promoting energy access as an explicit element of successful poverty reduction strategies. Bilateral and multilateral donors have increased their energy portfolios over the last five years, often with specific attention to increasing access to modern energy. 1.3 However, energy sector operations, in general, and Electrification Programmes (EP) in, particular, frequently yield less than optimal results.4 Typical problems of past EPs include: low connection rates; lower than predicted energy demand of connected users over time; slow disbursements - and a lack of evidence on development impact in the target region, especially regarding increased productivity. 1.4 Policy makers and practitioners largely agree that improved attention to uptake, productive use5 and “complementary services�6 during design and implementation could improve the impact of EP. Practitioners often refer to a set of “common sense-based� causalities to underpin this hypothesis - but most of these are not based on systematic evidence or immediately pose further questions. Box One lists typical practitioners’ questions around the energy-productivity nexus, some of which have originally prompted the present study. 1.5 The literature provides virtually no robust evidence on the hypothesis under 1.4 - nor on the impact of EPs in general. This is because EPs rarely implement rigorous impact evaluation. This striking absence of evidence has recently prompted the World Bank’s Independent Evaluation Group to demand better EP impact evaluation.7 1.6 To produce such evidence in the future, two basic approaches are possible: (i) ambitious stand-alone research, typically on national level and with project-independent funding (examples of such stand-alone research include LSMS, national firm-level surveys, large household energy surveys, broader infrastructure analysis (such as the Africa Infrastructure Country Diagnostics AICD, the PPI database or Chong/Hentschel/Saavedra 2003) or other studies 4 World Bank 1995, IEG 2008 5 FAO 2000, Fishbein 2003, IEG 2008 6 Motta/Reiche 2001, Peters/Harsdorff/Ziegler 2007 7 IEG 2008: “The evidence base remains weak for many of the claimed benefits of [rural electrification]. Tailor-made surveys, designed to test these benefits, need to be built into a greater number of Bank projects and designed to allow rigorous testing of the impact of electrification. […]While stimulation of productive enterprise is claimed to be among the benefits of electrification, few studies have tried to quantify these benefits using an impact evaluation methodology� Introduction 3 with large data panels) or (ii) more modest add-on modules for better impact evaluation in new and ongoing EPs, usually with project funding. The former are relatively expensive (typical costs of type (i) national surveys range from US$ 0.5M to US$1M) and rarely implemented.8 The latter have a huge potential for closing the EP knowledge gap identified above, as they cost less and could be made standard requirements for future EPs. However, examples are scarce, probably because it is difficult to strike the right balance between cost and minimum requirements for reliable results, or because EP managers shun the transaction costs of developing their own add- on survey instruments.9 1.7 INGENS has proposed (and started to test) a pragmatic, low-cost approach to improve impact evaluation for future EPs: an example for a simple, “Add-on Electricity Impact Evaluation Instrument� (AddIE) for the specific case of firm level productivity which shall (i) be applicable to a broad range of EPs10 and (ii) cost well under US$100k per EP. This tool allows EP managers to move from simple monitoring and reporting of numerical targets achieved to evaluation of the (pro-poor) impact achieved among an important segment of the target group. 1.8 In Phase One, INGENS has produced a pilot survey instrument suited for evaluating (A) the impact of electricity access on MSME performance and (B) the potential of complementary services (especially finance and BDS services) to “boost� such impact.11 Obviously, other EP “AddIEs� will be needed for a better understanding across the full range of expected EP impacts and country contexts (for instance for impacts on the all-important household level, or the role of energy in food production). A recently approved parallel ESMAP study has identified the same gap we have described in 1.5 and will produce such additional add-on M&E tools with a focus on the household level (Barnes/Khandker - see chapter 5 for synergies). 1.9 To test the proposed “AddIE� instrument under real-life conditions (and to minimize ESMAP funding for Phase One), it was decided to try out the draft questionnaire in actual ongoing EPs in Africa, using local consultants wherever possible. 8 However, it would make sense to implement such stand-alone studies as regularly scheduled multi-donor country efforts in the future (an idea originally developed by D. Barnes). 9 Foster 2002 proposes a similar approach to energy subsidy research and energy demand surveys. 10 The immediate application would be in the BMZ/DGIS-funded Energising Development programme, other GTZ- executed EPs and the World Bank’s Africa Action Plan. 11 This corresponds to two Hypotheses: (A) The impact of electricity access on SME performance (and thus regional productivity) is positive and significant. (B) Complementary services (such as microcredit and BDS) can significantly increase this impact. 4 INGENS STATUS REPORT JUNE 2008 1.10 Several EPs in Africa have accepted this draft instrument and are currently applying it, with project co-funding. The INGENS country cases are: Benin (GTZ case), Ghana (GTZ case), Uganda (GTZ case), Nigeria (WB case) and South Africa (WB case). 1.11 Emerging lessons from INGENS Phase One are: a. Literature with quantitative data on the nexus between MSME productivity, energy access (and usage) and complementary services is virtually non-existent. b. Substantial modifications were required to adopt the survey instrument to country boundary conditions, EP schedules and budgets. This has limited the possibility of advanced cross country comparison, although efforts were made to maintain a “core methodology� throughout all country cases. c. Due to budgetary restrictions, it was impossible to include valid control groups for BDS and microcredit in the first surveys. In addition, it is extremely difficult to find areas without access to BDS or microcredit that might serve as comparison groups in a pilot survey. In phase one, electricity access therefore is the only treatment for which robust results can be derived. Thus, impacts of electricity access can be investigated in several countries (Hypothesis A, see 1.8). For the impact of other services and their interactions, only simple correlations can be observed that do not allow deriving conclusions about causal relationships. d. Capacity of local consultants was very low (while consultant fees were relatively high compared to other regions) and significant capacity building was required on the go. This has sometimes delayed implementation and has reduced the content of the specific survey results in some country cases. e. In our case studies, it was difficult to find hard evidence on actual causalities between electricity access and firm performance, be this measured by profits, employment, or other indicators. However, “absence of evidence is not the evidence of absence� of such causalities - hence, more intensive work on this issue is clearly warranted. Looking beyond access as a binary variable and analyzing the level of usage on the firm level in more detail might be an option to gain better insights. Introduction 5 1.12 The present document is merely a status report, as fieldwork in several countries is still ongoing and data analysis has not been finalized yet. However, some first interesting findings are given in chapter 3 and chapter 4. The INGENS core methodology is summarized in chapter 2; and the AddIE questionnaire is given in Annex One. 1.13 A Final Report is scheduled for FY09, which would be jointly produced by GTZ and the World Bank, assuming additional funding for the pending tasks on the ESMAP-side can be found.12 In case no additional ESMAP-funding were available, GTZ funding would suffice to provide the following “minimum version� of the “final report: a. present final survey results and data from about five African country cases (country list see above; Nigeria and Uganda need additional field visits to improve analysis); b. draw first conclusions regarding the future use of the proposed Add-On Evaluation Instrument “AddIE� for the nexus of electricity access and firm performance; c. provide a “background section� with underlying methodological, economical and practical issues of relevance for task managers who are interested in adopting these Add-on Impact Evaluation instruments to their EPs. 1.14 Additional ESMAP funding is sought for finalizing the full Final Report jointly with GTZ (details see chapter 5). This full report would allow to (i) add a “Practitioners’ Manual� with recommendations on better design of EPs by adding complementary services and focusing on rural productivity; (ii) cooperate more closely with two recently approved parallel ESMAP studies (“Africa Electrification Experiences Initiative - AEE� and “Monitoring and Impact Evaluation of Rural Electrification�) and present results at the ESMAP AEE workshop in early 2009; (iii) include a new country study on Senegal (also by early 2009); and (iv) conduct advanced analysis for several of the existing cases in order to interpret some of the more controversial emerging results. 12 Several of the local consulting firms have delivered less than expected quality in their analysis, so that additional survey experts had to be hired. In addition, findings from the first round of field visits make it desirable to go back for complementary survey work in new target areas so as to allow for advanced data analysis. 6 INGENS STATUS REPORT JUNE 2008 BOX 1: Electrification Impact on Firm Performance – typical task manager questions  Can energy access loans/grants/projects only be successful if additional revenue is created?  Is a lack of “productive electricity uses� the main reason why (rural?) tariff schemes are often not cost covering?  Does “Productive Use� require increased productivity?  The really promising business ideas have probably already been picked up by the more promising entrepreneurs, and if those need electricity they probably already have a generator. So, can there be a positive effect of grid “access� at all apart from windfall profits to those using generators, by lowering their cost of electricity?  Is it a satisfactory impact if one local vendor sells more small consumer goods to local customers because his products are illuminated, while his competitors don’t have electricity – or is this only local redistribution of income, without net increase in value added? Is “export� necessary to create local value? Which part of regional productivity increase is typically due to (i) “exports� to outside markets (trade); (ii) changes in local industry structures, (ii) increased firm efficiencies (due to technology changes); (iv) increased local consumer demand and total output (without improved firm efficiencies), and (iv) multipliers and spill-over effects? How do local value chains work?  Is it more important to create local income or more employment?  What’s more promising for my project: lighting for household-based micro enterprises (who can extend their working hours); cell phones for small farmers (who can check market prices and reduce transaction costs); or machines for medium-sized food producers?  How can I measure (firm-level, local, regional) value added if no owner answers my questions truthfully?  Is it “access� if several small workshops use the machinery of a larger “service supplier� in the area who has a diesel genset?  Should appliances be targeted by my EP? How - with TA, grants, microcredit lines – or refinancing through the utility bill?  Should an EP aim at maximizing productive demand (measured in kWh) to maximize utility profits or at efficient energy usage?  What’s more promising for increased productivity: Energy or Electricity?  What’s the difference between “complementary services� and Integrated Rural Development of the 70s?  Which are the most important success determinants of EP (of SME): BDS, microcredit, road,, ICT – or others (leadership, investment climate, energy sector conditions, etc)  Can you give me some concrete examples of productive uses that work?  What do we know about lighting, ICT and machinery as intermediary outputs of firms’ production? Can we measure a Consumer Surplus for firms? Introduction 7 2 2. Methodology By Jörg Peters and Colin Vance (RWI Essen) Objectives The rationale underlying this study is a better understanding of how to boost productive energy demand and thereby increase income generation and employment in regions targeted for electrification projects. Providing access to modern energy services is ultimately about improving human welfare, and although this objective underpins most development projects, its evaluation is fraught with methodological challenges. These range from the complexities inherent in designing surveys that quantify the net impacts of electrification projects to the selection of appropriate analytical techniques. To facilitate measurability, Ravallion (2008) argues in favor of investigating intermediate outcomes that are of direct relevance to welfare. One possible intermediate outcome is per-capita income growth,13 which can be measured by drawing data on the income of households directly. While recognizing the importance of household surveys in measuring welfare improvements (World Bank 2002 and Bensch and Peters 2008), this study investigates the potential contribution of small businesses to this development process. Therefore, the focus is on indicators at the firm level that are assumed to ultimately increase per-capita income. 13 The underlying assumption is that increased income does not induce other negative impacts on human welfare that outdo the achieved positive income effects. Such negative impacts might be substantial environmental destructions or strong distributional distortions. Moser (2001) provides evidence that economic growth translated into an improvement of living standard indicators. Methodology 9 In addition to analyzing the impact of electricity provision on firm performance, the study also considers how this impact may be mediated by two ancillary services: (a) business development services (BDS)14 that help businesses to better utilize their energy access or inform them of possible new income-generating uses and markets; (b) financing services that help to cover the energy-related up-front investment costs of Micro, Small and Medium Enterprises (MSMEs), such as the investment costs of new machinery.15 This secondary focus aims to assess the widespread consensus among practitioners that providing electricity alone is not sufficient. There are a number of channels through which electricity and productive energy use might translate into economic growth and, hence, into improved welfare.16 In the long run, the most important impetus is the increase of productivity, as this has a direct and positive impact on the profitability of hiring labor and, ultimately, on wages. Higher labor productivity makes labor less costly, so that firms will find it profitable to expand employment (Walsh 2004). Moreover, while productivity reduces the cost of labor relative to other factors in the short term, the long term expansion of employment increases wages as firms compete to hire labor. An additional means through which productivity contributes to growth in rural areas is via increased labor input. In developing countries, electricity can mobilise a significant amount of labor by reducing the workload of time-intensive household activities like milling. Furthermore, improved access to lighting services can prolong working hours per day and thereby raise labor input. Shop owners, for instance, may prolong working hours because electricity provision renders the additional working time more profitable. This is often the most important effect of electrification projects in the short run, since it is frequently observed that businesses extend their working hours due to improved lighting services. 14 For the purpose of this study, Business Development Services include training and advisory services, marketing assistance, and business linkage promotion with the objective of improving MSMEs’ critical business processes. For other definitions, see CDASED 2001 and Motta/Reiche 2001. 15 Other services that potentially enhance the effect of energy projects include transportation, telecommunication, irrigation infrastructure, or general technical assistance. Yet, in order to keep relations between indicators measurable, such services are included only for control purposes and are not investigated extensively. The focus of the survey is instead on a narrow definition of BDS and MFI. 16 While it is beyond the scope of this study to join the controversial discussion about the „true“ impetus to growth, the literature on this debate can offer some useful insights. See Abu-Qarn and Abu-Bader (2007), Easterly (2001) and Wacziarg (2002) for a survey of theoretical literature. 10 INGENS STATUS REPORT JUNE 2008 With regards to capital accumulation, economic growth can additionally be driven by increasing savings and thereby investments. Again, the increased productivity makes further capital input profitable. It is often argued that modern energy provision removes bottlenecks to the extension of capital in firms and stimulates investments in new machinery, thereby increasing output and enhancing the productivity effect (World Bank 2007). However, increased labor productivity and capital accumulation might induce painful effects in the short run, as both can cause net decreases in labor demand. While the liberated work force could theoretically be employed in more productive ways, this potential is limited in developing countries due to restricted education and market access. A possible solution to this dilemma might be an integrative approach that complements rural electrification with, for example, BDS (Peters, Harsdorff and Ziegler 2007). A. Conceptual Framework The proceeding discussion illustrates how the usage of electricity, as well as other services, can translate into per-capita income growth through improved performance of MSMEs. Performance, in turn, can be described by a standard production function f that determines the firm’s output Y as a function of the complementary factors capital K, labor L and services S. Y  f ( L, K , Si ) (1) While Si represents various services like transport or telecommunications, our focus is on BDS (SB), MFI (SF) and in particular electricity (SE). Given the complementarity of these factors, an increase in, for example, energy infrastructure leads to higher productivity of the other factors.17 For our research objective it is important to highlight the simultaneity in this model. While it is intuitive that the usage of any service Si increases output Y, causation also runs in the reverse direction. Firms that perform better are more inclined to connect to the electricity grid or use BDS offers. Therefore, we face a system of simultaneous equations consisting of (1) and 17 See Straub, S. (2008) for presentation of the general case of infrastructure in a production function. Methodology 11 Si  g i (Y , K , L, Z S ) (2) In other words, a firm’s decision pertaining to the use of a service depends on output, capital, labor and, potentially, a vector of additional determinants ZS, (e.g. distance to the grid, sector, etc). Furthermore, capital and labor also depend on output and usage of different services: K  h1 (Y , Si , L, Z K ) (3) L  h2 (Y , Si , K , Z L ) (4) Indicators for the output Y can be figures that measure the volume of economic activity, such as sales or profits. While this is straightforward, it remains an open question as to how exactly to collect information to determine these variables. While some survey practitioners argue in favour of eliciting profits or sales by a direct question, others opt for calculating profits from income and expenditure figures.18 The justification for this latter approach is that many respondents feel uncomfortable about revealing their income due to tax-related concerns or to an aversion of exposing details about their financial situation. Furthermore, directly reporting output figures requires some ad-hoc calculation, since many entrepreneurs do not keep books and are not aware of the actual figures. Indicators for labor can be either the number of workers or, if possible, working hours summed up over workers. Simply counting the employees is in many cases very inaccurate, since MSMEs frequently hire workers on an irregular, stand-by basis. Therefore, this study elicited information on working hours per employee. In addition, we accounted for seasonality by asking for employment changes over the year. Like output and employees, information on the value of the capital stock can be collected in the aggregate or itemized. As the former approach transfers the calculation burden to the respondent, we instead opted for eliciting each asset individually. This still requires some effort on the respondent’s part, since the monetary value of the individual assets has to be assessed. In order to 18 See S. De Mel, D. McKenzie and c. Woodruff (2007) and Daniels (2001). 12 INGENS STATUS REPORT JUNE 2008 avoid recall errors and depreciation assumptions, we asked for the resale value of each asset. To capture long-run effects on growth, we elicited information on investment outlays made since receiving electrification. Although it is only an intermediate measure with respect to firm performance, it indicates long term growth prospects. Furthermore, comparing investments since electrification to investments before suffers relatively little from endogeneity, although anticipation effects have to be taken into account (see section Identification Strategy). The definition, measurement, and interpretation of the service variables Si is crucial for the purposes of this study. There are three possible definitions. The first would be to interpret Si as an access variable, meaning that Si equals one if the firm is in a region that is principally covered by a service provider and Si equals zero otherwise. According to this understanding, SE would equal one for a firm situated in a grid-covered region without being connected, while SE would equal zero for a firm disposing of a generator and situated in a non-grid-covered region. This interpretation is in line with the definition used by most rural service programs, which typically aim to generate benefits for the whole region via spillovers, not only those firms that are served directly. One might, however, also be interested in the effect of directly receiving a service. In this case, Si equals one if the firm actually uses the service and Si equals zero if it does not. Now, SE equals zero for the firm situated in a grid-covered region without being connected, while it equals one for the firm in a non-grid-covered region if it uses a generator.19 The third possibility is to interpret Si as a continuous variable by considering the level of usage. SE would than represent the amount of energy, measured in kWh-equivalents, for example. In this case, one does not need to differentiate between connected or non-connected firms. The difference would rather be reflected in the level of SE. Energy consumption of traditional appliances like petroleum lanterns would have to be translated into kWh-equivalents, which is, in 19 One might, of course, differentiate between generator usage and grid usage, depending on the quality both deliver. This is a notion with important implications for the study at hand, as it shows that access (to grid electricity) is in reality only a matter of quality and costs: for instance, rural firms in areas not covered by the grid that realize the importance of electricity have the opportunity to buy off-grid electricity capacities. The main difference between, for example, generators and the grid are rather quality and, ultimately, costs. Methodology 13 principle, easily done. In order to maintain this possibility, we collected detailed data on usage and consumption of both electric appliances and traditional ones. Additionally, the quality of service provision can be taken into account and reflected in Si. This is particularly appropriate when assessing the frequently unreliable electricity grids in Africa. One indicator for this purpose, for example, is the number and duration of unannounced outages per month. B. Identification Strategies The key aim of this study is to determine empirically the effect of Si on Y. The different interpretations presented in the section “Indicators� pose different problems for identifying Si in an empirical model. For reasons of clarity, we focus our discussion of how to identify Si on the questions of access and use, although our conclusions are in some cases transferable to the level of usage, as well. In principle, what we would like to know is what happens to the firm’s output if it is connected or has access to the Service i, i.e. Si equals one. In the following, we refer to this as the treatment. For this purpose, one has to compare the outcome variable after having received the treatment to the counterfactual situation of not having received it. Following Frondel and Schmidt (2005) and Ravallion and Chen (2008),20 we denote the post-treatment output by Y if the firm hypothetically had not received treatment and by Y+G if is has received treatment. G is therefore the gain attributable to the project and reflects the causal effect. While we present several strategies to identify the causal effect that make use of either cross-sectional or over-time data, for this study we are limited to the usage of cross sectional approaches (see Applied Research Approaches section). Since the conducted surveys serve as a baseline for ex-post evaluation at the same time, it is nevertheless worth discussing the whole set of identification possibilities. As a matter of course, the frequency of outcome Y and Y+G across the population of firms depends on a set of characteristics X (e.g. firm size or education of firm owner). The interest in an impact analysis now is on the average individual output change resulting from the project 20 Frondel, M. and C. M. Schmidt (2005), Evaluating Environmental Programs: The Perspective of Modern Evaluation Research. Ecological Economics, 55 (4), 515-526 14 INGENS STATUS REPORT JUNE 2008 intervention, which is often called the mean effect of treatment on the treated: M  E (Y  G | X , Si  1)  E (Y | X , Si  1) (5) where E(.) denotes the expected values. As is obvious, we can never observe Y and Y+G simultaneously for the same firm, since it either receives the project’s treatment or not. While E(Y+G|X,Si=1) can be easily estimated from a sample of treated firms, E(Y|X,Si=1), which measures the hypothetical output of these treated firms had they not been treated, is not observable This is what Frondel and Schmidt (2005) refer to as the core of the evaluation problem. To solve this, we have to formulate assumptions that allow replacing the unobservable and, hence, not estimatable E(Y|X,Si=1) with something than can be obtained by estimation from any existent dataset. In practice, this is only possible by finding a comparison or control group that serves to simulate the counterfactual situation for the treatment group. A frequently pursued approach is the before-after comparison, where E(Y|X,Si=1) is replaced by E(Yt-1|X,Si=1), i.e. the comparison group comprises the treated firms themselves before the implementation of the project. For example, the output of an electrified firm is compared with its output before electrification. The identification assumption in this case would be: E (Y | X , Si  1)  E (Yt 1 | X , Si  1) (6) That is, one assumes that the firm’s output would not have changed from t-1 to t if it had not received the treatment. This assumption can be violated if external factors affecting the firm’s output (e.g. national economic growth or crop yields) changed from t-1 to t. While this is not an issue if such factors are readily observable and accounted for, it can result in biased estimates of the treatment’s effects if the factors are not known. Furthermore, one has to exclude the possibility that firms anticipate the treatment and change their business strategy accordingly. For example, a firm might postpone productive activities or investments until the period after electrification. To the extent that this strategy reduces output in period t-1, the estimate of the treatment’s impact would be biased upward. If one can reasonably rule out such behaviour or Methodology 15 control for other relevant influences on output, the simple before-after comparison is a valid approach. In most cases, however, these assumptions are questionable. Since this vulnerability of the before-after comparison stems from the fact that it considers the treated group as control group for itself, one solution is to include non-treated firms in order to determine the counterfactual. This is the approach pursued under so-called difference-in-difference-estimation (DD),21 which in the traditional case compares output changes of firms that have received the service intervention to those that have not, as illustrated in Figure 1. While anticipation effects still cause a bias, DD controls for external changes affecting the firm’s output. Figure 1: The Difference-in-Difference Approach Y YtTreatment Yt Treatment 1 �t �t-1 Yt Control Yt Control Impact G= �t- �t-1 1 Electrification t 21 See again Frondel and Schmidt (2005), as well as Ravallion and Chen (2005). 16 INGENS STATUS REPORT JUNE 2008 Furthermore, unobserved heterogeneity between firms that is constant over time is automatically accounted for by calculating the differences in output for both treated and non-treated firms. This is reflected in the weaker identification assumption under which the output change of treated firms in the hypothetical no-project-intervention scenario equals the output change of non-treated firms in the no-project-intervention scenario. E (Y  Yt 1 | X , Si  1)  E (Y  Yt 1 | X , Si  0) (7) Remember that the first expression is by nature not observable, while the latter can easily be estimated from a control group sample. Returning to the different definitions of service Si presented in the above section, we encounter different identification possibilities using the DD-approach. Applying the access interpretation of service-providing projects, we require two regions that have to be surveyed before and after a project intervention: One that is not yet covered by a service provider, but that will receive access to the service soon (treatment group), and another that neither has nor will receive service coverage (control group). In order to meet the identification assumption (7), both regions have to fulfil certain conditions (see Box 1). Methodology 17 Box 2: Selecting Comparable Treatment and Control Regions Successful impact evaluation strategies require a survey design to gather data that allow for identifying counterfactual situations. Both difference-in-difference estimation (DD) and the cross-sectional approach call for adequate selection of control and treatment regions. In the case of DD, the treatment region is the target area of a service provision project that is not covered by the service at the outset. The control region does not and will not have access to the service. In the cross-sectional case, the treatment region already has access to the service, while the control region does not. In the case of an ex-ante impact analysis, this means that the target area of the intervention is referred to as the control group. In selecting treatment and control regions, it is crucial to assure sufficient comparability. Village level parameters like population, political importance, and access to other roads, transport or telecommunication have to be checked in both regions. Most importantly, the business environment has to be similar. This can be ensured by taking account of local market conditions, the availability of cash crops, infrastructure, etc. Generally speaking, differences in local characteristics that also influence firm performance have to be reduced as far as possible. In most cases, regions exhibiting comparable conditions can be identified. Rural Africa is replete with regions having low electrification rates, so that comparable non-electrified regions for the DD-approach should be available. Finding comparable electrified areas for the cross-sectional analysis might be harder. In the usual case, though, utilities and electrification projects follow an either virtual or physically existent priority list in accordance with national rural electrification plans. This list is compiled by taking into account characteristics like road access and business potentials. Therefore, the target areas selected for an electrification project are typically not deprived areas, but are rather similar in economic terms to those regions that were connected in recent years. This, however, might not apply if political considerations outweigh socio-economic indicators in the selection of regions to be electrified. 18 INGENS STATUS REPORT JUNE 2008 For the use interpretation of Si, surveying only the region of the project intervention might be sufficient. The treatment group then would consist of those firms that choose to use the provided service, while the non-users constitute the control group. Both have to be surveyed before and after the intervention. It is worth highlighting that this does not necessarily mean that the same firms have to be interviewed. As Ravallion (2008) notes, only the four means (output in t-1 and t for both control and treatment group) are required to calculate the estimator. These means need not to be calculated for the same sample over time. A problem that arises in applying the DD-approach is that it assumes homogenous reactions to changing external factors. Imagine that the treatment group consists of firms that choose to use the service due to some unobservable characteristic, such as political influence with local officials. Now, consider some environmental parameter that changes between t-1 and t that benefits those firms exhibiting the unobservable characteristic (e.g. the officials secure for their influential patrons cheaper access to particular inputs). This would lead to an overestimation of the impact, since DD ascribes the associated increase in profits among the treated firms wholly to the service provision, ignoring the role of political influence. One practical disadvantage of both before-after comparison and DD-estimation is the need of having data collected before and after the project intervention at hand. In this regard, many projects do not carry out adequate baseline studies at the time of the planning phase prior to the project’s implementation. While the surveys conducted for this study were designed to serve as a basis for such an ex-post analysis, our immediate priority is in generating indications for the impacts of the treatments using the currently available data. Therefore, for this study a cross- sectional approach is applied in order to assess the impacts of the project’s interventions. Furthermore, it is widely accepted among evaluation practitioners that ex-post surveys should be conducted only after sufficient time has elapsed since the firms received the service.22 In the case of electrification projects, consumers – be they households or firms – need time to adapt to the new situation. The monitoring horizon of electrification projects, though, typically only covers around three years, including the planning phase before the actual realisation of the project. To 22 ESMAP (2003). and Ravallion and Chen (2005) also highlight the fact that many effects of development projects emerge beyond the usual monitoring period, but argue in favour of evaluating intermediate indictors of long-term impacts. Methodology 19 some extent, cross sectional comparison affords a solution to this problem, as it captures the long term effect of a treatment. The intuition is that a treatment group comprising villages that are electrified since, say, five years simulates economic performance of the firms in the project’s target region five years after they have received access to electricity. Additionally, changing environmental parameters like GDP development or anticipation effects do not affect the accuracy of results. In formalised terms, this is reflected in the identification assumption for cross-sectional comparisons: E (Y | X , Si  1)  E (Y | X , Si  0) (8) In other words, it is assumed that electrified firms, if they – hypothetically – had no grid electricity, would behave and develop as the non-electrified do. In order to elaborate the implications for our survey design, we return to our different interpretations of Si. As in the case of DD, we need two regions to investigate the impacts of access to a service. Given a sufficient comparability of these two groups (see Box 1), the identification assumption holds and we are able to estimate the true impact G of access to service provision on the firm’s output. Certainly, the overall output of a region positively affects the likelihood of being electrified in electrification programs. This is how (2) can be interpreted in the case of the access interpretation of Si. However, there is no problem with selection bias as long as criteria like economic conditions, road access etc. are taken into account in selecting control and treatment regions (see Box 1). With regards to the application of the use definition of Si, the identification assumption (8) does not hold if using and non-using firms in the same region are compared. This is because we have to exclude the existence of variables that affect selection into treatment and the counterfactual no- treatment outcome at the same time, given control variables X. Taking the example of electrification, firms that are more motivated or risk-taking might be more inclined to get a grid connection. At the same time, this characteristic would certainly affect their outcome Y in the absence of an electrification project. Hence, comparing connected and non-connected firms from within the same region means comparing non-comparables. Differences in Y would be assigned to 20 INGENS STATUS REPORT JUNE 2008 the project, even though they are in fact due to some other unobservable differences in characteristics. An additional problem is the simultaneity reflected in (2). Being connected to the grid undoubtedly has a positive effect on output, as stated in (1). At the same time, however, better performing firms exhibiting a higher output are more likely to have the funds to get a connection. Again, this leads to a comparison of non-comparable firms. Altogether, only investigating the difference of performance between firms that use a service Si and those who do not leads to a strong selection bias.23 In methodological terms, the variable Si is endogeneous, since it is a choice variable of the firm that is a) determined simultaneously with output (see equations 1 and 2), and b) potentially correlated with non-observable variables that also affect output. For the cross-sectional approach, a solution to this problem is to find an identification variable that is correlated with the use of the service but is not correlated with firm output. Generally, such so-called instrumental variables (IV) are hard to find. For development projects, though, Ravallion (2008) proposes a procedure that – in theory – can be applied to many service interventions. This procedure creates a dummy variable, denoted here as service coverage (SC), that distinguishes regions according to whether they have access to the service.24 To the extent that SC is correlated with service use, but not with output Y or other unobserved variables, it can serve as an IV.25 In doing so, the selection process for service provision has to be considered. While overall economic performance of a region certainly plays an important role, this does not pose a big problem if the target region of the electrification project was selected according to macro economic indicators or an electrification priority list (see Box 1). This IV-approach can correspondingly be applied to the third interpretation of Si, the level of usage, a variable that is also likely to be endogeneous. As firms that are located in a region covered by the project are more likely to consume more of the service, SC again can serve as an instrument for level of usage. 23 This holds true for most other intermediate development indicators like household’s education, health or income. 24 See also S. M. MacKernan, Ravallion and Chen (2005) and M. Ravallion and Q. Wodon (1998). 25 While it might be argued that the grid attracts firms with some unobservable trait such as motivation, this effect is unlikely to be strong. Methodology 21 C. Applied Research Strategy The surveys conducted in the framework of the overall study serve foremost for analyzing cross- sectional differences between electrified and non-electrified regions before the implementation of electrification interventions. The results presented in this report can therefore be regarded as an ex-ante impact analysis (OECD 2007). For this reason, the surveys were intended to cover at least two regions: One with access to the electricity grid and one without. In order to additionally measure causal results for the two other services, BDS and MFI, we would have had to include five further regions: A region with access to only BDS, MFI or electricity respectively and all combinations thereof. As this was far beyond the budgetary possibilities, the focus is in most country case studies on measuring causal relationships between electrification and firm performance. The empirical analysis will avail both descriptive statistics and econometric models, with both servings as complementary tools for discerning causality. Referring back to Equation 1, we will explore alternative functional forms for modelling production, including the Cobb-Douglas and translog production functions (Blalock and Veloso 2007). Moreover, close consideration will be given to implementing the appropriate estimator. For example, a switching regression approach that orders observations into treated and non-treated groups may be employed if sample selectivity is deemed to be a problem (Vance and Geoghegan 2004; Ravllion 2008), while instrumental variable methods will be considered for handling endogeneity. Techniques that combine these two approaches may also be explored (Vance and Hedel 2008). In addition to the cross-sectional analysis conducted for this report, the surveys provide a basis for over time evaluation using before-after or DD methods. With regards to DD, though, the approach applicable in the future deviates from what was presented above. The reason is that in the usual case, two groups are surveyed as a baseline that are not electrified (i.e. untreated), one of which is going to receive the intervention. In our case, we have one region without and one with grid electricity today. The region without electricity will be connected to the grid in the years to come. Nevertheless, as illustrated in Figure 2, the already electrified region provides for a benchmark that allows comparing differences and thereby including external effects that have to be netted out. This can be applied both to the interpretation of Si as access or use of electricity. 22 INGENS STATUS REPORT JUNE 2008 Figure 2: The Difference-in-Difference using a treated (in t-1) Control Group Y Yt Control Yt Control 1 �t �t-1 Yt Treatment Yt Treatment Impact G= �(t-1) - �t 1 Electrification t D. Matching Approaches The idea of matching approaches is to improve the comparability of control and treatment groups. For this purpose, the firms from one group are matched to those from the other according to specific observable characteristics ZS. This can be either done individually so that exact matches are determined (See Frondel and Schmidt 2005), or as a group, which results in the creation of subgroups that are more comparable than the groups in total (Ravallion and Chen 2008). In the case of exact matching, the outcome indicator Y is compared for both firms to determine the individual gain. These individual gains can be summed up to obtain the average overall gain. If matching groups are created, these can be, in principle, used as described in the section above. Matching can thus be combined with before-after comparison, the DD- or the cross-sectional approach. The crucial step in applying matching approaches is the identification of appropriate characteristics ZS to base the matching on. In over time approaches like before-after comparison and DD, the pre-intervention outcome Yt-1 is a valuable source of information to create subgroups. Thereby, unobservable factors that are associated with the pre-intervention outcome can be Methodology 23 accounted for. In particular, for before-after comparison of connected and non-connected firms, the simultaneity bias resulting from (2) can be reduced. In addition to the pre-intervention outcome, further characteristics can be taken into account. These characteristics can either be used equally or weighted to find matching partners. One procedure to find weighted matches is the application of propensity scores. This method selects matching partners according to their probability to join the intervention P(Si=1| ZS). To do so, in a first step a probit model of program participation is estimated. It is important to highlight that ZS have to be variables that are not responsive to the intervention. This implies a challenge for the case of cross-sectional comparison, where the pre-intervention outcome Yt-1 is not available: Variables have to be found that are time-invariant and not supposed to be non-responsive to the intervention. At the same time they have to be part of ZS, meaning that they are determinants of the decision to participate (Si=1). Examples in the case of electrification might be the education of entrepreneurs, which can be assumed to be not affected by electrification in the short run. E. Further Methodological Issues Cross-regional approach While some (mostly qualitative) case studies on the productive usage of energy have been undertaken for particular projects, comparative appraisals for different countries or regions are difficult to come by. One purpose of this study is to fill this gap. A cross-regional approach can potentially reveal important insights about how different stages of economic development condition the effectiveness of energy provision. The target region of intervention in Uganda, for instance, exhibits very few commercial activities and nearly no manufacturing. By contrast, the Ghana target regions are industrial zones composed of firms that have a higher level of specialisation and mechanisation. If one considers these conditions as development stages instead of static regional attributes, it opens up opportunities to derive insights into the meaning of energy in different phases of economic progress. 24 INGENS STATUS REPORT JUNE 2008 Definition of MSME While several criteria for defining MSMEs are found in the literature (e.g. magnitude of turnover, number of workers) there is no universally recognized metric that applies to all contexts. For the present study, the aim was to first assure demarcation from home businesses. Home businesses were excluded in order to avoid further heterogeneity induced by households that only pursue commercial activities occasionally. Notwithstanding their fundamental importance for the rural economy, the induced heterogeneity would not be controllable given the relatively small sample sizes. Consequently, one precondition was that the business is in a permanent structure or kiosk outside the home. The categorisation into medium, small and micro enterprises was then done according to the number of workers (including owner-manager and family members): Medium Enterprises: between 20 and 49 workers Small Enterprises: between 5 and 19 workers Micro Enterprises: between 1 and 4 workers . The study focuses on micro and small enterprises. In principle, it is intended to draw representative samples from the population of firms. Concentrating on subgroups (with respect to size and industry) of enterprises that are comparable facilitates controlling for external factors and hence increases the probability of receiving meaningful results. Given the limited sample size and the intention to measure three different types of support (energy, BDS, MFI), it is not possible to get convincing research outcomes if all three firm sizes and many different kinds of industries are included. As illustrated in Figure 1, the number of sub-groups can multiply rapidly, which can pose challenges for statistically distinguishing between the groups. Assuming additionally that two separate BDS types have to be included, the number of subgroups would increase up to 28. Note that Figure 1 does not include the separation for different industries. Methodology 25 While it was envisaged to define uniform industries and firms sizes in all participating countries to allow cross-country comparisons, this strategy turned out to be impracticable because of the heterogeneity of the target regions as well as the project requirements in each country. Therefore, each country study has a different focus in terms of industries and sectors. Sampling To control for the different treatments and firm characteristics, the sample size has to be at least on the order of 200 responding firms per country. If sample size is reduced to fewer than 200 firms per country, included subgroups with respect to firm size and industry have to be further limited. Otherwise this would result in an excessively shallow representation that would not allow controlling for heterogeneity between different subgroups (e.g. size and sector) If the sample size were smaller, e.g. 50 MSMEs per country, it would be necessary to focus only on one or two industries among either small or micro firms. In order to obtain a representative sample, the ideal sample design would ensure that firms identified for inclusion in the sample are represented in approximately the same proportions as their actual representation within the population of MSMEs in the study zone. Presuming that an 26 INGENS STATUS REPORT JUNE 2008 enumeration of industries can be obtained,26 stratification is often a useful technique for achieving this aim, as it provides a means of guaranteeing the representation of particular subgroups of the population that might otherwise be left out if the selection was not stratified and due to pure chance. Unlike simple random sampling, where the elements of the sample are chosen individually and directly through a random process that gives each element an equal probability of selection, stratification divides the population into strata that are independently surveyed. In this way, it is assured that a particular subgroup of interest for the study – however small – is selected and, moreover, that its representation in the sample corresponds proportionately to its representation in the population. In technical terms, the key benefit of stratification is to reduce sampling variability (and improve accuracy) to the extent that adequate representation of the relevant subgroups in the population is achieved. Criteria that can be used for stratified sampling are firm size, sector, turnover, geography, or some combination thereof. Since firm traits are typically not available for the whole population of firms in advance, a pragmatic criterion is the location of the firms, as it is correlated with many of the other aspects. F. Deviations in Country Methodologies While an attempt was made to apply a uniform methodological approach across study regions, strict adherence to this ideal proved impractical, and particular deviations were adapted to accommodate country-specific conditions or project-specific demands.27 Most importantly, the Country Case Studies differ in terms of their identification strategy to measure the impact of service provision. The Benin and Uganda Studies apply the cross-sectional approach controlling for access to electricity and serve as a baseline for ex-post impact assessment at the same time. The Ghana Case was not able to implement the cross-sectional approach due to the specific set up and requirements from the GTZ project. It rather focuses on ex-post evaluation. The same applies to Nigeria, although efforts are still undertaken to find and conclude an appropriate cross-sectional control group. 26 Such an enumeration might be hard to obtain in a comprehensive form. However, it should be tried to get at least an idea about firm distribution within villages. Experience from orther surveys shows that this can be done in a satisfying way in cooperation with e.g. community representatives – although the obtained lists are rarely complete. 27 This is in line with both GTZ and World Bank’s expectation towards the study: one aim was precisely to design and apply a pragmatic, low-cost methodology to evaluate impacts in the context of real projects and their restrictions. Methodology 27 The definition of MSMEs was handled flexibly. While the Ghana team only includes firms that work away from their private home, in Uganda this strict condition would not have made sense, since too few MSMEs would have qualified. In this case, firms were included if they run their business in a separate building, which might be on a household’s lot. Given the different stages of development in the participating countries, it turned out to be reasonable in some countries to limit the investigation to only one or two of the services of interest (electricity, BDS, MFI). One reason is that it was not possible to find an appropriate number of MSMEs that benefit from electricity, MFI, BDS, and combinations thereof. While it would have been very interesting to uniformly define industries and firms sizes to be included in all participating countries, this strategy turned out to be impracticable because of the heterogeneity of the target regions. While the Ghana region, for example, features many small and micro manufacturing firms, the country team in Uganda found few firms employing more than two workers. Additionally, almost all firms in Uganda are in the retail and service sector, while there is a significant number of firms operating in other sectors in the target area in Ghana. As an alternative to stratified sampling in the absence of comprehensive firm enumerations, most country case studies applied simple random sampling. This is legitimate, given that a systematically bias of the sample is avoided. Examples for such biases would be, for example, only including firms next to the village’s main road or by deliberately including more firms for which a strong impact of energy provision is expected. 28 INGENS STATUS REPORT JUNE 2008 References Abu-Qarn and Abu-Bader (2007) Sources of Growth Revisited: Evidence from Selected MENA Countries, World Development, Vol.35, 752-771 Bensch and Peters (2008) Private Sector Participation in Micro-Hydro Power Supply for Rural Development: Baseline Study and Impact Assessment. RWI Essen and GTZ. Blalock, G. and F. M. Veloso (2007) Imports, Productivity Growth and Supply Chain Learning, World Development, 35, pp.1134-1151. Daniels, L. (2001) Testing Alternative Measures of Micro-Enterprise Profits and Net Worth, Journal of International Development, 13, p.599-614. De Mel, D. McKenzie and c. Woodruff (2007) Measuring Microenterprise Profits: Don’t ask how the sausage is made; World Bank Policy Research Working Paper 4229, May 2007. Easterly, W. (2001) The Elusive Quest for Growth, MIT Press ESMAP (2003) Monitoring and Evaluation in Rural Electrification Projects: A Demand Oriented Approach; Report of the Joint UNDP/World Bank Energy Sector Management Assistance Programme (ESMAP). Frondel, M. and C. M. Schmidt (2005), Evaluating Environmental Programs: The Perspective of Modern Evaluation Research. Ecological Economics, 55 (4), 515-526 MacKernan, S. M. The Impact of Microcredit Programs on Self-Employment Profits: Do Noncredit Program Aspects Matter? Review of Economics and Statistics 84 (1), p.93-115 Moser, G. and T. Ichida (2001) Economic Growth and Poverty Reduction in Sub-Saharan Africa, IMF Working Paper 01/112. Methodology 29 Motta, M. and K. Reiche (2001) Rural Electrification, Micro-finance and Micro and Small Business (MSB) Development: Lessons for the Nicaragua Off-grid Rural Electrification Project. Peters, J.; M. Harsdorff and F. Ziegler (2007) Complementary Services must accompany Rural Electrification – Anecdotal Evidence from Benin, in: Appropriate Technology, Vol. 34 (3). OECD (2007) Promoting Pro-Poor Growth: Ex Ante Poverty Impact Assessment. Ravallion, M. and Chen (2008) Are there lasting impacts of aid to poor areas ? Evidence from rural China? Policy Research Working Paper 4084, The World Bank. Ravallion, M. and Q. Wodon (1998) Evaluating a Targeted Social Program When Placement is Decentralised, World Bank Policy Research Paper 1945. Ravallion, M. (2008) Evaluating Anti-Poverty Programs, in: Handbook of Development Economics, Vol. 4, forthcoming. Straub, S. (2008) Infrastructure and Growth in Developing Countries: Recent Advances and Research Challenges. The World Bank Research Department, Policy Research Working Paper 4460 Vance, C. and J. Geoghegan (2004), Modeling Semi-Subsistence and Commercial Land-Use Decisions in an Agricultural Frontier of Southern Mexico: A Switching Regression Approach, International Regional Science Review, July, 326-347. Vance, C. and R. Hedel (2008), On the Link between Urban Form and Automobile Use: Evidence from German Survey Data, Land Economics, 84: 51-65. Wacziarg (2002) Review of Easterly’s “The Elusive Quest for Growth�, Journal of Economic Literature, Vol.40, 907-918 Walsh, C.E. (2004). The Productivity and Jobs Connection: The Long and the Short Run of It, 30 INGENS STATUS REPORT JUNE 2008 FRBSF Economic Letter, Number 2004-18, July. World Bank (2002) Rural Electrification and Development in the Philippines: Measuring the Social and Economic Benefits. World Bank (2007) World Development Report 2008 – Agriculture for Development, World Bank. 3 3. Status and First Findings from Country Studies 32 INGENS STATUS REPORT JUNE 2008 A. BENIN: The Impact of Electrification on Micro-Enterprises in Rural Benin By Marek Harsdorff, Jörg Peters and Colin Vance - RWI Essen 1. Introduction The agricultural sector in Benin employs 90% of the rural working population, leaving a rather marginal role to the service- and industry sector at first glance. Apart from subsistence crops, cotton – one among few products designated for export – determines the short term performance of Benin’s rural economy. Nonetheless, the amount of small service and manufacturing enterprises in rural villages is increasing – notably within the informal sector – as a growing number of farmers are trying to diversify their income. It is frequently argued that non-farming activities can provide a major source of household income security, and ultimately also support local economic growth. Access to electricity is often named as a bottleneck for an accelerated development of these firms in rural Benin. In practice, though, take-up rates and visible impacts of electrification projects in the commercial sector are disappointing in many cases. Nevertheless, there is evidence that mainly lighting and small machinery are starting to reform the service and small scale industry sector. In light of these considerations, the INGENS Benin case study investigates the impacts of electricity provision on rural microenterprises in Northern Benin. The focus is on manufacturing firms and parts of the service sector. By applying a cross-sectional control group approach, we intend to derive causal results about the impact of electricity access on firm performance. In addition, the relevance of both micro-finance and BDS use is analyzed with regards to their interaction with electricity access. For this purpose, around 350 firms in two regions – one with and one without access to the grid – were surveyed between April and May 2008. The remainder of this summary is structured as follows: First, the country specific methodology is presented. Second, preliminary findings are discussed and some impact hypotheses to be examined during future data analysis are formulated. Finally, the report concludes by providing an outlook on research methods to be applied for the final report. Emerging Findings from Country Cases 33 2. Methodology Identification Strategy The Benin case study applies a cross-sectional analysis that intends to identify causal relationships between firms having access to the grid and those that do not. To this end, five selected villages in the target region of the GTZ electrification project were surveyed which form the control group. In addition, five villages were visited that were connected to the electricity grid between 5 and 8 years ago. These electrified villages are referred to as the treatment group. In order to avoid heterogeneity in variables across these two regions that might influence firm performance, seven key characteristics for selecting the villages to be included were identified: Key characteristics: - Geographic location: located in a rural area with a distance to Cotonou between 400 and 600 km - Population: between 500 and 1500 households - Educational institutions: existence of secondary school - Economic relevance: existence of a regular market in the village - Political relevance: existence of communal administrative office - BDS: Enterprises have access in principle, even if partly and irregularly only - Financial institutions: access to micro-finance Hence, the only visible difference between these two groups is access to grid electricity, thereby establishing a sufficient basis for comparability. In doing so, small villages exhibiting limited business opportunities had to be excluded beforehand, since comparable electrified villages do not exist. Therefore, results from the study have to be interpreted with respect to this selection process and can not simply be extrapolated to rural areas in general. As BDS have been provided very rarely, there is no real control for BDS access. Controlling for access to finance is hardly possible as microfinance structures already exist in all of the surveyed villages and in most rural areas in Benin. Therefore, the Benin study exclusively aims to identify causal relationships between access to electricity and firm performance. Nevertheless, use of BDS as well as microfinance can be investigated. 34 INGENS STATUS REPORT JUNE 2008 Sampling The Benin case study focuses on manufacturing micro-enterprises in order to reduce heterogeneity within the sample. Only firms that are located in permanent structures – be these on the household’s plot or elsewhere – were interviewed. For the core manufacturing sample, carpenters, saw mills, welders, mechanics and tailors were determined beforehand to be included. Applying simple random sampling, every second manufacturing enterprise in both electrified and non electrified villages was interviewed. In total, 277 manufacturing firms were visited, of which 130 are located in the electrified region and 147 in the non-electrified region. In addition, micro-enterprises from other sectors were surveyed using simple random sampling. The sample size for this group, though, is considerably smaller: In the non-electrified villages, 51 out of 150 micro-enterprises belonging to the service sector were interviewed, 40 out of 170 in the electrified region. With regards to the data analysis, we plan to apply the more sophisticated methods that require a higher number of observations on the manufacturing sector, while descriptive analysis can equivalently be conducted for service enterprises. To get a somewhat comprehensive impression of all the non-agriculture income generating processes in the survey regions, a small number of agricultural processing firms was visited. Six firms in non-electrified and 5 in electrified villages were interviewed, which amounts to around 10% of total agricultural processing firms. Altogether, 379 enterprises were interviewed. Furthermore, 15 key interviews with local resource persons were conducted to collect information about the overall socio-economic situation, the availability of energy, the main problems in the villages, and to assess potential long term trends in enterprise development. Practical Survey Approach The Benin survey was conducted by well trained enumerators under close supervision. First, it was ensured that each enumerator understands the intention of the study as well as the intention of each question. Accordingly, enumerators provided explanations during the interview if they realized that a question was not understood or not answered correctly. Furthermore, the enumerators were trained to sit down after each interview to review the whole questionnaire. If Emerging Findings from Country Cases 35 responses were missing or had been apparently misunderstood, the enterprise was revisited. The field supervisors reviewed all questionnaires at the end of each day. If information was missing or incoherent, the respective enumerator was asked to go back to the firm. The fact that the enumerators were paid according to each fully completed and coherent questionnaire provided them with additional incentives to carry out the research carefully and thoroughly. 3. Preliminary findings The field work for the Benin case study was finalised in late May. Therefore, the collected data has not yet been analysed profoundly. Preliminary findings, though, that are based on qualitative impressions gained during field work and that could be confirmed by simple figures from the dataset are presented in this section. It is envisaged to further elaborate on these preliminary results in the main report. At the first glance, one intriguing result is the low electricity take up rate among manufacturing firms: After 5 to 8 years of electrification, only 29% of those manufacturing enterprises that already existed before electrification are connected to the grid. In contrast, 98% of the enterprises in the service sector use grid electricity. The main usage of electricity in both sectors is lighting, which might explain the difference between the two sectors: Manufacturing firms pursue their businesses mainly during daytime. Service firms like bars, shops and phone cabins mostly operate in the evening hours. Incentives for manufacturing firms to extend working hours seem to be lacking as increased production cannot be sold. Income generation through electric lighting and its productive use is higher in the service than in the manufacturing sector. While only few enterprises in the electrified villages that already existed before electrification use electric machinery, businesses that were created afterwards do in high proportion. One can even conclude that these firms were created thanks to the new availability of grid electricity. The reason is that they depend on electricity, since businesses like sawmills, photocopy or fish shops cannot work without electric appliances. As a consequence, they do not exist in non electrified villages. The number of service businesses in the electrified villages grew by 65% after electrification, the manufacturing enterprises by 23% (See table). 36 INGENS STATUS REPORT JUNE 2008 Firm Creation in Electrified Villages Total that existed Electricity before using Created after electrification and Total thereof electrification use electricity now Manufacturing sector enterprises Carpenter 76 11 0 11 Mechanic 57 6 2 4 Tyre vulcaniser 30 13 3 10 Forger 17 2 0 2 Tailor (men/women) 82 44 7 37 Welder 25 24 18 6 Electrician for electric equipment 15 15 11 4 Electrician for motorbikes 17 17 15 2 Saw mill 4 4 4 0 TOTAL 323 136 60 76 Service sector enterprises Bar 27 27 3 24 Shop 33 33 4 29 Hair dresser (men/women) 54 52 14 38 Photocopy 10 10 10 0 Fish shop 8 8 8 0 Telephone cabin 22 22 17 5 Photograph 2 2 2 0 Shop for electric equipment 6 6 6 0 TOTAL 162 160 64 96 only Agriculture sector enterprises lighting Mill for cereals 75 17 8 9 Mill for fresh vegetables 10 1 1 0 TOTAL 85 18 9 9 Emerging Findings from Country Cases 37 Within the service sector, phone cabins were installed in large number because of cell phone companies that often decide to extend their network to a village as soon as it is electrified. Concerning the manufacturing sector, notably welders tend to establish new businesses in the electrified villages. Apparently, demand for some new products, in this case iron goods that were not previously available, is strong. In general, investment in new machinery seems to emerge if an unsatisfied demand for new products exists, for example for photocopies, iron doors or frozen fish. It comes as a surprise, though, that these firms do not emerge in non-electrified villages by using generators. One explanation might be the considerable higher costs, in particular for large machines requiring powerful electricity. Investment costs to acquire a generator are prohibitive in many cases. In addition, the quality of electricity is not comparable. A photocopy shop would have to either turn on the generator for each copy or to run it all the time. Both options are not only inconvenient but also expensive. Existing manufacturing enterprises abstain from machinery investments, as the currently produced array of products does not necessarily require electricity. Machinery could therefore only be used to increase productivity and, hence, to increase output at given labor input or reduce labor at a given level of output. This, however, is not attractive for many rural firms: Increasing output is not an option, since demand on local markets is limited and regional or even national markets cannot be accessed. On the other hand, reducing labor input does not reduce costs to a sufficient extent, as labor is very cheap in rural areas. Income generation through electric machinery seems to be realised notably by newly created enterprises rather than by existing ones. These new firms address either an existing but up to now unsatisfied demand (e.g. photocopy shops) or contribute to an increased division of labor by producing preliminary products (e.g. sawmills).This again supports the hypothesis that access to markets is the most important prerequisite for any development in rural areas. Even 5 to 8 years after electrification, machinery that is run by diesel or petroleum (mills, fridges) seems not to be replaced by electric ones. Taking the example of agricultural processing, 98 % of mills that have access to electricity are still diesel driven, although 38 INGENS STATUS REPORT JUNE 2008 operating costs are much higher than for electric mills. As access to financial institutions exists, the lack of information and missing BDS might be the reason for the absence of economic investments and income generation after electrification. 4. Research outlook The preliminary findings and hypotheses presented in section 3 will serve as a basis for further analysis. We intend to investigate these issues by differentiating for firm types and sectors. In addition, firm performance will be analysed profoundly using descriptive as well as econometric methods and, if required, matching approaches. For this, the core dataset of manufacturing firms will serve as a good basis. Given satisfactory data quality, we will estimate production functions to get an impression of the impact of electrification controlling for other factors. An instrumented variables approach using grid coverage as identifying variable for electricity connection will be followed. If firms from different groups, in particular access and non-access firms, turn out to be too heterogeneous, matching methods will be applied to create more appropriate control possibilities. With regards to BDS and MFI, it will be extremely difficult to derive causal results. Nevertheless, insights about the patterns of usage can still be illuminating. Additionally, matching methods again can be used to approximate causal relationships by comparing the service using firms to comparable counterparts only from the non-using group. Furthermore, it might be possible to develop an indicator for quality of microfinance services by, for example, taking into account the different levels of bureaucratic red tape that an enterprise faces. Such an indicator might change over regions in sample, so that variation would be available to examine the impact of service quality. In general, recommendations for the implementation of the GTZ electrification project will be elaborated. Emerging Findings from Country Cases 39 B. GHANA: Summary of Key Findings from MSE Survey in the Brong Ahafo Region By William Steel and Marco Hüls I. Introduction Ghana has a favorable environment for access to electricity and development of micro and small enterprises (MSEs) relative to many African countries. With political and macroeconomic stability prevailing for two decades, growth has been maintained at over 5 percent per annum since 2001. About half the population has access to grid electricity, including all 166 district capitals, although weak supply growth combined with cutbacks in hydropower due to poor hydrological conditions in the Volta Lake led to severe and prolonged nationwide load-shedding in 2006 and 2007. Connection fees and tariffs have been subsidized. Looking ahead, the future growth of electricity usage may be constrained by: (i) insufficient supply system capacity to meet existing demand; (ii) the high costs of providing service to sparsely populated areas not yet served by the grid; (iii) steadily rising tariffs to close the gap between revenues and costs of investments in generation and distribution; and (iv) loss of confidence in the reliability of supply. Ghana’s overall business climate has been improving under the current private sector-oriented administration. MSEs (including self-employed enterprises) have faced few restraints, apart from periodic attempts at “decongestion� of major urban centers by moving certain activities out of the central business district to outlying areas where electricity and other services are not always available. Access to financial services has improved significantly over the last five years due to a liberalized and increasingly competitive and diversified financial sector. Locally-owned Rural and Community Banks (RCBs), with over 600 total outlets, cover the whole country and many of them offer group-based microfinance programs suitable for the “entrepreneurial poor.� Business development services (BDS), however, remain driven primarily by subsidized government and donor programs, with access outside the major urban centers limited to training arranged by the government run Business Advisory Centers. The GTZ-supported Program for Sustainable Economic Development is assisting districts in the study region to establish or upgrade industrial zones to help develop MSEs through increased productivity, growth and employment. The existing and newly established industrial zones cater 40 INGENS STATUS REPORT JUNE 2008 to a limited range of sectors including woodwork, food processing and car-repair related activities. The project will provide these zones with (improved) electricity networks, roads, water supply and sanitary facilities. The companies located in the industrial clusters will receive BDS tailored to their needs and will be linked with micro finance institutions. The underlying hypothesis of the project is that a service package comprising improved electricity supply, systematically provided BDS, access to microfinance and operating in a designated clustered area leads to better enterprise performance. It is assumed that providing this package would improve the impact on business performance compared to a provision of each service individually. The current survey serves as a baseline for a temporal comparison after the project’s end. For the cross-sectional purposes of the overall study, the Ghana case study will contribute a comparison of service-using and non-using firms. The reason is that cross-sectional control groups for access to any service were not available. That said, there are some preliminary conclusions on the use of the electricity and complementary services in MSEs that can be derived from the existing baseline data of the Ghana study. As will be presented below, the common-sense expectation that the use of electricity, BDS and micro finance is positively correlated to firm performance does not hold true in all cases. Affordability and reliability issues in the case of electricity and market distortion caused by supply driven BDS provision might be one explanation that needs to be considered. Emerging Findings from Country Cases 41 II. Methodology and Survey Area A simple random sample of 271 MSEs with fewer than 20 workers was interviewed in 2007 for the INGENS study in three district capitals, one small, medium and large respectively, in Ghana’s Brong-Ahafo Region. The district capitals all report strong growth and improvement in infrastructure, with even more rapid growth of population through the influx of migrants from surrounding rural areas, resulting in a rising density of MSEs and informal employment. Since the manufacturing sector seemed to offer more interesting insights due to a higher variation among firms, it was sampled at double the rate than commerce, which is the dominant subsector. In the analysis, these were weighted to arrive at a projected, representative sample. Moreover, for the purposes of providing a baseline for the GTZ project and further analyzing specific sectors and clusters that are of special interest for the project, existing industrial zones were oversampled. Additional observations in an industrial zone outside the project area were interviewed to serve as a control group for ex-post impact evaluation of the project. The main methodological drawback in designing the Ghana survey was the unavailability of suitable cross-sectional control groups for the three treatments (electricity, BDS and microfinance) investigated by INGENS. As an alternative, the analysis focuses on the “use� of electricity and other services. This means that all services (electricity, finance, BDS) are available in all regions, whether or not individual firms actually choose or can afford to make use of it. It was not feasible to find a suitable non-electrified town to serve as a control group because the remaining off-grid towns in Ghana are not comparable in terms of size, remoteness, and economic environment. Hence, the impact of access to grid electricity on enterprise performance cannot be assessed. Similarly, as the RCBs and other microfinance institutions are distributed widely, it would be difficult to find a comparable town with no access to financial services. Therefore, the Ghana data do not allow for analyzing the “access� interpretation of the main treatments. As outlined in the overall methodology, investigating the use of a service inside of a given socio- economic environment instead of investigating access to the treatments in separate but comparable areas poses a problem in deriving meaningful conclusions from the data. It is unlikely 42 INGENS STATUS REPORT JUNE 2008 that being connected or unconnected to the electricity grid inside a town where access is generally available is randomly distributed among enterprises. The decision of connecting to the available grid or using business or financial services may well be influenced by external factors not captured in the survey. Moreover, the “use� of a service is very likely to be simultaneously determined with firm performance, which would render any causal conclusions impossible. In the absence of a non-access control region, applying the identification strategy presented in the overall methodology is not feasible. Therefore, all analysis presented in this report is based on correlations between variables among different groupings within the sample at a point in time, and do not necessarily demonstrate causality. III. Preliminary Results The uptake rate of electricity among the MSEs sampled is reasonably high at 73%, with three- quarters of this group using it for machinery and equipment.28 Grid connection is associated with a 50% increase in the number of hours per day that MSEs use lighting devices (especially fluorescent tubes and energy-saver light bulbs). Nevertheless, grid connection and its associated additional lighting yields only a marginal increase in operating hours per day (11.7 vs. 10.5 hours for firms not using electricity). In spite of these slightly extended working hours, electricity use per se is not necessarily associated with higher performance. Those firms that do not need electricity directly for their particular activity (e.g., food preparation for sale, tire repair, battery-operated telephone services) do not necessarily perform worse than those connected to the grid. Contrary to the presumption that the use of electric machinery generates higher productivity, there is also no consistent evidence that enterprises using electricity for machinery and equipment perform better than those with electricity for lighting only. This result is even more astonishing taking into account the above mentioned simultaneity bias. 28 Subsectors with relatively high use of electricity for lighting only are food preparation/processing, commerce, and woodworking. Emerging Findings from Country Cases 43 This bias normally leads to an exaggeration of the impact: by correlating electricity use and profit, one neglects that the more profitable firms are more inclined to connect and thereby assigns too much of the profit difference between connected and non-connected firms to the treatment of electricity provision. The results thus seem to support rejecting the hypothesis that electricity-using firms perform better. There are, however, two potential explanations for the lack of a difference between connected and non-connected firms in the sample. First, the firms who are connected to the grid belong to other subsectors than those who are not connected. Comparing those two groups might be not appropriate if the subsectors are too heterogeneous. Indeed, different subsectors vary widely in terms of electricity use. For example, food preparation and processing firms operate an average of 12.1 hours a day as against 10.4 hours for wood and construction firms, even though 60% of firms in both subsectors are connected to the grid. This indicates that the impact of electricity on hours of operation and on performance in general is not independent of subsector effects. Therefore, further analysis has to be done within the subsectors. The second potential explanation for the absence of a difference between connected and non- connected firms might be the unreliability of the grid. Given the load shedding policy in Ghana – firms have recently faced announced 12 hours outages every second day – unreliability of electricity could have indeed offset the potential productivity-enhancement effects of electrification. Many machinery using firms (38%) cease operation until electricity resumed. Furthermore, a quarter of all respondents reported damage to machinery and equipment from outages and fluctuations, costing an average of $160 per year. Generators or batteries as back-up capacities are used by only 15% of those using machinery and 8% of those who use lighting only. For the surveyed firms, use of both finance and BDS together is positively associated with good enterprise performance, compared to firms using neither. The availability of various financial institutions in the towns surveyed yielded high take-up rates, with 82% of enterprise owners having a savings account, of which half had applied for a loan. This indicates that in Ghana the variety of types of financial institutions and the spread of microfinance methodologies has made access to finance more common than in countries where financial and cultural barriers limit their use of financial services. As much as 35% of respondents had received loans, although they do 44 INGENS STATUS REPORT JUNE 2008 not appear to use them as a major source of financing for start-up and working capital for their business. Nevertheless, “access to credit� is the business constraint (64%) most frequently cited by survey respondents. This, however, can also be due to the fact that respondents were informed about the objective of the interview, making a reporting bias very likely. BDS by itself (i.e., usually management training in this context) does not appear to be an important factor for firm performance. The reason could be that there is little access to market- oriented BDS in these towns. Another explanation is that the Government-sponsored training programs appear to be targeted toward older, relatively weak firms, which tend to perceive such training as an entry to obtain credit from government schemes rather than as a means of improving productivity. IV. Research Outlook With regards to the intriguing result of lacking differences between connected and non-connected firms, within sector analysis has to be intensified. Mainly subsectors like car repair or welding offer potentials for this. Sample sizes of these groups would also allow for sample specific regression analysis. To mitigate the influence of self-selection and simultaneity in comparing connected and unconnected firms, an internal control group can be constructed from the sample that only contains those firms that can be considered as “un-voluntary� non-users. These firms applied for a connection, and hence are capable to pay, but have not received a connection yet. There are, however, two concerns about this approach. First, only 12 firms belong to this group. Second, to use this group as a control, we have to assume that there is no unobserved variable that affects both involuntarily not receiving a connection and firm’s output. Nevertheless, it is worth investigating the performance of this subgroup compared to the connected firms. In particular, given the small size of this group, matching approaches can be used to create an appropriate comparison group to the involuntarily not-connected firms from the connected group. Since the Ghana data exhibits many missing values in important variables, imputation techniques Emerging Findings from Country Cases 45 to fill gaps will be applied. This will avoid the loss of many observations just because a few variables of interest are missing and offer important information for analysis. The variety of types of financial institutions and the spread of microfinance methodologies observed in the surveyed regions offers interesting research possibilities. This might allow for investigating the impact of different levels and qualities of MFI provision. After finalization of the GTZ project, treatment and control groups will be available for ex-post analysis. This will make it possible to use the Ghana sample for a more systematic analysis of the impact of access to a service package, including reliable electricity and other services and infrastructure in small-scale industry clusters over time. However, difficulties remain in controlling both the application of specific “treatments� and the role of other uncontrolled factors in the economic environment. Although we will be able to assess the impact of the whole service package as provided by the project, disentangling the individual impacts of each of the treatments will pose difficulties. 46 INGENS STATUS REPORT JUNE 2008 C. NIGERIA: Summary of the Key Findings from the MSE Survey in Lagos State By William Steel and Abiodun Akinkunmi I. Introduction Since its 1999 transition to civilian rule, the Nigerian Government has put in place a series of economic reforms and policies that significantly opened up the economy and stimulated growth. Although the country’s power policies have mostly focused on electrification of metropolitan areas, the power sector in these areas is performing poorly. With demand for electricity more than double the available grid generation capacity of the Power Holding Company of Nigeria (PHCN), unscheduled outages are a daily occurrence. This imbalance of power supply and demand was already prominent in the 1990s, as demand growth outpaced the addition of new generating capacity, currently at 3200 MW. It was aggravated by the creation of new states and local governments in the 1990s, which further strained the electricity supply system as they were connected to the grid. The unreliability of supply means that most industrial and commercial businesses must depend on expensive diesel generators as back-up (or virtually primary) systems. PHCN customers have less than one hour grid usage per day on average, and thus rely mainly on diesel generators for the rest of their needs. Furthermore, the incentive to obtain a new connection is reduced by the unreliability, despite the availability of the grid and other electrical infrastructure such as step-up and step-down transformers. Many Nigerians do not bother to connect to the grid even when available, and the micro and small medium enterprises (MSMEs) that depend on electricity are severely disadvantaged if they cannot afford to purchase or operate expensive diesel generators. Meanwhile, over 60 percent of the rural areas are not connected to the grid, and those connected do not have access to regular supply of electricity. The lack of electricity supply in rural areas, where raw materials for agro-allied industries are produced and a large share of the population lives and earns a living, has limited the potential for socio-economic development of these areas. Emerging Findings from Country Cases 47 II. Methodology and Survey Area The survey work was undertaken in Lagos State, a coastal region that is the commercial and industrial center of Nigeria. About 80% percent of the country’s manufacturing companies and headquarters of service and commercial companies are located in Lagos. This economic concentration has attracted a heavy migration from the rest of the country, resulting in a huge informal sector, including a large number of micro and small enterprises (MSEs). Agege and Surulere, two areas within Lagos State, were targeted by the World Bank’s program to improve reliability of electricity supply and were consequently surveyed for INGENS. These two areas represent the diverse characteristics of Lagos State: Agege includes both industrial and residential areas; while Surulere is mainly a residential and more elite area. Agege is made up of two local government areas (LGAs), Agege and Ifako-Ijaye, including five towns (Agege, Ifako, Ijaye, Ikeja and Ogba); while Surulere is a local government area on its own with only one town. Since the electricity grid is available throughout Lagos State, it was not possible to find a comparable non-electrified area as a control group. An initial enumeration identified 906 MSEs in these areas that met the criteria of being in a fixed place of business and having fewer than 20 workers (excluding petty trade). Using simple random sampling, 170 of these MSEs were selected for interviewing. 48 INGENS STATUS REPORT JUNE 2008 III. Preliminary Results Irregular supply of electricity by the power authority makes accessibility a problem in practice: i.e., being connected to the grid does not mean that enterprises have regular electricity supply. Hence, some MSEs may choose not to apply for connection to the grid because it is so unreliable. Although the survey does not explicitly capture unreliability as a direct reason for non- connection, of the 31% of sampled firms that are not connected, only 3% say they cannot afford it; another 12.5% want a connection but have not yet received it; and 15% say they “don’t need� the grid (either because it is so unreliable, they have alternate sources of energy, or simply do not utilize it in their activity). Self-employed enterprises are more than twice as likely as firms with 5-19 workers not to have a grid connection, and are also somewhat more likely to say they “don’t need� grid electricity. This may be because the high cost of backup energy sources (especially diesel generators) leads the self-employed to naturally select to work in subsectors that are less dependent on energy. Of the total sample, 43% of MSEs use electricity for lighting only, while 25% use it additionally for machinery and equipment. Surprisingly, in designated “industrial zones�, only 11% of MSEs use electricity for machinery and equipment, compared with a substantially higher proportion approaching 34% in market areas and town centers. The restaurant, bar and lodging subsector is most likely to use electrical machinery and equipment (57%). The subsectors least likely (below 17%) to use electricity for machinery and equipment are: hairdressing, miscellaneous manufacturing, food preparation and processing, and ICT (including telephone services using battery-operated telephones – which nevertheless must be recharged elsewhere). The poor quality of supply has undoubtedly impeded greater use of electricity for productive activities by MSEs in Nigeria. “Reliability of energy� is the most often cited major business constraint to greater use of electricity by survey respondents (62% of the sample); access to and cost of energy are also cited as major constraints by more than half of the firms. This has led to considerable investment in back-up energy equipment, with as much as 61% of MSEs in the survey having a generator and many others using industrial voltage stabilizers. Some 30% of the MSEs surveyed have had equipment damaged by electricity outages or power oscillations, Emerging Findings from Country Cases 49 reporting an average amount of N2,290 spent on damaged equipment on a monthly basis.29 As much as 55% of the MSEs respond to interruptions of grid electricity primarily by switching to a back-up energy source (with this percentage slightly higher for those using electrical equipment), while 13% stop business operations until grid power is restored. Other studies have found that unreliability of electricity in Nigeria has led to the closing of many MSEs and even large enterprises, and has probably reduced the relative share of production by smaller enterprises, which are less able to afford back-up energy systems than larger ones (or whose profits are more adversely affected by back-up energy costs). Perhaps because of the overriding factor of unreliability, no clear relationship was observed between the use of grid electricity and profitability or growth of sales and employment over the previous year. With respect to complementary services, the Federal Government has emphasized finance as a keystone of its policies to support MSMEs. In 1999, the Government established the Small and Medium Industries and Equity Investment Scheme (SMIEIS), which is funded by all licensed banks by setting aside 10% of their profits before taxes. This is intended to improve the share of liquidity available for MSME loan portfolios, and (together with some State investment funds), has helped make financing available for Microfinance Banks (MFBs). The data provide inconclusive evidence that the use of financial or business development services (BDS) is associated with positive enterprise performance. Overall, 56% of the respondents reported that their sales exceeded expenditures from the previous month. Roughly 47% indicated that sales had grown over the previous year, while 20% reported growth in employment. Although these shares were slightly lower for firms that did not use either finance or BDS, statistically significant differences could not be discerned. While nearly half (49%) of the MSEs surveyed have a bank account, only 13% percent had applied for loans (of which 87% were successful). Unusually for MSE surveys, ‘access to credit’ was not among the top ten major constraints to business. Demand for credit may be restrained by high interest rates, complex procedures, and collateral requirements. ‘Cost of credit‘ was cited as 29 The exchange rate of Nigerian Naira per U.S. dollar is roughly 116. 50 INGENS STATUS REPORT JUNE 2008 a major problem by 17% of all firms, and as the main reason for not applying for a loan by 29% of the respondents who had not applied; the complexity of procedures was the main reason for 20% of the non-applicants, and collateral requirements was the reason cited by 8%. The impact of using electricity and complementary services is affected by other factors, particularly infrastructure. Problems with transportation, communication and availability of potable water adversely affect MSEs throughout the country. Most main and feeder roads are in poor condition, raising the cost of bringing in raw materials and conveying finished goods to the market. Information technology systems are not affordable by many of the MSEs. Problems with security due to high crime rates in Nigeria cannot be overemphasized as a high risk and cost factor for MSEs. “Crime� includes extortion of money by the police and local government officials from shop owners and demand for “royalty� by “area boys� and street urchins.30 Shop owners reported cases of fraud and scams where they were duped by promises of services to their businesses, as well as cases of National Food, Administration and Control officials and other regulatory agencies asking them for bribes. Nevertheless, despite the general increase in cases of armed robbery, crime and theft were viewed as “no problem� by a majority of the survey respondents. However, as in the case of unreliable electricity, as increasing numbers of MSEs are directly affected by crime, the need to adopt alternative security measures is likely to impose additional costs and reduce profits. IV. Research Outlook Several possible avenues for future analysis emerge from the results presented here. With respect to data collection, the possibility of identifying potential control regions either exhibiting more reliable grid access or no access at all merits further exploration. If a clear distinction along the lines of access or reliability proves is not feasible, then two alternative analytical foci could be 30 A measure has been put in place by federal government to combat corruption among government officials. A palliative measure has been put in place by the Lagos state government to withdraw area boys from the street for rehabilitation. Emerging Findings from Country Cases 51 pursued. The first would compare the use of generators with non-use, with the aim of identifying the firm attributes that characterize each group. Building on this, the second would focus on quantifying the costs of power outages by examining the responses of manufactures, potentially by deriving willingness-to-pay estimates from revealed preference data. 52 INGENS STATUS REPORT JUNE 2008 D. UGANDA: The Impact of Electrification on Micro-Enterprises in Rural Uganda By Sven Neelsen, Jörg Peters and Colin Vance - RWI Essen I. Introduction Thirty percent of Ugandans today are exposed to absolute poverty. The economic potential of rural areas, in particular, is debilitated by a lack of basic infrastructure, with a national electrification rate of about 5% and frequent shutdowns of power due to a lack of supply. Uganda and the survey region (encircled) Against this background, the Uganda case study focuses on the impact of grid-access- treatment on rural micro-enterprises. For this purpose, 224 enterprises in three different areas were interviewed in mid-2007: one with access to the grid, a second with no current access but a planned connection, and a third with no access and no current plans for electrification. All areas are located in Mukono District, along the shoreline of Lake Victoria and some 60km to the East of Kampala. Their population density is high for rural African standards, and the main livelihoods are subsistence and commercial farming and fishing as well as small-scale trade in retail and services. A second field survey is planned for June 2008 to rectify problems identified during the first survey with respect to the comparability of the study areas. By drawing a dataset from more comparable sample areas, the aim will be to improve quantitative data analysis possibilities for INGENS. The process of sampling and other methodological specifics of the Uganda study are set out in section II. Despite the limitations of the 2007 data, a number of tentative results give an indication of how grid-access affects the rural micro- Emerging Findings from Country Cases 53 enterprise. Among the more interesting findings is that the sectoral distribution of enterprises does not differ between the access and non-access areas. The survey additionally revealed that lighting and communication are the primary productive uses of electricity. These and other preliminary findings are presented in sections III and IV. With a new dataset, they will be addressed with a set of more rigorous methods. II. Methodology and Survey Area Identification Strategy The primary aim of the Uganda case study is to investigate the impacts of electrification interventions on the performance of rural firms. As set out in the overall methodology, the establishment of causal relations here requires sufficient comparability of treatment and control samples. To this end, an attempt was made to identify regions with and without access to the grid that are otherwise similar in their economic conditions. With the study's limited budget, the identification of additional control areas for Business Development Services (BDS) and microfinance was impossible without reducing the sample size per group to unacceptable levels. Straightforward control possibilities for the effects of the two services are thus not applicable within the given setup: BDS are not available in any of the surveyed areas, while microfinance products are accessible to all sampled enterprises, with a large number of institutions, credit vehicles and minimal collateral requirements. Hence, while no quantitative results for BDS can be derived from the Uganda case, the impact of microfinance is measured in a use vs. non-use scenario, rather than in the methodologically preferable access vs. non-access setting. Simultaneity and endogeneity problems can therefore not be ruled out. Still, comparing use and non-use as well as different levels of usage might offer interesting insights, in particular when interacted with electricity access and usage. The data collection was carried out in the summer of 2007 in three survey regions of Uganda distinguished by access to the grid. Region (1) has no grid-access but a planned grid connection. Region (2) has no grid-access and no planned connection, and region (3) has had access to the grid since 2003. The first region is selected mainly as the over-time control for electrification impacts. For cross-sectional analysis, observations from (1) and 54 INGENS STATUS REPORT JUNE 2008 (2) are lumped together to form the control group. These two regions are compared to (3), which is referred to as the treatment group. Sampling During the first part of field work in late August 2007, 224 non-farming enterprises working in a permanent structure on the household’s plot or elsewhere were surveyed using stratified random sampling. The applied stratification criterion was proximity to the main road. None of the interviewed enterprises employs more than five employees, so that the Uganda case considers micro-enterprises only. In addition to the firms, interviews were also conducted with key informants in public administration units, the national utility, and a number of microfinance and business development institutions. During the fieldwork it became clear that the crucial identification requirement concerning the comparability of treatment and control region was not sufficiently satisfied. Regions (1) and (2) consist of a number of small trading centres along 30km of dirt-road, with limited access to regional or national markets. In contrast, the already electrified/treatment area (3) is one larger settlement and a gazetted31 landing site for commercial fishing, with extensive interregional trade. For cross-sectional analysis, it is hence impossible to disentangle the effects of landing site status and access to the grid. Therefore, the project team decided to return to the field in June 2008 and raise a suitable control group (4) from a gazetted landing site without grid-access. The identified site is situated in the same sub-county as (3) and is comparable in physical extent and population. One factor complicating the identification of a proper cross-sectional treatment group in Uganda was the fact that the target region of the electrification intervention was not selected according to its economic potential or any priority list of regions to be electrified (see overall methodology). Therefore, most electrified rural areas in Uganda exhibit much better economic conditions than the target region and it was not possible to find a comparable treatment region. 31 To be gazetted, a landing site has to fulfil the sanitary and general infrastructural requirements set by the Ugandan government and the European Commission. Emerging Findings from Country Cases 55 III. Preliminary Results The descriptive results presented in the following are derived from the first round of fieldwork in 2007. The non-access samples (1) and (2) are compared to enterprises in the grid-access area (3). In addition, comparisons are carried out between the connected and non-connected within (3) - recognizing the methodological limitations as addressed in the overall methodology. Against this background and with the aforementioned concerns on comparability of the survey areas, the findings can only serve as a first orientation. More profound and rigorous analysis will follow once the data of the second round of fieldwork is available. As indicated in Table 1, standard indicators of firm performance like the number of employees, labour compensation, sales, and owner income are higher in the treatment area (3), though the differences are not statistically significant for the former two variables. With respect to owner income, roughly 76% of sampled enterprises with grid access earned over 200,000 UGX in the previous month, while only 38% of enterprises without access crossed this threshold.32 These differences notwithstanding, the combination of higher demand potentials on local markets and grid access does not correlate with substantial differences in the industry structure compared to the control group of (1) and (2). In all cases, the service sector is dominant (>90% of enterprises), with a strong focus on small retail-shops. Bars and restaurants, and hairdressing salons are other relevant sectors. 32 The exchange rate of Ugandan shillings (UGX) per U.S. dollar is roughly 1,721. 56 INGENS STATUS REPORT JUNE 2008 TABLE 1: Differences in firm performance by grid access Standard Difference in Mean Error. n Means t-Test Number of employees Access to grid 0.910 0.120 100 0.121 0.864 No access 0.789 0.081 123 Weekly pay (UGX) Access to grid 16691 2788 38 698 0.190 No access 15993 2328 35 Weekly sales (UGX) Access to grid 225778 37832 97 97903 2.33* No access 127875 21492 113 Count of Firms chi2 Owner income (UGX) >200000 <=200000 15.84* Access to grid 25 8 No access 69 111 * denotes significance at 5% level In the treatment region (3), around 60% of interviewed firms are connected to the grid. Electricity is mainly employed for lighting, communication, entertainment media and cooling, while its use for manufacturing-machinery is scarce. The energy share in total expenditures does not vary substantially between the connected and non-connected enterprises or between the treatment and control group. This may indicate that firms react to the higher energy efficiency of electric devices with increased demand rather than exploitation of the full-savings potential.33 Moreover, even connected firms do not abstain 33 Electric devices generate one lumen hour much cheaper than traditional lighting sources. This lower unit price can result in an increased demand for the service and thereby offset the decrease in consumption attributed to increased efficiency, a behavioural response referred to as the rebound effect. Emerging Findings from Country Cases 57 completely from using traditional energy sources like kerosene. One of the reasons for the continuous use of traditional energy sources and the limited application of electricity in production is certainly the unreliability of electricity supply: In the treatment region, grid electricity is temporarily shut down 2 out of 3 days and 35% of connected enterprises claim damages to equipment due to fluctuating voltage. Electric energy - with its current use - appears to be rather a convenience than a requirement, as almost all enterprises report they are able to operate without electricity. Furthermore, in spite of improved lighting services, electricity-using enterprises do not show more working hours per day than the non- connected. The same applies for working hours across the treatment and control group. Potentials to expand business activities seem to be limited and not restricted by a lack of access to electricity. Access to (micro-)finance also does not appear to be a crucial bottleneck for investments in electricity-using machinery, as it is widely accessible in the area.34 The micro-data and the information from key interviews indicate that the lack of investments in electric machinery is rather due to a shortage of local demand for manufactured products and limited access to external markets. This product market constraint is exacerbated by the dismal condition of the road infrastructure in the survey region. 34 However, many respondents considered the interest rate of 25-30% too high to acquire a micro-loan. 58 INGENS STATUS REPORT JUNE 2008 IV. Research Outlook The surveying of a new area that is suitable as a control group for non-access to grid electricity (4) allows for conducting more rigorous impact evaluation. As set out in the overall methodology, this will go above a descriptive analysis and can include the estimation of production functions and instrumented variable approaches. If matching methods are feasible, they can be applied to create control groups within each of the surveyed areas. This enables us to approximate causal relationships even in the absence of a proper access-control, for instance for the usage and non-usage of micro-credit. Furthermore, the appropriateness of the applied questionnaire as a primary research tool will be intensely discussed with particular regard to the Ugandan set up for regions that show limited economic activity and variation in firm types. Alternative methods to address the encountered problems will be proposed. To enrich the learning potential here, some additional questions that allow for the crosschecking of responses will be integrated into the second round of interviews. Against the tentative results from the first round of sampling, the analysis of the complete data will focus on four aspects: (a) The impact of grid-access on standard indicators of firm performance; (b) the use of electricity in production; (c) the failure to trigger more investment in electricity-using applications and machinery; and (d) indicators of the complementary effect of having both access to grid electricity and to outlet channels like the landing site on economic activity. Emerging Findings from Country Cases 59 E. SOUTH AFRICA No country status summary available in time for June status report 4 4. Emerging Overall Results and Lessons from Country Studies By Jörg Peters and Colin Vance - RWI Essen Taken together, the initial results from across the countries sampled in this study seem to suggest that electrification has a small impact on the majority of MSEs - or at least that such impact is very hard to measure. In the Ghana case, for example, this is suggested by the negligible difference in the performance of grid-connected firms compared to non-connected ones. In the Benin study, alternatively, it is indicated by the extremely low connection rates among manufacturing firms, which one might suppose would be most interested in using electricity for machinery usage. As these early results seem to confirm the pessimistic appraisal expressed at the outset of the study, based on anecdotal evidence from practitioners (that is, the conviction that electricity programmes often have no impacts on productivity), the next step of this study should involve a more detailed investigation to find explanations. One possible explanation may simply be methodological. The example of Ghana, for instance, might be misleading due to an ultimately inadequate comparison resulting from a selection bias: Initial results show that connected and non-connected firms are in different subsectors. Hence, disparities between these two groups undermine the firm performance comparison. Furthermore, the unreliability of the electricity grid might equalize the impacts of electricity access. As some insights from the Ghana case indicate, excessive outages and machinery damage due to voltage fluctuations induce severe problems for the entrepreneurs – so that investments in Emerging Overall Results 61 electricity connection and machinery might ultimately turn into liabilities, compared to competitors who continue to rely on traditional methods and manpower. In general, increasing energy demand that outpaces supply leads to outages and decreasing reliability in many African countries. This is reflected in several of the country cases. While the importance of investigating impacts of electricity access is still beyond discussion, research efforts have to be intensified to examine the consequences of announced and unannounced outages, how firms react, and which solutions are available. Methodologically, reliability problems might even offer interesting research perspectives, since firms’ back-up behaviour could reveal willingness to pay and elasticity of demand approximations. Usually, such figures are difficult to obtain, since the absence of variability in electricity tariffs within a country complicates the estimation of elasticities. The case studies from the rural areas surveyed in Benin and Uganda confirm the widely accepted hypothesis that access to markets is the major barrier to rural development. Local demand for many traditional products is limited so that existing firms targeting local markets do not expand their production. It is, however, somewhat surprising that in Benin several new firms were seen to be created after electrification. These firms sell products that could not be produced without electricity and for which a hidden demand existed already. In Uganda, in contrast, comparable developments cannot be observed. This raises two interesting questions that should be investigated further in this and in future research projects: First, what are the patterns in preferences and consumption underlying these observations? Second, to what extent does this new demand crowd out traditional products? A crucial insight is the difference in the quality of gathered data depending on how the survey was actually implemented in the field. As the Benin example shows, close supervision of field work substantially increases the consistency of the data. The number of uncovered mistakes and inaccuracies by both the field supervisor and enumerators themselves suggest that such an approach is indispensable to collect credible data. There is some evidence that when enumerators refer to the nature of the development project underlying the conducted survey, it distorts the respondent’s answers to perception questions. 62 INGENS STATUS REPORT JUNE 2008 Respondents seem to overestimate in almost all country cases the relevance of electricity, MFI or BDS. This becomes evident if the respondent’s judgement stated in perception questions is compared to standardized questions on service usage, for example. Many firms frequently state access to finance as a serious problem, while they never applied for or even regularly receive credits. Furthermore, it is important that the regions targeted for surveying are pre-investigated objectively beforehand, particularly when the study encompasses several countries. If the differences between countries are too pronounced, then the application of a uniform methodology is unlikely to be warranted. The region surveyed during the first round in Uganda, for example, was largely populated by firms for whom many of the questions in the survey instrument were simply not relevant (though a pre-INGENS project survey had indicated otherwise). These included specifically questions on assets, as a large share of the firms was one-man operations with little in the way of capital equipment. In such cases of very small firms from the commerce sector, more meaningful data could only be gathered using alternatively designed survey instruments that allow closer inspection, for instance by using diaries. Better pre-survey field inspection could have helped to develop a more appropriate research strategy. The same applies to the Nigerian case, where the reliability of electricity provision is the crucial bottleneck. The study’s set up mainly focuses on access to the electricity grid, which is not a pressing issue in the survey region in Nigeria. Inspecting the region in advance could have revealed this interesting research topic. Examining the complementary services BDS and MFI poses several additional methodological problems. BDS, in particular, are extremely heterogeneous, both within and across countries. Comparing low quality vocational training to sophisticated market access consulting activities does not really make sense. Concentrating on a specific BDS branch seems to be more reasonable. Global availability of MFI makes it hard to identify regions that have no access at all, complicating the quest for identifying causal effects. One solution might be to instead examine the quality level, since this does indeed vary across regions and countries. Appropriate indicators and identification strategies for these purposes have to be developed. Emerging Overall Results 63 Finally, the most important overall result is that the objectives of the research study should be commensurate with the assigned budget. Overloading research objectives leads to weak results in the end. In order to avoid this, the INGENS team decided to limit the pursued questions according to feasibility. For instance, it would have been extremely interesting to investigate the interaction of different services, say, BDS, MFI and grid electricity, by identifying causal relationships as described in the overall methodology. Given the limited sample size potentials, however, it was immediately evident that one would end up with too few observations for each combination of service provision. 5 5. Outlook: Next Steps and Funding Request By Lucius Mayer-Tasch, Mohua Mukherjee and Kilian Reiche Additional ESMAP funding is sought for finalizing the full Final Report. Such funding would allow the following deliverables, in addition to the minimum content described in section 1.13:  Prepare the full final report jointly with GTZ, including the advanced analysis suggested in the previous chapter (additional funding for the GTZ contribution has been secured from BMZ).  Add a new chapter on concrete, specific recommendations for practitioners on how to improve EPs by adding complementary services and/or productive uses components. This practitioners’ chapter would complement the present focus on evidence-based Impact Evaluation with a new chapter (or a twin publication) on the more practical recommendations that can be given to EP task managers regarding (i) the design of add-on components for “complementary services� to improve EP project impacts and (ii) real-life examples of productive uses that might be replicable in new EP. This twin publication would be based on qualitative sources (practitioners experience, literature and structured interviews) rather than on application of the AddIE survey instrument.  Add Annex on concrete and replicable examples of Productive Uses from at least four countries, including business plans, financial analysis and lessons learned on project implementation.  Start direct cooperation with the new ESMAP initiatives “Africa Electrification Experiences Outlook 65 Initiative (AEE)� (TTL Rysankova and Tenenbaum) and “Monitoring and Impact Evaluation of Rural Electrification� (TTL Barnes and Khandker), as well as AICD and GTZ’s BMZ/DGIS-funded Energising Development.  Present INGENS results at the ESMAP AEI Africa seminar in early CY09  Conduct advanced analysis on Ghana, Benin and Uganda (e.g. role of lighting and ICT as part of firms’ production process)  Add Senegal as country case number six (this seems to be a case with more promising results regarding actual impact on the firm level) The following table summarizes two options for this additional tranche of ESMAP funding (details see Annex), and what would be the additional outputs for these two options, as compared to the baseline scenario (no new funding tranche). For the High Case, US$195k would be required from ESMAP, for the Low Case, US$ 140k would suffice. INGENS Modules - New Funding Needed ADDITIONAL OUTPUTS ADDITIONAL ESMAP FUNDING [US$ Thousands] Output No new Low Case High Case None Low Case High Case esmap tranche Limited Feedback Workshops / Dissemination x Output 1-10 can be done by GTZ … Results country + meta country x …with already committed GTZ funding Final Methodology + Conceptual framework (other determinants) x Annotated Biblio x Context Chapter (Intro+Background) x Results: Limited Lessons, Recommendations & Conclusions x Annex: (annotated) instrument, TORs, budget, training process+supervision guidelines x editing & printing ? Limited Annex: "cases + cash flows" x Extended Feedback Workshops + Country Dissemination two countries four countries 5 10 Diessemination Donors (AEI Presentation; Editing; "Glossy Publication" + CD + web) small full 10 20 link to Doug: HH AddIE + LongIE yes yes 5 5 Practitioners' Manual: Advice on EP Design and Complementary Services Recomm. Chapter Stand-alone Manual 20 45 Recommendations: Practical issues to whatch out for (e.g. example calculations - ) limited full Conclusions: Framework for Operations integrating complementary services yes Proper Case & Cash Flows yes Improved Manual on Evaluation: AddIE Toolkit! yes yes 20 20 Advanced Analysis (e.g. access vs level, production function, matching) yes yes incl above incl above new country: Senegal N=350, 2 areas N=500, 4 areas 25 35 more treatments (per country) - e.g., BDS, MFI, other infrastructure yes: senegal yes: senegal incl above incl above MDG AddIEs: food/agro AddIEs (NB: link to agro) (i) draft tool Draft instrument Draft instrument 15 15 (ii) Ditto - field work + prel analysis N=350; 3 areas N=350; 3 areas 25 25 Supervision + cooperation GTZ-WB 15 20 Total Additional ESMAP Funding needed 0 140 195 66 INGENS STATUS REPORT JUNE 2008 Draft Bibliography By Suzanne Maia and Ben Attigah Akesson, G. and V. Nhate. 2006. “Rural Electrification Project Ribáué/Iapala in Nampula Province, Mozambique: Study on the impact of rural electrification in the Ribáuè, Namiginha and Iapala Areas of Ribáuè District�. SIDA, Stockholm. Alderdice, A. and J. H. Rogers. 2000. “Renewable Energy for Microenterprise.� National Renewable Energy Laboratory, Golden, Colorado. http://www.nrel.gov/villagepower/vpconference/vp2000/handbooks/microenterprise_handbook.pdf. Alderdice, A., J. Winiecki and E. Morris. 2007. “Using Microfinance to Expand Access to Energy Services: A Desk Study of Experiences in Latin America and the Caribbean.� The SEEP Network, Washington DC. Ali, I. and E.M. Pernia. 2003 “Infrastructure and Poverty Reduction – What is the Connection?� ADB, ERD Policy Brief Series, Economics and Research Department, No. 13. Annecke, W. 1994. “Energising the Poor in South Africa: The Dilemma of an Emerging Economy.� Paper presented at the Conference towards Promoting Sustainable Development, Global Forum ’94, University of Manchester, England, 29 June – 1 July 1994. Annecke, W. 1999. “The Role of Electrification in the Development of Small and Micro Enterprises.� Paper presented at the Association for Women in Development, Washington, DC, 10- 14 November 1999. Annecke, W. 2000. “Assessment of the Impact of the Southern African Power Pool (SAPP) on the Poor of the Region.� Energy and Development Research Centre, University of Cape Town, Republic of South Africa. Annecke, W. 2000. “SMME Business Opportunities for Women in Energy.� Paper presented at the Technology for Women in Business Conference. Pretoria, Republic of South Africa, 24 July 2000. Asaduzzaman, M., D.F. Barnes, and S.R. Khandker. 2007. “Restoring Balance: Bangladesh’s Rural 68 INGENS STATUS REPORT JUNE 2008 Energy Realities�. Energy Sector Management Assistance Programme (ESMAP)/The International Bank for Reconstruction and Development/ The World Bank, Washington, DC. Asian Development Bank / World Bank. 2004. “Sri Lanka: Improving the Rural and Urban Investment Climate.� The International Bank for Reconstruction and Development (IBRD)/The World Bank, Washington, DC. Barnes, D. F. (xxx) “National Survey of Energy Use by Small Companies and Micro Companies.� The International Bank for Reconstruction and Development/ The World Bank, Washington, DC. Barnes, D. F., R. van der Plas, and W. Floor. 1997. “Tackling the Rural Energy Problem in Developing Countries.� The World Bank, Finance & Development, June 1997. Berthomieu, N. 2004. “Les Applications Photovoltaiques Génératrices de Revenus.“ Rapport pour la Promotion de l’Electrification Rurale et de l’Approvisionnement Durable en Combustibles Domestiques, PERACOD, Republique de Senegal et GTZ, Eschborn, Germany. http:///www.peracod.sn/IMG/pdf/les_applications_PV_generatrices_de_revenus.pdf Cabraal, A., D. F. Barnes, and S. Agarwal. 2005. “Productive Uses of Energy for Rural Development,� Annual Review of Environmental Resources, Vol 30, pp.117-44. Cabraal, A., M. Cosgrove-Davies and L. Schaeffer. 1996. “Best Practices for Photovoltaic Households Electrification Programs: Lessons from Experiences in Selected Countries.� Technical Paper No. 324, The World Bank. Washington, DC. Cecelski, E. 2000. “Energy and Poverty Reduction: The Role of Women as a Target Group.� Paper presented at the Debate on Sustainable Energy in Danish Development Assistance, Copenhagen, 5 September 2000, Energia, Washington, DC. Chong, A, J. Hentschel, and J. Saavedra. 2003. "Bundling Services and Household Welfare in Developing Countries: The Case of Peru" (November 2003). World Bank Policy Research Working Paper No. 3310. Available at SSRN: http://ssrn.com/abstract=610355 Cook, C. C., T. Duncan, S. Jitsuchon, A. Sharma, and W. Guobao. 2005. “Assessing the Impact of Transport and Energy Infrastructure on Poverty Reduction.� Asian Development Bank (ADB), Manila, The Philippines. page 69 Cyphelly, I. 2005. “Appui à la Définition des Stratégies de la Mise en œuvre d’une Politique d’Énergie Rurale Productive.� Rapport pour la Promotion de l’Electrification Rurale et de l’Approvisionnement Durable en Combustibles Domestiques-PERACOD, Republique de Senegal et GTZ, Eschborn, Germany. http://www.peracod.sn/IMG/pdf/appui_a_la_definition_des_strategies_de_mise_en_oeuvre_d_une_ polit._d_energie_rur._productive.pdf. Dardenne, L. and M. Connault. 2007 “Etude sur l’Électrification des Activités Non-Domestiques dans les 12 Villages du Projet d’Électrification Rurale.“ GTZ/DGIS/SBEE. Data Vision International (T) Ltd. 2005. “Study to Promote Productive Uses in Rural Electrification Projects in Tanzania.� Final Report, ESMAP/The World Bank, Washington DC. DataVision International (T) Ltd. 2004. “Study to Promote Productive Uses in Rural Electrification Projects.� Inception Report, ESMAP/The World Bank, Washington DC. De Gouvello, C. and L. Durix. 2008. “Maximizing the Productive Uses of Electricity to Increase the Impact of Rural Electrification Programs,� ESMAP/The International Bank for Reconstruction and Development/The World Bank, Washington, DC. ESD Africa. 2006. “A Study of Productive Uses of PV Electricity in Rural Areas of Kenya.� Final Report to The World Bank, ESD Africa, Nairobi. Etcheverry, J. 2003. “Renewable Energy for Productive Uses: Strategies to Enhance Environmental Protection and the Quality of Rural Life,� University of Toronto. FAO. 2000. “Solar Photovoltaics for Sustainable Agriculture and Rural Development.� Environment and Natural Resources Working Paper No. 2, http://www.fao.org/sd/egdirect/egre0057.htm. FAO. 2000. “The Energy and Agriculture Nexus.� Environment and Natural Resources Working Paper No. 4, FAO. Rome, Italy. http://www.fao.org/sd/EGdirect/EGre0058.htm. Farinelli, U. (ed.). 1999. “Energy as a Tool for Sustainable Development for African, Caribbean and Pacific Countries.� European Commission/UNDP, Brussels, New York. Finucane, J. 1998. “Promotion of Productive Uses of Rural Electrification in Indonesia.� 70 INGENS STATUS REPORT JUNE 2008 Presentation of the SHS Project Support Group, Indonesia, The World Bank. Fishbein, R. 2003. “Survey of Productive Uses of Electricity in Rural Areas.� Report for The World Bank/AFTEG, Washington, DC. GEF / FAO. 2002. “Lessons From Emerging Experience With Productive Uses,� Summary Report, GEF-FAO Workshop on Productive Uses of Renewable Energy, 18-20 June 2002, Rome, Italy. GTZ / Integration. 2005. "Productive Uses of Energy in Chitral District, Pakistan.� GTZ, Eschborn, Germany. GTZ / Integration. 2005. “More than 50 Micro and Small Business Ideas and Opportunities with Renewable Energy Components in Rural and Peri-Urban Mongolia,� GTZ, Eschborn, Germany. GTZ / Integration. 2006 “Options for productive use of energy in the villages of the Lokoho rural grid area.� Substudy 6, Feasibility Report, Lokoho Hydro Power for Rural Development Project, GTZ, Eschborn, Germany GTZ. 2004. “Tibet: Small-Scale Power Supply for Rural Development.� Impact Report, GTZ, Eschborn, Germany. Haeusler, S., N. N’Guessan, and C. Dessallien. 2007. “ Evaluation a Mi-Parcours du Programme Regional RAF/03/013 : Energie pour la Production de la Pauvrete/Plates-formes Multi- Fonctionnelles (PREP).� Rapport final, UNDP, Bureau régional pour l'Afrique. Herberg, D. 2003 “Photovoltaics for Productive Uses in Rural Areas - A General Overview.� Internal Paper, World Bank, Washington DC. Hilman, H., J. Gidwani, and E. Morris. 2007. “Using Microfinance to Expand Access to Energy Services: The Emerging Experiences in Asia of SEWA, SEEDS, NUBL and Amret.� The SEEP Network, Washington, DC. Jacobs, G. 2006. “Renewable Energy for Productive Use: What SMEs Should Be Targeted in Botswana?� Commissioned by InWEnt-Internationale Weiterbildung und Entwicklung gGmbH for the Integrated Southern Africa Business Advisory (INSABA), Botswana. Kabutha, J., M. Sengendo, J. Winiecki, and E. Morris. 2007. “Using Microfinance to Expand Access page 71 to Energy Services: The Emerging Experiences in East Africa of Faulu Kenya and Kenya Union of Savings and Credit Cooperatives (KUSCCO).� Working Draft, The SEEP Network, Washington, DC. Kapadia, K. 2004. “Productive Uses of Renewable Energy: A Review of Four Bank-GEF Projects� Draft, University of California/Berkeley. http://rael.berkeley.edu/old-site/kamal.pdf Kirubi, C. 2006. “How Important is Modern Energy for Micro-Enterprises? Evidence from Rural Kenya.� Master’s thesis, University of California, Berkeley, Maia, S. 2005. “Electrifying the Market.� United Nations Environment Programme (UNEP), Our Planet, 16 (3). Mathur, S. 2005. “Productive Use Study: Key Conclusions and Recommendations Tanzania.� The World Bank. Matthieu, P. 2002. “Identificación de las Aplicaciones Productivas de la Energía Eléctrica en Sitios Pilotos del Proyecto PERZA, Nicaragua.� Internal Report, The World Bank. Mbaiwa, J. 2004. “Domestic Use versus Income Generating Activities in Delivering Modern Energy to Rural Botswana.� Journal of Energy in South Africa, 15 (4). Meadows, K. and C. Riley. 2003. “Modern Energy: Impacts on Micro-enterprises - A Literature Review into the Linkages Between Modern Energy and Micro-Enterprise.� Department for International Development (DfID), London. Menelaws, R.D. 1999. “Best Practices Guide for Productive Use Marketing.� PT. PLN, Indonesia/The World Bank, Washington, DC. Modi, V., S. McDade, D. Lallement, and J. Saghir. 2006. “Energy and the Millennium Development Goals.� New York: Energy Sector Management Assistance Programme, United Nations Development Programme, UN Millennium Project and The World Bank. Morris, E., J. Winiecki, S. Chowdhary and K. Cortiglia. 2007 “Using Microfinance to Expand Access to Energy Services�. Working Draft, The SEEP Network, Washington, DC. Motta, M. and K. Reiche. 2001. “Rural Electrification, Micro-finance and Micro and Small Business 72 INGENS STATUS REPORT JUNE 2008 (MSB) Development: Lessons for the Nicaragua Off-grid Rural Electrification Project.� The World Bank, Washington, DC. Nordstrom, S. 2005. “UNDP/GEF Programming Kit on Productive Uses of Renewable Energy.� Draft Paper, Version 1. UNDP/GEF. Pouliquen, L. 2000. “Poverty and Infrastructure� Background Paper for 2000/01 World Development Report, World Bank, Washington DC. Qase, N. and W. Annecke. 1999. “Energy Provision for the Urban Poor: South African Country Case Study.� Energy and Development Research Centre, University of Cape Town, South Africa. Ramani, K.V. and E. Heijndermans. 2003. “Energy, Poverty, and Gender.� The International Bank for Reconstruction and Development/The World Bank, Washington, DC. Saghir, J. 2005. “Energy and Poverty: Myths, Links, and Policy Issues.� Energy Working Note No. 4, The International Bank for Reconstruction and Development/The World Bank, Washington, DC. Sarr, S. and J. P. Thomas. 2005. “The Role of Renewable Energy in the Development of Productive Activities in Rural West Africa: The Case of Senegal.� Final Report, Renewable Energy Technology Working Group, Global Network on Energy for Sustainable Development, Enda-TM. Songco, J.A. 2002. “Do Rural Infrastructure Investments Benefit the Poor? Evaluating Linkages: a Global View, a Focus on Vietnam.� World Bank Policy Research Working Paper No. 2796. The World Bank / ESMAP. 2001 “Peri-Urban Electricity Consumers – a Forgotten but Important Group; What Can We Do to Electrify Them?� Report No. 249/01, The World Bank/ESMAP, Washington, DC. The World Bank. 1995. “Indonesia: Second Rural Electrification Project.� Staff Appraisal Report, Report. 12920-IND, The World Bank Industrial and Energy Operations Division, Country Department III, EAP Regional Office,. The World Bank. 1996. “Rural energy and development: Improving energy supplies for two billion people.� The World Bank, Washington, DC. The World Bank. 1999. “United Mexican States: Renewable Energy for Agriculture Project.� Report No. 19912-ME, Project Appraisal Document (PAD), Sustainable Development Department, Mexico page 73 and Colombia Country Management Unit, LAC. The World Bank. 2001. “India Power Supply to Agriculture.� Volume 1, Summary Report, Report No. 22171-IN, The World Bank, Energy Sector Unit, South Asia Regional Office. The World Bank. 2003. “Nicaragua: Rural Electrification Project (PERZA).� Project Information Document (PID) 11399, The World Bank, Washington, DC. The World Bank. 2003. “Philippines: Rural Power Project.� Report No. 24284, Project Appraisal Document (PAD) P066397, Washington DC. The World Bank. 2004. “Indonesia: Renewable Energy Small Power Project.� Report No: 30244, Project Completion Note, Washington DC. The World Bank. 2005. “Electricity for All: Options for Increasing Access in Indonesia.� The World Bank, East-Asia and Pacific Region Energy and Mining Unit and Indonesia Country Program, Jakarta/Washington DC. The World Bank. 2006. “Bolivia: Decentralized Electricity for Universal Access.� Report 37936, Project Information Document (PID) P102479, Washington DC. The World Bank. 2006. “Indonesia: Electricity Access for Rural Transformation.� Report AB2431, Project Information Document (PID) P093047, Washington DC. The World Bank. 2007 “United Mexican States: Renewable Energy for Agriculture Project.� Report No. ICR-000044, Project Implementation Completion Results Report, Mult-23251, Sustainable Development Department, Mexico and Colombia Country Management Unit, LAC. The World Bank. 2007. “Mexico: (CRL) Rural Electrification Project.� Report No. AB2919, Appraisal Stage-Project Information Document (PID) P08896, Washington DC. The World Bank. 2007. “Sri Lanka: Renewable Energy for Rural Economic Development– Additional Financing.� Project Information Document (PID) P101699, Washington DC. UNDP / The World Bank. 2002. “Rural Electrification and Development in the Philippines: Measuring the Social and Economic Benefits.� Joint UNDP/World Bank Energy Sector Management Assistance Programme (ESMAP), Washington, DC. 74 INGENS STATUS REPORT JUNE 2008 Usher, E. 2007. “Financing the Path to Renewable Energy Enterprises.� Stockholm Environment Institute, Renewable Energy for Development. Vol. 20, No. 2. Wagner et al. 2003. “Projet de l’Electrification Rurale d’Initiative Locale (ERIL).� Rapport pour la Promotion de l’Electrification Rurale et de l’Approvisionnement Durable en Combustibles Domestiques-PERACOD. Republique de Senegal et GTZ, Eschborn, Germany. White, R. 2002. “GEF/FAO Workshop on Productive Uses of Renewable Energy: Experience, Strategies, and Project Development.� Workshop Synthesis, GEF/FAO Workshop on Productive Uses of Renewable Energy, Rome, Italy. ftp://ftp.fao.org/SD/SDR/SDRN/GEF- FAO_productive_uses_workshop.pdf World Bank Independent Evaluation Group (IEG). 2008 “The Welfare Impact of Rural Electrification: A Reassessment of the Costs and Benefits.� The World Bank, Washington, DC. page 75 ANNEX: THE INGENS QUESTIONNAIRE Provide Authors?! See Separate Document “Ingens_Questionnaire_CONSOLIDATED_fieldversion_Jan08� 76 INGENS STATUS REPORT JUNE 2008 ANNEX: INGENS Enumerator Training TBD for Final Report page 77 ANNEX: Lessons Learned from Ghana Fieldwork Provide Authors?! Length of Interview The time required to complete the interview improves significantly with experience of the interviewer. In the pre-test, interviews typically took one-and-a-half to two hours, which was too long both to maintain quality and to complete the intended number of interviews; hence some cut-backs had to be made in the amount of information requested. In the first couple days in the field, interviews were typically in the range of one to two hours – but most took more than the targeted 70 minutes. By the end of the week, it appeared that 70 minutes was about the median. The interviewers started out with about 4-5 interviews per day, and eventually reached 6-7 (the number depends also on whether the enterprises are concentrated in one location and how early the enumerators are able to begin). The sections that took the most time were Real Capital Endowment and Employment, especially for medium sized firms. Even when the questionnaire length is suitable for a 70-minute interview on average, the actual interview period is often actually longer, because of interruptions for the interlocutor to take care of clients. Thus, in programming the number of interviews that can be completed per interviewer per day, it is prudent to allow for about 90 minutes per interview. Incentive Offering to buy a drink in the middle of longer interviews is a good idea, but in some cultures might risk embarrassing the interviewee for not offering a drink as “host,� and in any case would have added a complication in Ghana of how to provide and manage the finances (can neither expect the enumerators to provide out of their pockets nor to be able to provide receipts). In the Ghana case, GTZ provided T-shirts celebrating the 50th Anniversary of Cooperation between Ghana and Germany as a gesture of appreciation for cooperating, and these were highly prized (and indeed sought after by others not involved in the exercise). 78 INGENS STATUS REPORT JUNE 2008 In principle, we think it is NOT a good precedent to provide monetary compensation for being interviewed, but some non-monetary incentive such as a drink, T-shirt, etc., is useful in building good will for future surveys. Making it Easier to Locate Firms for Future Studies In cases where the initial survey is a baseline in order to evaluate impact through follow- up tracer surveys in the future, it might be desirable to print out stickers with a logo that enterprises could put next to their door – both because of the prestige associated with having a GTZ or other sticker, and as an occasion to let them know that we may be seeking them again in a year or two to follow up. Although in Ghana an enumeration had been done the year before to serve as a sampling frame, it was belatedly discovered that the coding of the data in SPSS precluded randomly selecting the firm names and associated location information for the interviews. (Discussed further in next section.) It was necessary to have the original enumeration sheets that had the detailed information on Area of Town; Street; Cross-street; Other landmarks. Highly accurate data is especially important in countries where street names and house numbers are not commonly used. These are recorded in Part 1 of the INGENS questionnaire. Anticipating that we may wish to follow up with a sample of these same firms in the future, we can anticipate that it may be desirable to provide future enumerators with a copy of the front page of the original questionnaire form providing the detailed location information together with the firm name – which we suggest be included on p.1 for this purpose, not just on p.2. Often enterprises do not have business names or are rather known by their owner’s name. Therefore, it would be prudent to also include the owner’s name at the front page. (It is also important to investigate whether the full detailed information can be entered into the database; or at least, retain a detailed sheet giving the details corresponding to the codes used.) Questionnaire Modifications Line of Business Codes page 79 It would be better to have the subsectors listed in order by standard industrial classification (rather than alphabetically) and to have a clear rationale for disaggregation of subsectors into additional categories. In some cases, the rationale is probably related to energy (e.g., candle making, pottery, and brick-making). In other cases, it is not clear (e.g., cushion making, arts and crafts). Many of the detailed categories are unlikely to have enough respondents to be useful analytically – unless a survey is being undertaken in a location where a particular activity is exceptionally concentrated (e.g., handicrafts). The following subsector groupings are proposed for analysis and comparisons across countries: Subsector Groupings M Manufacturing M2, M7, M14,M16, M17, M18 Food Processing M13, M15, M22 Textiles, Shooes M5, M6, M9, M9, M10, M11, M12, M20 Wood, carpentry, construction M4, Metalworking M0, M1, M3, M8, M19, M21 Other manufactring S Services S4, S10, S17 Food and lodging S5 Hairdressng 35 S2, S6, S8, S9, S16 Repair and servicing (bicycles, automotive, electric, electronic S7, S12, S21, S22, S23 ICT S18, S24 Commerce S0, S1, S3, S11. S13, S14, S15, S19, S20 Other services In some instances, more disaggregation might have been useful, e.g., between restaurants, “chop bars,� and open-air grilling or boiling of food for sale. It might have been possible to get at these distinctions by type of structure (permanent building, kiosk, shed, open air), but unfortunately there is no question on structure. 35 Repair services (vehicle, electronic, shoe) are now usually counted with manufacturing in international industrial classifications, but it is too late to change, so can be left here for consistency across INGENS studies. 80 INGENS STATUS REPORT JUNE 2008 Type of Structure We suggest that type of structure be added in the future, as a possible variable to discriminate the degree of formality of the enterprise. The following codes are proposed, to be grouped broadly into “informal� (no permanent structure) and “permanent structures� (with kiosks as the dividing line): Informal (non-permanent) 36 1 Table, grill, cooking area: open air } Included ONLY if in 2 Table with umbrella } energy- related business: 3 Shed (covered, sides open or partly) } e.g., cooking, grilling; 4 Stall movable) } phone service (not just selling Permanent structures enumeration sheet 5 Kiosk 6 Container 7 Business operated out of a home (no separate structure for the enterprise) 8 Shop, permanent structure (in building, row of shops; or small shop by itself) 9 Separate building/factory/hospital/school in own premises with own grounds and entrance Part I There is a need to have indicated the specific and precise location of enterprises on the front page to be able to identify the same enterprises during subsequent surveys. The front page can then be photocopied to assist the interviewer to find the firm in future follow-up surveys. [It is difficult to capture all the location details in the database.] Area characteristics have been coded. [“1� is reserved for new industrial zones (being established by GTZ in Ghana as an intervention).] Section B: Growth constraints 36 To avoid being overwhelmed by petty retail trade, which generally accounts for the majority of MSMEs, the Ghana study included retail trade unless in a kiosk or permanent structure, and included “informal� activities only if they were in some way energy-using or –related (e.g., cooking, grilling, telephone services, vulcanizing, battery charging). page 81 Although it is desirable from the analysis and comparison standpoint to have this question in line with World Bank’s Doing Business, this proved problematic in the field. For our target group, some of the categories were too abstract and therefore not understandable. If compatibility to Doing Business should be maintained, for future surveys the categories need to be supplemented by real life examples to make them understandable. Section C.1: Electricity We dropped the mini-grid option from Q18, as we had no mini-grids in our target area. Section C.5: Lighting In the pre-testing, getting the wattage of individual bulbs was highly problematic due to several reasons: lamps inaccessible, old lamps, type unreadable or not printed on, bad quality lamps. Overall, the main issue in Ghana seemed to be the quality of lighting products rather than their wattage. The bulbs generally did not have the wattage visible on them, and there was no basis for determining it (unlike Uganda, enumerators had no electrical engineering background); nor was it desirable to have 8 different enumerators making guesses. Given the need to cut down the overly long interview time, it was decided only to count the number of bulbs by general category (energy-saver, incandescent, fluorescent), and to try to get an idea of which wattages are available by asking the sellers during the market survey (of bulb prices). In fact, the number of types available in the towns was rather limited. Therefore, based on market survey of availability, we propose to use the following standard wattages for analysis: Energy Saver 18 W, Incandescent 60 W, Fluorescent tubes 20 W. Question 26 was also adjusted so that it reads “How much was the last payment you made for electric energy from the grid?� 82 INGENS STATUS REPORT JUNE 2008 Section D: Water and Waste These questions were added at the suggestion of GTZ Ghana to get information on access to water and sewerage services (another aspect of infrastructure) and to assess the impact of MSMEs on the environment. Section E: Finance Q81-84: It was judged sufficient to have detailed information on the most recent loan, without asking for similar information on previous loans (of which there are likely to be very few, and recall ability would adversely affect the quality of data). We coded the Financial Institution Type to ease analysis. Section F: Business Development Services Technical services/maintenance was introduced as a further category of business services. Section G: Markets The question on sales margin was extended to also apply to services. Section I: Employment An additional question is needed Q128a to ask “Is the _____ an apprentice?� for two reasons:  There is strong interest in apprenticeship as a means of passing on skills, and also because the saturation of supply in some subsectors is attributed to the large number of graduated apprentices trying to establish their own businesses in competition with their former masters;  Accurate calculation of average wage will require calculating the ratio of weekly pay per employee ONLY for wage employees, not for unpaid workers such as family members and apprentices. page 83 [Note: In analyzing the Uganda data to obtain average wage, we would suggest removing those employees whose weekly pay is obviously below the minimum for wage workers.] Enumerators in Ghana reported that it was difficult to get a standard monthly or weekly payment for each employee, as wage differs from bad weeks to good weeks. Section J: Real Capital Endowment Age of equipment was dropped for some smaller items to speed up the interviews. On the other hand, we tried to include detailed information (age, wattage) for electricity consuming devices. Sometimes it was difficult to get the breakdown of all different materials from the respondents. Overall, this section worked better with retail than with other categories of enterprises. Section L: Owner Information This section was adapted to capture if the owner is not personally managing the enterprise and a manager is being interviewed. Possible answers for Question 150 was adjusted to include “Don’t own a house but leaves in a family house and was numbered 4� . 84 INGENS STATUS REPORT JUNE 2008 ANNEX: Productive Uses in Project Practice – Case Examples TBD for Final Report page 85 ANNEX: Some Pictures from INGENS Cases 86 INGENS STATUS REPORT JUNE 2008 page 87 88 INGENS STATUS REPORT JUNE 2008 page 89 90 INGENS STATUS REPORT JUNE 2008 page 91 92 INGENS STATUS REPORT JUNE 2008 page 93 94 INGENS STATUS REPORT JUNE 2008 page 95 96 INGENS STATUS REPORT JUNE 2008 page 97 98 INGENS STATUS REPORT JUNE 2008 page 99 100 INGENS STATUS REPORT JUNE 2008 page 101 102 INGENS STATUS REPORT JUNE 2008 page 103 104 INGENS STATUS REPORT JUNE 2008 page 105 106 INGENS STATUS REPORT JUNE 2008 page 107 108 INGENS STATUS REPORT JUNE 2008 page 109 110 INGENS STATUS REPORT JUNE 2008 ANNEX: INGENS Case Countries at a glance TBD for final Report INGENS Modules - New Funding Needed ADDITIONAL OUTPUTS ADDITIONAL ESMAP FUNDING [US$ Thousands] Output No new Low Case High Case None Low Case High Case Can be ask now & Priority done in do by Team esmap FY09 May09 tranche Limited Feedback Workshops / Dissemination x Output 1-10 can be done by GTZ … Results country + meta country x …with already committed GTZ fundin Final Methodology + Conceptual framework (other determinants) x Annotated Biblio x Context Chapter (Intro+Background) x Results: Limited Lessons, Recommendations & Conclusions x Annex: (annotated) instrument, TORs, budget, training process+supervision guidelines x editing & printing ? Limited Annex: "cases + cash flows" x Extended Feedback Workshops + Country Dissemination 5 10 x Ghana nation 1 Diessemination Donors (AEI Presentation; Editing; "Glossy Publication" + CD + web) small full 10 20 link to Doug: HH AddIE + LongIE yes yes 5 5 x Practitioners' Manual: Advice on EP Design and Complementary Services Recomm. Chapter Stand-alone Manual 20 45 x 1 Recommendations: Practical issues to whatch out for (e.g. example calculations - ) limited full x Conclusions: Framework for Operations integrating complementary services yes x Proper Case & Cash Flows yes x Improved Manual on Evaluation: AddIE Toolkit! yes yes 20 20 x Advanced Analysis (e.g. access vs level, production function, matching, etc) yes yes incl above incl above x new country: Senegal N=350, 2 areas N=500, 4 areas 25 35 x Senegal 1 more treatments (per country) - e.g., BDS, MFI, other infrastructure yes: senegal yes: senegal incl above incl above MDG AddIEs: food/agro AddIEs (NB: link to agro) (i) draft tool Draft instrument Draft instrument 15 15 x WB large, GTZ small (ii) Ditto - field work + prel analysis N=350; 3 areas N=350; 3 areas 25 25 difficult Other AddIEs: health, education, others (water, roads, ICT, …) non-electrical energy x 1 offgrid power x 1 other donors no Apply AddIE in additional country+analysis: ca 30k-50k per new country ? Supervision + cooperation GTZ-WB 15 20 1 Total Additional ESMAP Funding needed 0 140 195 ANNEX: Budget Estimates for Outlook page 111