Policy Research Working Paper 9035 Is Informality Good for Business? The Impacts of IDP Inflows on Formal Firms Sandra Rozo Hernan Winkler Social Protection and Jobs Global Practice October 2019 Policy Research Working Paper 9035 Abstract This paper examines the effects of large inflows of internally where people from their origin locations settled earlier. displaced persons (IDP), who are primarily absorbed by the The paper finds that large inflows of IDP induce sizable, informal sector, on the behavior of formal manufacturing negative effects on the intensive and extensive margins of firms in Colombia. To identify causal effects, the analysis production of formal firms. These effects are stronger for employs annual, firm-level panel data between 1995 and firms operating in sectors that face a stronger competition 2010 and exploits that when conflict intensifies, forcefully from the informal economy. displaced individuals tend to migrate to municipalities This paper is a product of the Social Protection and Jobs Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at hwinkler@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Is Informality Good for Business? The Impacts of IDP Inflows on Formal Firms Sandra Rozo∗ Hernan Winkler† JEL Classification: D22, J61, O17. Keywords: Forced Migration, Firms, Informality. ∗ USC Marshall School of Business. Corresponding author. Email: sandra.rozo@marshall.usc.edu † The World Bank. We would like to thank the Colombian Statistics Department [Departamento Nacional de Estad´ıstica] for providing the data for this study. We thank Stephen Trejo and two anonymous referees for their suggestions. We are also grateful to Maria Jose Urbina for her excellent work as a research assistant. We also thank the participants of the World Bank Policy Forum on Impacts of Refugees and IDP on Host Communities for their comments and suggestions. The findings, interpretations, and conclusions in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank Group, its Executive Directors, or the countries they represent. This paper uses confidential data from the Colombian statistics Agency (DANE). To obtain access to the micro data authors should send a formal letter to: contacto@dane.gov.co. Upon receipt, DANE will request the signature of an information agreement. DANE will allow the researchers to keep the results of their estimates, but not the micro data. The agreement does not pose any other restrictions on the investigation. There are no conflicts of interest or financial support to disclose. I Introduction Forced displacement due to violence and conflict has hit an all-time high. By the end of 2017, the number of individuals forcibly displaced reached 68.5 million (UNHCR, 2018). Sudden and large inflows of forced migration may induce sizable effects on hosting economies, not only by increasing demand for public services, but also by modifying the decisions made by workers and firms. To adequately guide public policy to help forced migrants and their hosts cope with these shocks, it is crucial to properly understand these effects. This article explores the effects of a sudden and large wave of internally displaced persons (IDP), who are primarily absorbed by the informal sector, on the behavior of formal firms in Colombia. Previous literature has largely focused on examining the effects of voluntary migrants on firm behavior within countries with a low incidence of informality.1 The effects of a large wave of low-skilled forced migrants, however, deserves a separate analysis, for several reasons. First, forced migrant inflows are disproportionately concentrated in developing countries with large in- formal sectors. In fact, according to a recent report of the United Nations Refugee Agency, 86 percent of the world’s forcefully displaced individuals are in low and middle-income countries close to conflict situations (UNCHR, 2015) and the World Bank estimates that this figure is ac- tually higher and closer to 99 percent (World Bank, 2016). If forced migrants join the informal sector upon arrival, formal firms may be negatively affected by the migration shock via the unfair competition of informal firms that do not pay taxes or comply with regulations. Second, forced migrants have characteristics that are drastically different from those of voluntary migrants. Forced migrants are less likely to be positively self-selected in terms of skills than are economic migrants, whose migration decisions are more tied to expected labor market success (Chiswick, 1999). They face great uncertainty about the duration of their stays in hosting economies and are more likely 1 Lewis, 2011; Kerr et al., 2015; Dustmann and Glitz, 2015; Ottaviano et al., 2018, for example, study the impact of voluntary migrants on firm outcomes. In these studies—as is to be expected when considering a standard labor demand and supply model—a increased number of immigrants pushes wages down, causing reductions in operation costs and changes in capital-labor ratios (see Lewis, 2011 and Dustmann and Glitz, 2015 for examples). Consequently, many of these studies find evidence of unskilled immigration having positive effects on total employment, firm output, and firm creation (Kerr et al., 2015; Dustmann and Glitz, 2015). 2 to be affected by trauma and distress, since they tend to have recently fled from wars (Moya et al., 2012). These characteristics may complicate the efficient integration of displaced individuals into formal reception markets. We focus on the case of Colombia because its unique characteristics make it an ideal case with which to pursue this study. The escalation of the Colombian armed conflict in the late 1990s and the 2000s induced large sudden flows of displaced individuals. According to data from the Human Rights Observatory, internal conflict in Colombia between 1995 and 2010 displaced approximately 5.8 million people. Colombia also collects among the most complete and rich firm-level panel data available for the study of firm behavior in developing countries. The Annual Manufacturing Survey, a census of all manufacturing firms of more than 10 employees, includes plant-year level information on sales, wages, employment, and capital as well as product-plant-year information on output and input prices. To identify causal effects, we use a panel-instrumental variable methodology. We construct the instrumental variable for inflows of IDP following the standard approach in the literature, which combines early settlements of migrants with time trends on migration outflows.2 Our geographic variation comes from the fact that IDP move disproportionately to municipalities where there are early settlements of populations from their municipalities of origin. Our time variation comes from observed outflows of displaced populations by municipality and year due to conflict shocks. We construct the predicted inflow of immigrants by combining municipal cross-sectional information from the Colombian population census of 1993 (the last population census before the intensifica- tion of the conflict) with time-varying data on the total number of individuals expelled from each municipality. Our instrument is a strong predictor of the observed inflows of IDP between 1995 and 2010.3 We find that large inflows of IDP induce sizable negative effects on formal firms’ intensive 2 See Card, 2001 and Altonji and Card, 1991 for the pioneer approaches and Lewis and Peri, 2015 for a review of the literature on applications. 3 A new criticism of the validity of this type of shift-share instrument was recently proposed by Jaeger et al. (2018). The authors suggest that using pre-settlements of migrants in countries where migration flows are stable in time confounds short- and long-term causal effects. Our identification strategy is not sensitive to their critique because the inflows of forced migrants were sudden and large in scale as a consequence of the intensification of armed conflict. 3 and external margins of production but no effects on firm prices or input demands. Our estimates suggest that when inflows of IDP increase by 1 percent, the production of formal firms drops by approximately 0.3 percent and the probability of firm exit increases by 0.2 percentage points. These correspond to sizable effects, given that the average municipality registered an annual growth of 22 percent in the number of forcefully displaced individuals received between 1995 and 2010 in ıve, back-of-the-envelope calculation would suggest that Colombia. Based on these numbers, a na¨ formal firms located in the average municipality should have seen an approximately 6.6 percent annual decrease in their formal production between 1995 and 2010 as a result of the large number of IDP. Our main estimates include fixed effects by firm and year, and as such are not sensible to aggre- gate time trends or time-invariant firm characteristics. We also show that our estimates are robust to the inclusion of a battery of controls, including regional time trends, violence and conflict- related covariates, and differential pre-trends in economic conditions, government size, and vio- lence/conflict levels. When exploring the mechanisms driving our results, we document that larger inflows of dis- placed populations are positively associated with an increase in the size of the informal sector. This may be explained by the characteristics of the forcefully displaced populations. Forcefully displaced individuals commonly arrive in new locations without legal identification documents, tend to have low education levels, lack experience in jobs that have high local demand (as many IDP move from rural to urban areas), and may be affected by traumatic events that complicate their integration into formal markets (see Ib´ an˜ ez and Moya, 2006). IDP thus tend to look for low-tier jobs, which are most likely available in the informal sector (as documented extensively by Amaral and Quintin, 2006; Perry, 2007; Galiani and Weinschelbaum, 2012; La Porta and Shleifer, 2014a; Meghir et al., 2015), increasing its size. An increase in the size of the informal sector may induce negative effects on the performance of formal firms, in several ways. Formal firms, for example, face higher costs relative to infor- mal firms as they pay taxes, fees, and higher wages for their employees (La Porta and Shleifer, 4 2014a). Informal firms can thus usually offer lower prices, thereby competing unfairly with for- mal businesses. Unfair competition from informal firms can also slow down the process in which inefficient firms can be replaced by more efficient competitors and negatively affect the incentives of formal firms to innovate and adopt new technologies, as these innovations can be easily stolen (Perry, 2007). At the same time, since informal firms are able to use public goods, but do not pay for them, they lower their quality and crowd out their use by formal firms (Besley and Persson, 2013). We also document that the negative effects of IDP inflows are concentrated on the group of formal firms operating in sectors where informality is more predominant, and that as such may face stronger informal competition. Our results are in line with anecdotal evidence from several media outlets, which consistently report the close connection between forced displacement and the size of the informal sector in Colombia. More particularly, journalists document that forcefully displaced individuals have a hard time integrating into formal markets given their previous experience in agriculture (which is not common in urban areas) and their low education levels, and that, consequently, often their only economic choice is to work as self-employed individuals and join the informal sector selling products on the streets (see IPS, 1999; El Espectador, 2009; El Tiempo, 2010; El Tiempo, 2014; for some examples). Reports also suggest that even if displaced populations manage to find a job in the formal or informal sectors, they have a high likelihood of receiving less favorable labor conditions than the rest of the population with comparable characteristics (see Huffington Post, ıa, 2012). 2010; La Silla Vac´ This paper contributes to three strands of economic literature. One strand it contributes to is the research exploring the effects of unskilled migration on firm-level outcomes. Papers examining the effects of unskilled migration on firms have focused mainly on developed economies.4 These stud- ies examine firm-level outcomes such as productivity, imports, exports, investments, wages, entry, exit, and relative skill mix. Their findings suggest that immigrants have positive effects on firm- 4 These include the cases of Spain (Carrizosa and Blasco, 2009), the United States (Lewis, 2011), Italy (Accetturo et al., 2012), the United Kingdom (Ottaviano et al., 2018), and Germany (Dustmann and Glitz, 2015). 5 level outcomes.5 This study contributes to this literature by documenting the effects of unskilled migration in an economy with a large informal sector and by examining the case of forcefully displaced migrants, whose characteristics differ drastically from those of voluntary migrants. This paper also adds to the growing number of studies that examine the impact of forced mi- grants in hosting economies. Most of this literature has been focused on the United States and the Middle East and documents the impact of refugee inflows on employment and prices (see Bor- jas and Monras, 2017; Clemens and Hunt, 2017; Del Carpio and Wagner, 2015; and Ceritoglu et al., 2017 for examples). We contribute to this literature by studying the effects of IDP on firm performance. The paper that is most similar to this study is that of Altindag et al. (2018), who study the impact of Syrian refugee migration on firm behavior in Turkey. In contrast to our re- sults, these authors document that refugee inflows had a positive impact on firm creation in the construction and restaurant sectors in Turkey and no impact in the other sectors of the economy. The difference between our findings and theirs may be explained by the fact we study the impacts of IDP migration and not refugees. IDP move within the frontiers of a conflict country with a weak economy and a weak state. On the other hand, refugees are presumably wealthier and better educated than IDP that are locked in their countries. Therefore, the effect of refugees and IDP on the performance of firms must be different. Particularly, Colombian IDP had previously worked primarily in agriculture, whereas the previous employment of Syrian refugees was more diverse. IDP in Colombia, consequently, were completely absorbed by the informal sector, whereas there was some integration of refugees into formal firms in Turkey (although not necessarily into formal 5 For instance, these studies typically document greater firm productivity, driven mainly by lower production costs and skill complementarities in the workplace. They also examine the effects of immigration on capital investments, where the results are mixed. Lewis (2011), for example, finds that plants in areas that received more unskilled im- migrants were less likely to adopt automation machinery, which served as a buffer for the effects of immigration on wages. Accetturo et al. (2012) and Ottaviano et al. (2018), in contrast, find that firms in Italy and the United Kingdom increase their capital investments in response to immigration from developing countries, arguably because firms tend to offset the skills-downgrading effect with more capital accumulation. The latter also finds that immigration acts as a substitute for offshoring—by lowering the intermediate imports from the immigrants’ countries of origin—and tends to increase exports to the immigrants’ countries of origin because it helps reduce information barriers and trade costs. Finally, Dustmann and Glitz (2015) find that the responses of firms to an influx of immigrants in Germany depends on their sector of economic activity. While firms in the non-tradable sector respond by lowering wages, their tradable sector counterparts primarily respond by scaling up their employment and changing their skill mix. In addition, they also find a positive net entry effect of firms in the tradable sector (i.e., firm creation minus exit). 6 jobs within those firms). Our paper thus contributes to this literature by identifying the effects of a internal forced displacement shock from a low-skilled population that, before the onset of conflict, had been primarily concentrated in agriculture and that was thus mostly absorbed by the informal sector upon arrival in hosting locations in the short term. Third our paper, contributes to the literature on the economic impacts of armed conflict (see Abadie and Gardeazabal, 2003, Guidolin and La Ferrara, 2007; Camacho and Rodriguez, 2013; uck et al., 2013 and Blattman and Miguel, and Serneels and Verpoorten, 2015 for examples and Br¨ 2010 for literature reviews on this topic). We contribute to this group of studies by focusing our analysis on the impacts of IDP migration caused conflict within a state. II Colombian Context According to information from the Colombian Human Rights Observatory, between 1995 and 2010 approximately 5.8 million people were internally displaced by violence in Colombia, accounting for approximately 11 percent of the total Colombian population of 2010 (see Figure I). The number of displaced individuals, however, may be even higher, since these figures only include individu- als who searched for governmental support when arriving at their new locations. The same data suggest that, between 1995 and 2010, 82 percent of municipalities received at least one of these migrants (see panel b of Figure I). The escalation of the Colombian conflict —-fought between the country’s guerrilla groups, paramilitary vigilantes, and armed forces-— was the main reason for forced displacement in the late 1990s and the early 2000s (Engel and Ib´ an˜ ez, 2007).6 Forced displacement, more specifically, 6 As documented in Rozo (2018), the Colombian internal armed conflict intensified in the middle of the twentieth century with the formal creation and growth of illegal armed groups. In 1964, adherents of a Cuban-style revolution founded the National Liberation Army (known by its Spanish acronym, ELN). Later, in 1966, a second left-wing group, the Revolutionary Armed Forces of Colombia (FARC in Spanish), was founded as the union of all the remaining communist guerillas. Initially, both groups claimed to defend the interests of the rural poor, aiming to overthrow the government and to install a Marxist regime. In time, however, the motivations of both groups became primarily economic. Paramilitarism began in the late 1980s as an anti-insurgent response by landowners and drug traffickers to left-wing guerillas’ actions in areas where the state was unable to provide security. In 1997, the paramilitary forces coalesced into the United Self-Defense Organization of Colombia (AUC in Spanish). By 2003, the AUC had declared a partial ceasefire, and some paramilitary blocs agreed to participate in a “disarming program” that concluded in 7 was not a causal by-product of the Colombian conflict, but an extremely common strategy of war used by illegal armed groups to weaken the enemy’s popular support, clear regions for illegal crop growing and drug trafficking, and expropriate lands and natural resources (Ib´ an˜ ez and V´ elez, 2008). According to Ib´ an˜ ez, Moya, and Vel´ asquez (2006), for instance, between 1993 and 2002 displaced individuals lost 1.2 million hectares of land. II.1 Characterizing internally displaced migrants ´ The annual cumulative displaced population from 1995 to 2010, according to the Registro Unico ıctimas (RUV) from the Colombian Human Rights Observatory is presented in Figure I.7 The de V´ figure shows that the number of displaced individuals began to increase dramatically between 1996 and 2002 in Colombia, when the internal conflict was most intense, as shown by the evolution of conflict-related variables such as armed actions and clashes between armed groups (see panel b of Figure II). Since then, forced displacement has been decreasing slowly in conjunction with a softening of conflict intensity and violence.8 Data from RUV also suggests that the cumulative population of forced migrants is balanced in terms of gender (51 percent women) and that most are of working age. Colombia’s forced migrants are also young. In particular, 39 percent of forcefully displaced individuals were 15 years old or younger at the time of displacement, this percentage is disproportionately larger than this age group within the population of Colombia as a whole (28 percent). Indeed, 15.5 percent of forced migrants were younger than 5 years of age at the time of migration. Households also tend to be bigger as several members of the extended family tend to live together to save on housing on y Reparaci´ costs (Unidad para la Atenci´ ıctimas, 2013). Previous studies, using surveys on de V´ given to migrants who were forcefully displaced, also report that this population has low education 2005. Many of the combatants that had been part of the AUC, however, later fused into new criminal groups that are known today as Bandas Criminales (BACRIM), illegal armed groups that obtained financial resources, mainly through extortion and drug trafficking, to carry on their activities. 7 The data is publicly available at: http : //rni.unidadvictimas.gov.co/RU V 8 The problem is still ongoing, however. At the time of writing, in mid-2017, the year’s forceful displacement rate has already surpassed 800. 8 levels (around 5 years of education) (see Ib´ an˜ ez and Moya, 2006; Garay, 2008; Carrillo, 2009). Several studies have attempted to characterize the migration decision of forced migrants using surveys. Their findings suggest that forced displacement in Colombia mostly originates from rural areas, where the internal armed conflict has taken place. In that sense, the migration decision of forced migrants is mainly driven by safety concerns related to the presence and activities of illegal armed groups. In particular, data from RUV suggests that for the 59 percent of individuals for whom information is available on the cause of displacement, 54 percent of migration cases were attributed to the activities of illegal armed groups and 84 percent were attributed to a death on y Reparaci´ threat (Unidad para la Atenci´ ıctimas, 2013). Most displaced individuals, on de V´ consequently, had previously worked primarily in agriculture (Ib´ an˜ ez and Moya, 2006; Carrillo, 2009). Households that had access to basic public services, had better economic opportunities, an or had private property showed a lower probability of migrating (Engel and Ib´ ˜ ez, 2007; Ib´ an˜ ez and Moya, 2010). In areas with extreme levels of violence, however, owning land increased the probability of displacement because these household were targeted for extortion by illegal armed groups (Engel and Ib´ an˜ ez, 2007). Forced migrants in Colombia have moved to areas where they had friends or relatives and that were closer in distance to their municipalities of origin (Ib´ an˜ ez and Moya, 2006; Carrillo, 2009; Lozano-Gracia et al., 2010). Yet, in regions with extreme violence, individuals preferred to relocate to more distant locations and to cities that were more populated; they were attracted to the sense of anonymity both provide (Carrillo, 2009; Lozano-Gracia et al., 2010). Other criteria migrants take into account when choosing their destinations includes the provision of public goods and the population density of the destination municipality (Carrillo, 2009; Lozano-Gracia et al., 2010). Households who migrate have incurred substantial losses in physical assets left behind in their municipalities of origin and have suffered human capital depreciation due to household disinte- gration or post-traumatic stress disorders (see Ib´ an˜ ez and Moya, 2010 for a quantification of these welfare losses). Most households, in addition, move to urban areas, where there is little demand for their agricultural experience. Many of them, consequently, face extreme hardship upon arrival 9 elez, in their new locations, facing living conditions less favorable than those of the urban poor (V´ an 2002; Ib´ ˜ ez and Moya, 2006). Ib´ an˜ ez and Moya (2006), more particularly, estimate that the con- sumption of displaced households falls by 35.7 percentage points upon migration and drops even further during the year following displacement. III Data III.1 Firm data Our main source of information is the Encuesta Anual Manufacturera (Annual Manufacturing ıstica (DANE), the Colombian statistics Survey), collected by the Departamento Nacional de Estad´ agency. These data set is a census of all manufacturing plants with ten or more workers or with a total output value larger than 65 million in 1992 Colombian pesos (approximately USD$95,000). Once a plant is included in the survey, it is followed over time until it goes out of business. The survey includes information on all production-related variables, including employment and wages. In conjunction with the standard plant information, the census contains information on all physical quantities and prices (valued at factory-gate prices) of each output and input used or produced by each plant. In this article, firms’ prices are defined as the plant-product-year observation estimated by dividing the value of revenues or expenditures by physical quantities. These data is regarded as one of the best and most complete sources of information for studying firm behavior in developing countries (Kugler and Verhoogen, 2011). The data set covers the years 1995 to 2010, though its data distinguishing employment and wages by white- and blue-collar workers is only available after the year 1999.9 Our sample consists of all municipalities where more than two formal firms with ten or more workers are observed.10 These municipalities are typically more populated and have higher eco- 9 The data also contains information on firms’ average output and input prices. We do not use this data, however, as it highly concentrated in the biggest cities for which geographic variation is considerably lost. 10 Municipalities where only one firm was observed were dropped from the sample by DANE due to confidentiality concerns. 10 nomic activity (see Appendix I for a comparison of the municipalities in the manufacturing sample and those not included). They also were strong recipients of IDP during the period of study (see Figure I), had lower levels of conflict-related violence, although higher levels of urban crime (mea- sured through homicide rates) relative to the rest of the country (see panel a of Figure II). Figure III shows the annual evolution of the mean of all the firm outcomes employed in this study. III.2 Internal forced displacement Municipal data on forced displacement caused by violent conflict was obtained from the Registro ´ Unico ıctimas (RUV, Registration of Victims) from the Unidad para la Atenci´ de V´ on on y Reparaci´ ıctimas of the Colombian government. This is the best source of information on in- Integral de las V´ dividuals who were displaced by violence because it combines all official sources available where individuals registered upon arrival in the new locations (even if they registered with authorities a long time after their arrival, in which case they are included in the year in which they arrived in the data). Although not all individuals register, these data are an excellent approximation of the total number of individuals received in each municipality because being registered is a condition for ac- cessing any type of governmental support from local authorities. To be registered, individuals have to declare their displacement under oath, attest to the exact dates of the event, their municipality of origin, and some of their socioeconomic conditions, as well as to describe the facts leading to their displacement. The RUV data, consequently, offers information on the municipalities from which migrants were expelled and into which they were received. It is available annually by municipality between 1984 and 2016. In this paper, however, we focus on the period between 1995 and 2010 because it is the period for which firm data is also available. The time evolution of total displacement is presented in Figure I and shows strong variation between 1995 and 2010. The total number of receiving municipalities is presented in panel b of the same figure (Colombia has 1,122 municipal- ities). 11 Figure IV presents the geographic distribution of the intensity of migration outflows and in- flows of individuals as a share of the mean population between 1995 and 2010. The upper panel of the figure presents the intensity index, which reveals that the municipalities that lost a significant portion of their population to forced displacement were mainly located on the west and between the middle and the south of Colombia. The mean intensity index for all the municipalities is 0.11, suggesting that, on average, all Colombian municipalities lost approximately 11% of their popula- tion to forced migration between 1995 and 2010. The variance, however, is high (s.e. 0.25). There are also approximately 45 municipalities that saw an intensity index of 100 percent, suggesting that their population was depleted by forced migration. The figure also shows that the forced migration that originated in the center of the country was of relatively low intensity. The lower panel of the figure shows the pressure index, defined as the total number of forcefully displaced individuals who arrived in a municipality as a share of the municipal mean population between 1995 and 2010. The pressure index is a good proxy for the congestion of public goods and services as a consequence of forced displacement in the municipalities where these populations ar- rive. The pressure index has a mean value of 10 percent (s.e. 0.19). Although forced migration was substantial during the period of analysis, the figure reveals that most of the receiving municipalities did not face pressure indexes that are extremely large. In fact, only 38 municipalities have pres- sure indexes of more than 50 percent, suggesting that forcefully displaced migrants tend to move to urban areas with large populations. Similar results were documented by Carrillo (2009) and a, the Colom- Lozano-Gracia et al. (2010). For instance, according to the data from RUV, Bogot´ bian capital city, received almost 10 percent of the total number of migrants. Migrants may decide to move to urban areas in search of better economic opportunities (i.e., a larger labor market), a sense of safety due to anonymity, and to move away from conflict areas, which were predominantly rural during the period of analysis. 12 IV Empirical Strategy Identifying the effects of IDP inflows on firms’ behavior is challenging, since migrants do not move to random locations. The displaced populations, for instance, may have chosen to move to more populated, less violent areas where local authorities have better control of the territory, where there are better economic opportunities, and where illegal armed groups are not active. To correct for these biases, we use a panel-instrumental variable approach. Our main specification is given by the following equations: Log (Yjmt ) = γ1 IDPmt + Xmt Γ + γt + γj + jmt (1) IDPmt = θ1 Predicted Inflowsmt + Xmt Θ + θt + θj + µjmt (2) where j stands for firm, m stands for municipality, and t stands for year; Y represents the firm decisions (including production, prices, and input demands);11 IDPmt represents the ratio of IDP inflows to population of working age (multiplied by 100 to ease interpretation);12 Xmt is a vector of municipal controls;13 γt , γj , θt , and θj represent year and firm fixed effects; and 11 For production we use the variables of gross production (which includes the value of all products produced by the firm valued at factory-gate prices plus inventories in production process), intermediate consumption (which includes the value of all inputs used in the production process), energy consumption (measured in kilowatts), and probability of firm entry and exit; for prices we use the variables of nominal input and output prices (where firms’ prices are defined as the plant-product-year observation estimated by dividing the value of revenues or expenditures by physical quantities); and input demands include employment (total, blue- and white-collar), gross investment (measured as the sum of new and used assets minus the sales of assets) and net investment (which excludes depreciation). 12 Working age population includes individuals 12 years and older as defined by the Colombian statistics agency. 13 The variables in this vector include interactions of year dummy fixed effects and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995; department and year fixed effects; homicide rates; clashes between armed groups; and armed actions. 13 Predicted Inflowsmt is the instrumental variable, defined as Migrants1993 mk Predicted Inflowsmt = Forced Migration Outflowskt × × 100 (3) k=1 K Total Migrants1993 m where Forced Migration Outflowskt measures the number of individuals who were displaced by violence in municipality k and year t; K represents the total group of municipalities; Total Migrantsm is the total number of individuals who live in municipality m, but who were not born there in 1993; and Migrantsmk are the total number of individuals born in municipality k who are living in municipality m in 1993. We use the year 1993 to construct the instrument because in that year the Colombian statistics agency collected the last population census before the large wave of forced displacement took place in Colombia (see Figure I). Predicted Inflowsmt , therefore, is constructed following the original idea by Card (2001) and Altonji and Card (1991) (see Lewis and Peri, 2015 for a literature review), which exploits the fact that individuals tend to migrate disproportionately into regions in which they have relatives, friends, or family (commonly known in the migration literature as early settlements of migrants) because these people provide the mi- grants with support networks. In this specification, γ1 will identify the percentage change in firm outcomes when the inflows of displaced individuals increases by 1 percentage point. IV.1 Validity of the identification strategy The first condition that must be met to guarantee the validity of our estimates is the relevance assumption. It requires that our instrument (Predicted Inflowmt ) should be strongly correlated with the inflows of internally displaced populations (IDPmt ). Figure V shows the geographic distribution of inflows of forcefully displaced individuals and the predicted inflows constructed using equation (3) for the years 1995, 2000, 2005, and 2010. The figure suggests that there is a positive and strong correlation between both variables. A formal test is presented in Tables I, II, III, VII, and IV which show the estimates of the first stage equation confirming a positive and 14 significant correlation between Predicted Inflowsmt and IDPmt . The tables also show that the F- statistic for excluded instruments is always higher than 10, alleviating concerns of biases induced by a weak instrument. The second condition that must be met to guarantee that our estimates are valid is the ex- clusion restriction. It implies that the interaction of the aggregate time component of our instru- ment (forced migration outflows) and the geographic municipal component (earlier settlements of migrants), should only be correlated with firm outcomes through IDP inflows. Given that our estimates include fixed effects by year and firm (or municipality) aggregate time components or time-invariant firm characteristics are not a threat for our identification strategy. Our estimates will only be threatened by time-variable covariates (not controlled for in Xmt ) that may be correlated with the instrument and directly affect firm outcomes. One relevant threat to our identification strategy is that since forced migrants are fleeing vio- lence and conflict, they may be moving to areas with presumably lower levels of conflict. It is also possible that, upon arrival in their new locations, displaced individuals may be increasing local violence levels or eroding the rule of law either by becoming perpetrators or victims of violence. In that sense, higher violence and conflict could also affect firm performance. In fact, Camacho and Rodriguez (2013) and Rozo (2018) show that more violence increases firm exit, using the same data we employ in our analysis. To account for this possibility we control in all our estimates by homicide rates (as a proxy for violent crime) and conflict variables (including clashes between armed groups and armed actions).14 Our results are robust to the inclusion of these controls. Another possible threat to our identification is that municipalities that had larger earlier settle- ments of migrants also had different prevalent characteristics relative to the other municipalities before the conflict intensified, and these differences may be inducing divergent time patterns which are not explained by IDP inflows. It is possible, for instance, that municipalities with larger pre- 14 Data on homicide rates come from the Observatory of Human Rights from the Colombian Vice Presidency. Data on the Colombian civil conflict was obtained from the Centro de Memoria Hist´ orica (CMH). As of today, this is the best source of data on civil conflict as it combines information from NGOs and governmental records with testimonies from the victims of the civil conflict. We used data from CMH to construct two measures for the intensity of the Colombian civil conflict: total number of clashes between armed groups and total number of armed actions or attacks which took place between armed groups (i.e., illegal armed groups or the Colombian military). 15 settlements of migrants were more prosperous, less violent, had a different sector composition, or had larger government presence before the conflict induced the large forced migration wave. We account for all of these possibilities by including controls for interactions of year dummy fixed effects and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. We also include controls for department and year fixed effects to account for any regional time trends affecting firm behavior. All our estimates are robust to the inclusion of these controls.15 V Effects of IDP on Formal Firm’s Decisions We estimate equations (1) and (2) for firm production, prices, and input demands using the Annual Manufacturing Survey. The survey only includes information on formal firms that comply with government regulations. The results presented in this section, consequently, describe the effects of IDP inflows only on the Colombian formal sector. We expand our analysis to the effects of IDP in the informal sector in the next section. V.1 Effects on production First, we explore the impacts of IDP inflows on the intensive and extensive margin of production. The estimates of equations (1) and (2) using several proxies for the intensive margin of firm pro- duction are presented in Table I.16 We find negative effects of IDP inflows on all proxies of firm production. Our most preferred estimates are presented in panel C column (3) and include fixed 15 We also re-estimate all of our specifications scaling IDP inflows and our measure of Predicted Inflows by to- tal population instead of working age population. The results of these exercises are extremely similar to our main estimates. We do not include them in the manuscript due to space concerns, but they are available upon request. 16 The combination of controls presented in the columns that are multiples of three are our preferred estimates as they seem to pose the stronger test in the validity of our estimates. 16 effects by firm, year, and controls for differential time pre-trends.17 They suggest that when IDP inflows in a municipality increase by 1 percent, gross production of the firms located in that mu- nicipality decreases by approximately 0.3 percent. Table I also presents estimates of the effects of IDP inflows on oil and energy consumption pointing to similar results although slightly larger in size. We also test for dynamic effects of IDP inflows on the intensive margin of production and find no evidence of a long-term impact of IDP inflows one or two years after the IDP shock takes place. In fact, the coefficients become very small and close to zero as we lag the IDP shocks one or two years. When we estimate a model combining the contemporaneous effect and the effects one and two years after the IDP shock took place, we only observe significant effects for the contemporaneous impacts of IDP inflows on the intensive margin of production (see Appendix III). Table II presents the estimates of the effects of IDP inflows on the extensive margin of pro- duction estimated as the probability of entry and exit for each firm. The results confirm that firms are adjusting their external margin of production in response to larger IDP inflows. In fact, we observe that larger IDP inflows are associated with a larger exit and lower entry probability for all firms. Particularly, the estimates in columns (3) and (6) suggest that when the share of IDP as a percentage of the working age population increases in 1% the probability of firm entry falls by 0.01 percentage points and the probability of firm exit increases by 0.02 percentage points. The estimated coefficients represent a small effect taking into account the size of the migration shock (which corresponds to an average increment of 4,000 individuals in each municipality and year) and previous results from other studies that use the same data in Colombia.18 The literature that studies the impacts of immigration on firm performance generally finds 17 The number of observations drops for column (3) because all the additional controls are not available for all the municipalities in the manufacturing survey. However, these controls are available for the largest municipalities that host the bulk of the economic activity in Colombia. The same reduction in our sample occurs for the estimates presented in Tables II and III. 18 For instance, Camacho and Rodriguez (2013) uses the same sample to study the effect of conflict on firm exit. The authors find that a one-standard deviation increase in the number of guerrilla and paramilitary attacks in a municipality increases the probability of plant exit in 5.5 percentage points or .28 SD. 17 positive effects (see Ottaviano et al., 2018 and Kerr et al., 2015 for examples). However, most of these studies focus on the impacts of highly educated migrants in developed countries with small informal sectors. The effects of migration on firm outcomes are also sizable in these studies. Ottaviano et al. (2018), for instance, find that a one percentage point increase in the share of immigrant workers in the local labor force produced a 2.8 to 3.4% increase in firm exports and a 3% effect on labor productivity. The sign difference between their results and ours may be explained by the type of migrants we are analyzing. Particularly, highly educated individuals are typically absorbed by the formal sector whereas IDP are predominantly absorbed by the informal sector. Overall, our estimates suggest that larger IDP inflows are detrimental for the intensive and the extensive margins of production of formal firms in Colombia. We explore the potential mecha- nisms driving these effects in the next section. V.2 Effects on prices We next explore the effects of IDP inflows on firm prices. Considering that IDP are a population shock, it is plausible as was pointed earlier that they may be impacting output or input prices directly through an increase in consumption; that is, through an aggregate demand shock in output and input markets. We consider this possibility by estimating equations (1) and (2) using the average sale or purchase value of all outputs and inputs used or produced by all formal firms in the manufacturing census between 1995 and 2010. All estimates include firm and year fixed effects as well as product fixed effects, which correspond to the four-digit classification of the International Standard Industry Classification (113 four-digit codes). The results of this exercise are presented in Table III. Although some of the results are statistically significant for nominal output prices, these effects are not consistent and disappear once a more robust set of controls is included in column (3). Consequently, our results suggest that IDP inflows have not had a significant impact on input or output prices. 18 V.3 Effects on input demands Next, we explore whether IDP inflows had any impact on the firm’s demand for inputs including labor and capital. The results of these exercise are useful to study the possibility that IDP may be increasing labor supply and as such may be modifying firms’ optimal combination between labor and capital. Our estimates of the impacts of IDP in total employment are presented in Table IV. Addition- ally, we also estimate the effects of IDP inflows on employment and wages by type (blue- and white-collar) in Tables V and VI, albeit only between 2000 and 2010 due to data availability. We find no consistent evidence of a significant effect of higher IDP inflows on formal em- ployment or wages of any kind. Although some of the results are statistically significant for em- ployment effects, these effects are not consistent and disappear once a different set of controls are included in the estimates. Our results are in line with the argument suggesting that IDP in Colombia have low levels of education and little experience on manufacturing jobs (see Ib´ an˜ ez and Moya, 2006). In fact, as forced migrants in Colombia mostly moved from rural to urban areas, they previously worked on agricultural activities which have low demand on urban regions. Dis- placed individuals, hence, may be searching for low quality jobs as they lack experience, which ultimately may suggest that they will end up joining the informal sector. We also explore the effects of IDP inflows on firms’ capital demand measured as the logarithm of gross and net investment in Table VII. Similar to the estimates for employment, we are not able to distinguish a consistent statistically significant effect of IDP inflows in capital demand.19 In sum, our results so far suggest that the effects of IDP on manufacturing formal firms are limited to their negative effects on the intensive and extensive margin of production with no other observed effects on the prices or input demands in the formal sector. 19 As is common in most firm surveys a handful of firms report capital and investment figures, this explains the drop in the number of observations on Table VII. To address the concern that this sample was not selected we imputed capital figures to all the firms available in Table I by doing a two-step exercise: i) we estimate a regression of firm investment on production, intermediate consumption, and sector for the firms that have information available on all variables, ii) we use the estimated coefficients to predict investment figures for the firms that had missing information. We then re-estimate Table VII for all the firms included in Table I. We still find no effects of IDP migration on net investment. 19 VI Mechanisms IDP inflows may affect formal firm production via several channels. First, large IDP inflows can prompt a positive supply shock, which could reduce wages (as documented by Card, 2001 and Borjas, 2003 for economically-driven migration in developed countries) or cause input substitution (as Lewis, 2011 has found for the United States), and through these mechanisms, lower the costs of production or increase productivity which could rise levels of production and firm entry. We test this mechanism in subsection V.3 by analyzing the impacts of IDP inflows on labor demand and nominal wages finding no supporting evidence for its validity within the formal sector. Second, large IDP inflows can also increase local demand for products as a result of the larger local population. We argue however, that this channel is unlikely as our results point to the opposite direction and because IDP may be consuming proportionally more goods and services from the informal sector where prices are typically lower. In fact, La Porta and Shleifer (2014b) and Perry (2007) show that the informal sector is able to carry lower prices as it does not have to comply with regulations and legal fees, it also many times produces goods of lower quality relative to the ones produced in the formal sector. Third, forcefully displaced individuals themselves may come up with new ideas or create new businesses or even increase the productivity of firms that employ them with their know-how. We argue that the size of this last channel, however, is small, since most forcefully displaced indi- viduals only had labor experience in agriculture, had lower levels of education relative to local populations, and had few to no assets (see V´ an elez, 2002; Ib´ ˜ ez and Moya, 2006;Ib´ an˜ ez and Moya, 2006; Garay, 2008; Carrillo, 2009 for details). Fourth, IDP inflows may be fully absorbed by the informal sector increasing the informality market share, as informal businesses are able to offer lower prices than the formal sector due to lower regulations, taxes, or quality. Our results are in line with the last channel and are also consistent with the characteristics of the IDP in Colombia. Considering that displaced individuals arrive in new locations without legal 20 identification documents, have low experience in urban jobs that have high local demand, and tend to have lower education relative to the rest of the Colombian population (Ib´ an˜ ez and Moya, 2006), it is plausible that upon arrival at urban centers, IDP take lower tier jobs, which are most likely to be available in the informal sector (Amaral and Quintin, 2006; Perry, 2007; Galiani and Weinschelbaum, 2012; La Porta and Shleifer, 2014a; Meghir et al., 2015), increasing its size. A larger informal sector may end up hurting formal firms’ performance in several ways. Formal firms, for example, face higher costs relative to informal firms as they pay taxes, fees, and higher wages for their employees (La Porta and Shleifer, 2014a). Informal firms, thus, can usually carry lower prices and costs competing in an unfair way with formal businesses. Beyond reducing the demand for formal firms’ products via lower prices, unfair competition from informal firms could slow down the process in which inefficient firms can be replaced by more efficient competitors and negatively affect the incentives of formal firms to innovate and adopt new technologies (Perry, 2007). Competition with informal firms, consequently, may lead to productivity losses for the formal firms. In addition, since informal firms are able to use public goods but do not pay for them, this may lower the quality public goods and services and crowd out their use by formal firms (Besley and Persson, 2013). At the same time, higher informal competition may force formal sector firms to lower the quality of their products (as proposed by Banerji and Jain, 2007). If our argument is correct, then we should observe that those formal firms operating in sectors that are more prone to informality competition should be disproportionately affected by IDP in- flows relative to the other firms. To test this hypothesis we use the manufacturing survey firm data and split the sample according to industry codes in two groups of firms according to whether they are part of a sector that tends to have high levels of informality (e.g., retail sales and construction) or low levels of informality (e.g., financial services, machinery production, production of chemical or pharmaceutical products). The results are presented in Tables VIII through XI, and strongly support our hypothesis. Particularly, we only find negative effects of IDP inflows on the intensive and extensive margins of production for the group of firms that operate in sectors that are more 21 prone to face a larger informal competition.20 Further evidence supporting our argument can be established when dividing our sample be- tween the firms that never exit in our period of analysis (which are part of the balanced panel) and the firms who exit at some point. Our main estimates use the unbalanced panel of firms (all firms), as such, the results observed in Tables I and II correspond to the firms that are in the market each period of analysis (but may exit in the future). The size, in terms of production, of the firms who exit anytime between 1995 and 2010 is smaller relative to the firms that stay during the whole pe- riod of analysis. In particular, the mean production of the firms in the balanced panel is 20,300,000 (6,376 observations) and the mean production of the firms who exit sometime between 1995 and 2010 is 10,900,000 (115,885 observations). To clarify how the sample composition is affecting our main results we estimate the effects of IDP migration for the sample of firms in the balanced panel (i.e., the largest firms) in Appendix IV. We find no effects of IDP migration on any of the outcomes that we study. These results imply that the negative effects that we observe on Tables I and II are driven by the smaller firms, which ultimately exit the market at some point during our period of analysis. These results also support our argument as smaller firms should be disproportionately affected by the competition of an enlarged informal sector relative to larger firms. VI.1 Do other sources of data support our hypothesis? We also use the Colombian household surveys to test how higher IDP inflows affect the size in- formal sector. Particularly, we examine whether larger IDP inflows are indeed affecting the size of the informal sector by classifying all workers age 15 and 62 into two groups according to whether their primary job is part of a sectors that tends to have high levels of informality (such as retail sales and construction) or low levels of informality (such as financial services, machinery pro- duction, production of chemical or pharmaceutical products, and highly technical jobs).21 Sector classification codes are only available in the labor force surveys beginning in 2002. Our sample, 20 The only exception is firm exit for which we find significant effects for both groups of firms. 21 We use the same classification employed when we divide the manufacturing survey into firms more and less exposed to informal competition for consistency. 22 consequently, spans between 2002 and 2010. Data from 2002 to 2005 comes from the Encuesta Continua de Hogares and the Gran Encuesta Integrada de Hogares from 2006 to 2010. Both sur- veys are comparable across time, but the later introduced improvements such as new questions and a sharp increase in the number of municipalities surveyed. From 2002 to 2005 we can only identify the exact location of workers located in the 13 main cities of Colombia (i.e., 13 municipalities), but beginning in 2006 we observe their locations in 609 municipalities. We estimate a linear probability model for the probability of being employed in these sectors on IDP inflows including fixed effects by year, municipality, month (when the survey was collected), individuals covariates (such as gender, marital status, education level, and household size), and municipal controls (including all the controls included in our previous estimates). The results are presented in Table XII and largely suggest that the probability of begin employed in a highly informal sector increases with higher IDP inflows. Additionally, the effects of IDP inflows on the employment probability within highly formal sectors are not significant and even has a negative sign. Particularly, our estimates in Panel C and column 2 suggest that when the share of IDP increases in 1 percent, the probability of being employed in a highly informal sector increases by 0.2%. Considering that the labor force surveys are only available between 2002 and 2010 and that they only include data for a handful of municipalities, we also re-estimate all of our previous analysis restricting the sample in the manufacturing surveys in two ways. First, by restricting the sample period (to either 1995-2010 and 2002-2010) and second, by restricting the municipalities to only include those available in the manufacturing census and the labor force surveys. Our results are presented in Appendix II and are consistent with our previous results. VII Discussion This paper investigates the effects of inflows of IDP on the behavior of firms in an economy seg- mented into a normal-sized formal sector and a large informal sector as is commonly observed in 23 most developing countries. Our findings strongly suggest that larger inflows of displaced individu- als who are fully absorbed by the informal sector have sizable negative effects on the performance of formal firms. Our estimates suggest, more particularly, that when inflows of IDP increase by 1 percent, the production of formal firms drops approximately by 0.3 percent and the probability of firm exit increases by 0.02 percentage points. We argue that the effects of inflows of IDP on formal firms are mainly driven by the positive association of larger inflows of displaced individuals and the size of the informal sector. As forcefully displaced individuals tend to have low education levels and a lack of experience in occupations that have high local demand (because many IDP move from rural to urban areas), they may be taking low-tier jobs that are most likely offered in the informal sector, thus increasing its size. We also find that the effects of IDP migration on formal firms are primarily concentrated in the short-term, suggesting that developing economies adjust rapidly to large labor shocks. One interesting question for future research, is thus, to examine the distribution consequences of these type of shocks in the short-run. Our results highlight the importance of national policies and international cooperation efforts in facilitating the integration of IDP—and also international refugees—into formal labor markets. They suggest that in contexts where IDP or refugees are not allowed to work formally, their par- ticipation in the informal economy could negatively affect firms operating in the formal sector. In addition, our results highlight the importance of the initial conditions in shaping the effects of IDP and refugees, as they could exacerbate the negative effects of informality on the host economy. Despite the unique data employed for this study, our analysis is limited to the effects of IDP on the formal sector because there is no similar data available for firms operating in the informal sector. Empirically exploring the effects of IDP on informal firms remains an important area for future research. One important limitation of our study is that it only focuses on the partial equilibrium effects of internal forced displacement on firm performance. A promising line for the future is to explore the 24 general equilibrium effects of internal forced displacement in the Colombian economy. As found by Calder´ on-Mej´ ıa and Ib´ an˜ ez (2015), for instance, other parts of the economy are also affected by these migration shocks. Particularly, the authors document that as a consequence of internal forced displacement native workers in the informal sector receive lower wages. Additionally, the positive population shock has also been shown to be inflating housing prices for low-income individuals and reducing them for high-income individuals (see Depetris-Chauvin and Santos, 2018). These another potential effects may end up interacting in interesting ways worth exploring with more detail. References Abadie, A. and J. Gardeazabal (2003). The economic costs of conflict: A case study of the basque country. American economic review 93(1), 113–132. Accetturo, A., M. Bugamelli, and A. R. Lamorgese (2012). Welcome to the machine: firms’ reaction to low-skilled immigration. Bank of Italy Temi di Discussione (Working Paper) No 846. Altindag, O., O. Bakis, and S. Rozo (2018). Blessing or burden? The impact of refugees on businesses and the informal economy. SSRN. Working Paper N. 3188406. Altonji, J. G. and D. Card (1991). The effects of immigration on the labor market outcomes of less-skilled natives. In Immigration, Trade, and the Labor Market, pp. 201–234. University of Chicago Press. Amaral, P. S. and E. Quintin (2006). A competitive model of the informal sector. Journal of Monetary Economics 53(7), 1541–1553. Banerji, A. and S. Jain (2007). Quality dualism. Journal of Development Economics 84(1), 234– 250. 25 Besley, T. and T. Persson (2013). Taxation and development. In Handbook of Public Economics, Volume 5, pp. 51–110. Blattman, C. and E. Miguel (2010). Civil war. Journal of Economic literature 48(1), 3–57. Borjas, G. J. (2003). The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. The Quarterly Journal of Economics 118(4), 1335–1374. Borjas, G. J. and J. Monras (2017). The labour market consequences of refugee supply shocks. Economic Policy 32(91), 361–413. uck, T., W. Naud´ Br¨ e, and P. Verwimp (2013). Business under fire: entrepreneurship and violent conflict in developing countries. Journal of Conflict Resolution 57 (1), 3–19. Calder´ on-Mej´ ıa, V. and A. M. Ib´ an˜ ez (2015). Labour market effects of migration-related supply shocks: evidence from internal refugees in colombia. Journal of Economic Geography 16(3), 695–713. Camacho, A. and C. Rodriguez (2013). Firm exit and armed conflict in Colombia. Journal of Conflict Resolution 57 (1), 89–116. Card, D. (2001). Immigrant inows, native outows, and the local labor market impacts of higher immigration. Journal of Labor Economics 19(1), 22–64. Carrillo, A. C. (2009). Internal displacement in Colombia: humanitarian, economic and so- cial consequences in urban settings and current challenges. International Review of the Red Cross 91(875), 527–546. Carrizosa, M. T. and A. S. Blasco (2009). Immigration and firm performance: a city-level ap- on y comportamiento empresarial: una aproximaci´ proach/inmigraci´ on a escala de ciudad. In- vestigaciones Regionales (15), 111. 26 Ceritoglu, E., H. B. G. Yunculer, H. Torun, and S. Tumen (2017). The impact of Syrian refugees on natives’ labor market outcomes in Turkey: evidence from a quasi-experimental design. IZA Journal of Labor Policy 6(1), 5. Chiswick, B. R. (1999). Are immigrants favorably self-selected? American Economic Re- view 89(2), 181–185. Clemens, M. A. and J. Hunt (2017). The labor market effects of refugee waves: Reconciling conflicting results. NBER Working Paper N. 23433. Del Carpio, X. V. and M. C. Wagner (2015). The impact of Syrians refugees on the Turkish labor market. World Bank Policy Research Working Paper (7402). Depetris-Chauvin, E. and R. J. Santos (2018). Unexpected guests: The impact of internal displace- ment inflows on rental prices in colombian host cities. Journal of Development Economics 134, 289–309. Dustmann, C. and A. Glitz (2015). How do industries and firms respond to changes in local labor supply? Journal of Labor Economics 33(3 Part 1), 711–750. El Espectador (2009, April). La situaci´ on de los desplazados. http://www.elespectador.com/articulo137806-situacion-de-los-desplazados. El Tiempo (2010, August). Fuerza laboral de desplazados impacta salarios informales. http://www.eltiempo.com/archivo/documento/MAM-4088271. El Tiempo (2014, February). Desempleo entre los desplazados es del 35,5%. http://www.eltiempo.com/archivo/documento/CMS-13528618. Engel, S. and A. M. Ib´ an˜ ez (2007). Displacement Due to Violence in Colombia: A Household- Level Analysis. Economic Development and Cultural Change 55(2), 335–365. Galiani, S. and F. Weinschelbaum (2012). Modeling informality formally: households and firms. Economic Inquiry 50(3), 821–838. 27 on de los derechos de la poblaci´ Garay, L. J. (2008). Proceso nacional de verificaci´ on desplazada. First Report to the Colombian Constitutional Court. Guidolin, M. and E. La Ferrara (2007). Diamonds are forever, wars are not: Is conflict bad for private firms? American Economic Review 97 (5), 1978–1993. Huffington Post (2010). Colombia’s Internally Displaced People. Ib´ an˜ ez, A. M. and A. Moya (2006). C´ omo el desplazamiento forzado deteriora el bienestar de los hogares desplazados? an´ on. alisis y determinantes del bienestar en los municipios de recepci´ Ib´ an˜ ez, A. M. and A. Moya (2010). Vulnerability of victims of civil conflicts: empirical evidence for the displaced population in Colombia. World Development 38(4), 647–663. Ib´ an˜ ez, A. M. and C. E. V´ elez (2008). Civil conflict and forced migration: The micro determinants and welfare losses of displacement in Colombia. World Development 36(4), 659–676. Ib´ an˜ ez, A. M., A. Moya, and A. Vel´ ıtica proactiva para la poblaci´ asquez (2006). Hacia una pol´ on desplazada. Bogot´ a: Universidad de los Andes, Secretariado Nacional de Pastoral Social Cari- tas Colombia. IPS (1999, September). Colombia: ıa informal de Desplazados engrosan econom´ a. Bogot´ http://www.ipsnoticias.net/1999/09/repeticion-ciudades-de-america-latina-colombia- desplazados-engrosan-economia-informal-de-bogota/. Jaeger, D. A., J. Ruist, and J. Stuhler (2018). Shift-share instruments and the impact of immigra- tion. NBER Working Paper N. 24285. Kerr, S. P., W. R. Kerr, and W. F. Lincoln (2015). Skilled immigration and the employment struc- tures of US firms. Journal of Labor Economics 33, S147–S186. Kugler, M. and E. Verhoogen (2011). Prices, plant size, and product quality. The Review of Economic Studies 79(1), 307–339. 28 La Porta, R. and A. Shleifer (2014a). Informality and development. The Journal of Economic Perspectives 28(3), 109–126. La Porta, R. and A. Shleifer (2014b). Informality and development. Journal of Economic Perspec- tives 28(3), 109–26. ıa (2012, December). Desplazados dentro de su misma ciudad: las v´ La Silla Vac´ as in- ıctimas m´ visibles en Colombia. http://lasillavacia.com/historia/desplazados-dentro-de-su-misma-ciudad- las-victimas-mas-invisibles-en-colombia-40579. Lewis, E. (2011). Immigration, skill mix, and capital skill complementarity. The Quarterly Journal of Economics 126(2), 1029–1069. Lewis, E. and G. Peri (2015). Immigration and the economy of cities and regions. In Handbook of Regional and Urban Economics, Volume 5, pp. 625–685. Lozano-Gracia, N., G. Piras, A. M. Ib´ an˜ ez, and G. J. Hewings (2010). The journey to safety: conflict-driven migration flows in Colombia. International Regional Science Review 33(2), 157– 180. Meghir, C., R. Narita, and J.-M. Robin (2015). Wages and informality in developing countries. American Economic Review 105(4), 1509–1546. Moya, A. et al. (2012). Violence, emotional distress and induced changes in risk aversion among the displaced population in Colombia. In Pacific Development Economics Conference, Univer- sity of California, Davis. Ottaviano, G. I., G. Peri, and G. C. Wright (2018). Immigration, trade and productivity in services: Evidence from UK firms. Journal of International Economics 112, 88–108. Perry, G. (2007). Informality: Exit and exclusion. World Bank Publications. Rozo, S. V. (2018). Is murder bad for business? Evidence from Colombia. Review of Economics and Statistics. Forthcoming. 29 Serneels, P. and M. Verpoorten (2015). The impact of armed conflict on economic performance: Evidence from rwanda. Journal of Conflict Resolution 59(4), 555–592. UNCHR (2015). Global trends: Forced displacement in 2015. Technical report. UNHCR (2018). Global Trends. Forced Displacement in 2017. Technical report. on y Reparaci´ Unidad para la Atenci´ ıctimas (2013). Informe nacional de desplazamento on de V´ forzado en colombia 1985 a 2012. Technical report. elez, C. E. (2002). Colombia poverty report. World Bank. V´ World Bank (2016). Forcibly displaced: Toward a development approach supporting refugees, the internally displaced and their hosts. Technical report, World Bank Report. 30 VIII Tables and Figures 31 Table (I) Impacts of IDP Inflows in the Formal Intensive Margin of Production Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A. Reduced Form Predicted Inflows -0.010** -0.021*** -0.0111** -0.013*** -0.020*** -0.0151*** -0.019*** -0.021** -0.0194*** (0.0042) (0.006) (0.00476) (0.005) (0.007) (0.00495) (0.005) (0.008) (0.00575) R-squared 0.934 0.935 0.951 0.924 0.924 0.942 0.919 0.920 0.939 Observations 122,261 122,231 82,715 122,222 122,192 82,704 122,104 122,074 82,572 Panel B. OLS Share of IDP (% Working Age Pop) -0.011** -0.010 -0.003 -0.014*** -0.007 -0.004 0.004 -0.001 -0.002 (0.004) (0.006) (0.005) (0.005) (0.007) (0.006) (0.007) (0.012) (0.006) R-squared 0.934 0.935 0.951 0.924 0.924 0.942 0.919 0.919 0.939 Observations 122,267 122,237 82,719 122,228 122,198 82,708 122,110 122,080 82,576 Panel C. 2SLS Share of IDP (% Working Age Pop) -0.027** -0.068*** -0.029** -0.034*** -0.067*** -0.040*** -0.050*** -0.069** -0.051*** (0.011) (0.020) (0.012) (0.012) (0.024) (0.013) (0.013) (0.028) (0.015) Observations 122,261 122,231 82,715 122,222 122,192 82,704 122,104 122,074 82,572 32 Panel D. First Stage Dep. Variable: Share of IDPs (% Working Age Pop) Predicted Inflows 0.374*** 0.306*** 0.328*** 0.374*** 0.306*** 0.328*** 0.374*** 0.306*** 0.328*** (0.015) (0.019) (0.025) (0.015) (0.019) (0.025) (0.015) (0.019) (0.025) First Stage F-statistic 627.24 257.18 446.21 627.24 257.18 446.21 627.24 257.18 446.21 Observations 122,269 122,239 82,719 122,269 122,239 82,719 122,269 122,239 82,719 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No No Yes No Homicides Rates No Yes No No Yes No No Yes No Conflict Controls No Yes No No Yes No No Yes No Additional Controls No No Yes No No Yes No No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by firm are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (II) Impacts of IDP Inflows in the Formal Extensive Margin of Production Dependent Variables (in logs) Entry Exit (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.00912*** -0.0111*** -0.00399** 0.00961*** 0.0123*** 0.00964*** (0.00131) (0.00197) (0.00197) (0.00133) (0.00202) (0.00216) R-squared 0.602 0.611 0.608 0.225 0.232 0.229 Observations 122,211 122,181 82,572 122,211 122,181 82,572 Panel B. OLS Share of IDP (%Pop in Working Age) -0.00865*** -0.00359** -0.00312** 0.00319*** 0.00212 0.00213 (0.00111) (0.00161) (0.00130) (0.00120) (0.00180) (0.00150) R-squared 0.602 0.610 0.608 0.224 0.232 0.229 Observations 122,217 122,187 82,572 122,217 122,187 82,572 Panel C. 2SLS Share of IDPs (%Pop in Working Age) -0.0244*** -0.0362*** -0.00845** 0.0257*** 0.0402*** 0.0204*** (0.00361) (0.00708) (0.00417) (0.00377) (0.00766) (0.00472) 33 Observations 122,211 122,181 82,572 122,211 122,181 82,572 Panel D. First Stage Dep. Variable: Share of IDP (% Pop in Working Age) Predicted Inflows 0.374*** 0.306*** 0.473*** 0.374*** 0.306*** 0.473*** (0.0150) (0.0191) (0.0233) (0.0150) (0.0191) (0.0233) First Stage F-statistic 626.99 257.17 412.22 626.99 257.17 412.22 Observations 122,211 122,181 82,572 122,211 122,181 82,572 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No Homicides Rates No Yes No No Yes No Conflict Controls No Yes No No Yes No Additional Controls No No Yes No No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by municipality are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (III) Impacts of IDP Inflows in Input and Output Nominal Prices Dependent Variables (in logs) Nominal Output Price Nominal Input Price (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.006* -0.012** -0.002 0.005 0.004 0.008 (0.004) (0.006) (0.004) (0.005) (0.009) (0.006) R-squared 0.395 0.406 0.445 0.944 0.944 0.943 Observations 799,738 799,737 391,289 731,502 731,503 731,255 Panel B. OLS Share of IDP (% Working Age Pop.) -0.002 -0.002 0.003 0.009 0.011 0.008 (0.003) (0.005) (0.005) (0.006) (0.007) (0.006) R-squared 0.395 0.406 0.445 0.944 0.944 0.945 Observations 799,738 799,737 391,289 731,502 731,503 731,255 Panel C. 2SLS Share of IDP (% Working Age Pop.) -0.020* -0.047** -0.008 0.014 0.013 0.020 (0.011) (0.023) (0.013) (0.016) (0.036) (0.016) 34 Observations 799,741 799,737 391,289 731,509 731,513 731,255 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop.) Predicted Inflows 0.314*** 0.247*** 0.293*** 0.315*** 0.252*** 0.392*** (0.017) (0.018) (0.016) (0.017) (0.019) (0.019) First Stage F-statistic 335.28 178.56 371.78 330.51 187.35 458.81 Observations 799,741 799,737 391,289 731,509 731,513 731,255 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Product FE Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No Conflict Controls No Yes No No Yes No Homicides Rates No Yes No No Yes No Additional Controls No No Yes No No Yes Notes: Each coefficient corresponds to a separate regression. Product fixed effects correspond to the four-digit classification of the International Standard Industry Classification, which account for 113 four-digit codes. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the firm level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (IV) Impacts of IDP Inflows in Labor Demand Dependent Variables (in logs) Total Employment Nominal Wages (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.010*** -0.016*** -0.001 0.001 0.001 -0.002 (0.0027) (0.004) (0.004) (0.0016) (0.002) (0.002) Observations 82,696 82,676 0.932 82,267 82,247 0.862 R-squared 0.931 0.932 82,657 0.861 0.862 82,228 Panel B. OLS Share of IDP (% Working Age Pop.) -0.001 -0.008* 0.0003 0.003 0.0003 0.002 (0.00344) (0.00492) (0.004) (0.00190) (0.00246) (0.002) Observations 0.931 0.932 0.932 0.861 0.862 0.862 R-squared 82,700 82,680 82,661 82,271 82,251 82,232 Panel C. 2SLS Share of IDP (% Working Age Pop.) -0.028*** -0.055*** -0.00394 0.002 0.003 -0.00637 (0.007) (0.016) (0.010) (0.005) (0.008) (0.006) 35 Observations 82,696 82,676 82,657 82,267 82,247 82,228 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop.) Predicted Inflows 0.353*** 0.288*** 0.328*** 0.353*** 0.288*** 0.328*** (0.013) (0.023) (0.025) (0.013) (0.023) (0.025) First Stage F-statistic 711.12 151.57 446.21 711.12 151.57 446.21 Observations 82,758 82,738 82,719 82,758 82,738 82,719 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No Homicides Rates No Yes No No Yes No Conflict Controls No Yes No No Yes No Additional Controls No No Yes No No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the firm level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (V) Impacts of IDP Inflows in Nominal Wages by Type (2000-2010) Dependent Variables (in logs) Nominal Wages Blue-collar Nominal Wages White-collar Nominal Wages (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows 0.001 -0.002 0.003 -0.002 0.001 -0.0002 (0.002) (0.002) (0.002) (0.002) (0.004) (0.003) R-squared 0.862 0.862 0.765 0.764 0.825 0.825 Observations 82,247 82,228 79,652 79,633 78,619 78,628 Panel B. OLS Share of IDP (% Working Age Pop.) 0.0003 0.002 0.001 0.002 0.002 0.006 (0.002) (0.002) (0.002) (0.002) (0.004) (0.004) R-squared 0.862 0.862 0.765 0.765 0.825 0.825 Observations 82,251 82,232 79,656 79,637 78,623 78,632 Panel C. 2SLS Share of IDP (% Working Age Pop.) 0.003 -0.006 0.009 -0.006 0.003 -0.0005 (0.008) (0.006) (0.008) (0.006) (0.013) (0.009) 36 Observations 82,247 82,228 79,652 79,633 78,619 78,628 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop) 0.288*** 0.328*** 0.288*** 0.328*** 0.288*** 0.328*** Predicted Inflows (0.023) (0.025) (0.023) (0.025) (0.023) (0.025) First Stage F-statistic 151.57 446.21 151.57 446.21 151.57 446.21 Observations 82,738 82,719 82,738 82,719 82,738 82,719 Controls (for all Panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year × Department FE Yes No Yes No Yes No Homicides Rates Yes No Yes No Yes No Conflict Controls Yes No Yes No Yes No Additional Controls No Yes No Yes No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the firm level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (VI) Impacts of IDP Inflows in Labor Demand by Type (2000-2010) Dependent Variables (in logs) Employment Blue-collar Employment White-collar Employment (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.016*** -0.001 -0.020*** -0.003 -0.018*** -0.004 (0.004) (0.004) (0.005) (0.004) (0.005) (0.005) R-squared 0.932 0.932 0.919 0.919 0.911 0.910 Observations 82,676 82,657 79,823 79,804 81,250 81,229 Panel B. OLS Share of IDP (% Working Age Pop.) -0.008* 0.0003 -0.006 0.001 -0.010* 0.0002 (0.005) (0.004) (0.005) (0.004) (0.006) (0.005) R-squared 0.932 0.932 0.919 0.919 0.911 0.910 Observations 82,680 82,661 79,827 79,808 81,254 81,233 Panel C. 2SLS Share of IDP (%Working Age Pop.) -0.055*** -0.004 -0.068*** -0.007 -0.063*** -0.010 (0.016) (0.010) (0.017) (0.011) (0.019) (0.013) 37 Observations 82,676 82,657 79,823 79,804 81,250 81,229 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop.) Predicted Inflows 0.288*** 0.328*** 0.288*** 0.328*** 0.288*** 0.328*** (0.023) (0.025) (0.023) (0.025) (0.023) (0.025) First Stage F-statistic 151.57 446.21 151.57 446.21 151.57 446.21 Observations 82,738 82,719 82,738 82,719 82,738 82,719 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year × Department FE Yes No Yes No Yes No Homicides Rates Yes No Yes No Yes No Conflict Controls Yes No Yes No Yes No Additional Controls No Yes No Yes No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the firm level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (VII) Impacts of IDP Inflows in Capital Demand Dependent Variables (in logs) Net Investment Gross Investment (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.053*** -0.070** -0.040 -0.027** -0.030* -0.008 (0.019) (0.027) (0.025) (0.013) (0.018) (0.017) Observations 23,680 23,655 23,669 48,725 48,687 48,709 R-squared 0.707 0.711 0.710 0.691 0.694 0.692 Panel B. OLS Share of IDP (% Working Age Pop.) 0.001 0.013 -0.004 -0.012 -0.014 -0.010 (0.0229) (0.0296) (0.026) (0.0143) (0.0191) (0.016) Observations 0.707 0.711 0.710 0.691 0.693 0.692 R-squared 23,680 23,655 23,669 48,725 48,687 48,709 Panel C. 2SLS Share of IDP (% Working Age Pop.) -0.146*** -0.240** -0.0942 -0.075** -0.107 -0.0197 (0.055) (0.104) (0.062) (0.036) (0.065) (0.044) 38 Observations 23,680 23,655 23,669 48,725 48,687 48,709 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop.) Predicted Inflows 0.353*** 0.288*** 0.328*** 0.353*** 0.288*** 0.328*** (0.013) (0.023) (0.025) (0.013) (0.023) (0.025) First Stage F-statistic 711.12 151.57 446.21 711.12 151.57 446.21 Observations 23,680 23,655 23,669 48,725 48,687 48,709 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No Homicides Rates No Yes No No Yes No Conflict Controls No Yes No No Yes No Additional Controls No No Yes No No Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the firm level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (VIII) Impacts of IDP inflows on Firms in the Informal and Formal Sectors - Manufacturing Sample Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption Type of sector Formal Informal Formal Informal Formal Informal (1) (2) (3) (4) (5) (6) Panel A. Reduced Form -0.0125 -0.0109* -0.0148 -0.0157*** -0.0162 -0.0214*** Predicted Inflows (0.00808) (0.00570) (0.00929) (0.00577) (0.0103) (0.00685) R-squared 0.946 0.954 0.931 0.948 0.937 0.940 Observations 34,395 46,572 34,390 46,566 34,342 46,487 Panel B. OLS Share of IDPs (%Pop. in Working Age) 0.00187 -0.00190 0.00228 -0.00562 0.00611 -0.00417 (0.00794) (0.00822) (0.00943) (0.00809) (0.00802) (0.00879) R-squared 0.946 0.954 0.931 0.948 0.937 0.940 Observations 34,395 46,572 34,390 46,566 34,342 46,487 Panel C. 2SLS Share of IDPs (%Pop in Working Age) -0.0397 -0.0271* -0.0471 -0.0389*** -0.0514 -0.0532*** 39 (0.0256) (0.0138) (0.0293) (0.0140) (0.0327) (0.0167) Observations 34,395 46,572 34,390 46,566 34,342 46,487 Panel D. First Stage Dep. Variable: Share of IDPs (% Working Age Pop.) Predicted Inflows 0.314*** 0.403*** 0.314*** 0.403*** 0.314*** 0.403*** (0.0225) (0.0245) (0.0225) (0.0245) (0.0225) (0.0245) First Stage F-statistic 195.38 270.84 195.4 270.8 194.61 270.14 Observations 34,395 46,572 34,390 46,566 34,342 46,487 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Yes Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by firm are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (IX) Impacts of IDP inflows on Firms in the Informal and Formal Sectors (cont’d) - Manufacturing Sample Dependent Variables (in logs) Entry Exit Formal Informal Formal Informal (1) (2) (3) (4) Panel A. Reduced Form Predicted IDP Inflows -0.00271 -0.00486** 0.0122*** 0.00878*** (0.00361) (0.00239) (0.00427) (0.00251) R-squared 0.531 0.671 0.264 0.240 Observations 47,273 71,656 47,273 71,656 Panel B. OLS Share of IDP (%Pop in Working Age) -0.00309 -0.00331* 0.00586** 0.00193 (0.00215) (0.00185) (0.00297) (0.00208) R-squared 0.531 0.671 0.264 0.240 Observations 47,273 71,656 47,273 71,656 Panel C. 2SLS Share of IDP (%Pop in Working Age) -0.00615 -0.0104** 0.0277*** 0.0187*** 40 (0.00819) (0.00510) (0.00976) (0.00563) Observations 0.531 0.671 0.262 0.239 Panel D. First Stage Dep Variable: Share of IDPs (% Pop in Working Age) Predicted Inflows 1993 0.441*** 0.468*** 0.441*** 0.468*** (0.0374) (0.0291) (0.0374) (0.0291) First Stage F-statistic 139.57 259.05 139.57 259.05 Observations 47,273 71,656 47,273 71,656 Controls for all Panels Year FE Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by firm are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (X) Impacts of IDP inflows on Firms in the Informal and Formal Sectors (cont’d) - Manufacturing Sample Dependent Variables (in logs) Nominal Input Price Nominal Output Price Formal Informal Formal Informal (1) (2) (3) (4) Panel A. Reduced Form 0.00314 -0.00133 0.0184 0.00748 Predicted IDP Inflows (0.00657) (0.00412) (0.0116) (0.00555) R-squared 0.447 0.506 0.939 0.952 Observations 146,234 236,513 305,008 413,198 Panel B. OLS Share of IDP (% Pop in Working Age) 0.0179* 0.00522 0.0224* 0.00578 (0.00945) (0.00531) (0.0125) (0.00622) R-squared 0.447 0.506 0.939 0.952 Observations 146,234 236,513 305,008 413,198 Panel C. 2SLS Share of IDPs (% Pop in Working Age) 0.0108 -0.00478 0.0475 0.0226 41 (0.0226) (0.0148) (0.0298) (0.0169) Observations 0.447 0.506 0.939 0.952 Panel D. First Stage Dep Variable: Share of IDPs (% Pop in Working Age) Predicted Inflows 1993 0.291*** 0.278*** 0.389*** 0.325*** (0.0169) (0.0177) (0.0257) (0.0200) First Stage F-statistic 296.44 246.09 313.34 255.36 Observations 146,234 236,513 346,081 441,206 Controls for all Panels Year FE Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by firm are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (XI) Impacts of IDP inflows on Firms in the Informal and Formal Sectors (cont’d) - Manufacturing Sample Dependent Variables (in logs) Net Investment Gross Investment Total Employment Nominal Wages Formal Informal Formal Informal Formal Informal Formal Informal (1) (2) (3) (4) (5) (6) (7) (8) Panel A. Reduced Form Predicted Inflows -0.0279 -0.0337 0.0265 -0.0173 -0.00226 -0.00128 -0.00325 -0.00122 (0.0446) (0.0324) (0.0309) (0.0220) (0.00557) (0.00500) (0.00450) (0.00240) Observations 0.706 0.715 0.670 0.706 0.927 0.936 0.853 0.870 R-squared 9,251 13,841 19,244 28,384 34,370 46,543 34,168 46,325 Panel B. OLS Share of IDP (%Pop in Working Age) -0.0241 -0.0311 0.0119 -0.0281 0.00508 0.00170 0.00408 0.00161 (0.0545) (0.0359) (0.0290) (0.0225) (0.00542) (0.00562) (0.00423) (0.00287) Observations 0.706 0.715 0.670 0.706 0.927 0.936 0.853 0.870 R-squared 9,251 13,841 19,244 28,384 34,370 46,543 34,168 46,325 Panel C. 2SLS Share of IDP (%Pop in Working Age) -0.107 -0.0717 0.0834 -0.0412 -0.00720 -0.00317 -0.0103 -0.00304 42 (0.172) (0.0706) (0.0973) (0.0524) (0.0178) (0.0124) (0.0143) (0.00594) Observations 9,251 13,841 19,244 28,384 34,370 46,543 34,168 46,325 Panel D. First Stage Dep Variable: Share of IDPs (% Pop in Working Age) Predicted Inflows 0.314*** 0.403*** 0.314*** 0.403*** 0.314*** 0.403*** 0.314*** 0.403*** (0.0225) (0.0245) (0.0225) (0.0245) (0.0225) (0.0245) (0.0225) (0.0245) First Stage F-statistic 86.25 44.17 156.48 131.79 189.1 269.77 188.79 268.9 Observations 9,251 13,841 19,244 28,384 34,398 46,543 34,398 46,325 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Yes Yes Yes Yes Notes: Each coefficient corresponds to a separate regression. Additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors by firm are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Table (XII) Impacts of IDP inflows in Employment in Informal and Formal Sectors Dependent Variable Pr (Employed) Sector All Sectors Highly Informal Sectors Highly Formal Sectors (1) (2) (3) Panel A. Reduced Form Predicted Inflows 0.001*** 0.002*** -0.00002 (0.0004) (0.0004) (0.00005) R-squared 0.328 0.398 0.017 Observations 2,797,614 2,323,262 474,347 Panel B. OLS Share of IDPs (% Working Age Pop.) 0.0003 0.0003 -0.00001 (0.0002) (0.0002) (0.00002) R-squared 0.328 0.398 0.017 Observations 2,797,614 2,323,262 474,347 Panel C. 2SLS 43 Share of IDPs (% Working Age Pop.) 0.001** 0.002** -0.00003 (0.0006) (0.0007) (0.00008) Observations 2,797,614 2,323,262 474,347 Panel C. First Stage Dep. Variable: Share of IDPs (% Working Age Pop.) Predicted Inflows 0.774*** 0.794*** 0.656*** (0.177) (0.179) (0.166) First Stage F-statistic 16.12 16.69 13.68 Observations 2,797,614 2,323,262 474,347 Controls for all Panels Year FE Yes Yes Yes Municipality FE Yes Yes Yes Month FE Yes Yes Yes Individuals Covariates Yes Yes Yes Additional Controls Yes Yes Yes Notes: Each coefficient corresponds to a separate regression. Individuals covariates include gender, marital status, education level and household size, and additional controls include interactions of year dummies and i) armed attacks by illegal armed groups in 1995, ii) fatal victims of attacks in 1995, iii) homicide rates 1995, iv) public expenditures in 1995, v) central government transfers to health, education, and other expenditures in 1995, vi) number of financial institutions in 1995, vii) number of tax collection offices in 1995, viii) 2000 GDP share in agriculture, services, and industry, ix) municipal tax income in 1995, and x) night light density in 1995. Clustered standard errors at the municipality level are reported in parentheses. ∗∗∗ significant at the 1%, ∗∗ significant at the 5%, ∗ significant at the 10%. Figure (I) Inflows of IDP in Colombia (1995-2010) (a) Total Number of Individuals (b) Total Number of Hosting Municipalities 44 Figure (II) Violent Crime and Conflict in Colombia (1995-2010) (a) Violent Crime: Homicide rates (per 100K individuals) (b) Armed Internal Conflict 45 Figure (III) Mean Firm Outcomes (variables in logs) (a) Production ($COL) (b) Intermediate consumption ($COL) (c) Electric Energy (Htz) (d) Number of firms (e) Gross investment ($COL) (f) Net investment ($COL) (g) Employees (h) Wages ($COL) 46 Figure (IV) Intensity and Pressure Migration Indexes (a) Intensity Index: Total Outflows / Mean Population (1995-2010) (b) Pressure Index: Total Inflows / Mean Population (1995-2010) 47 Figure (V) Predicted and Observed IDP Inflows 48 Appendix I: Characterizing the Municipalities in the Manufacturing Sample Mean Values of all Municipalities in Colombia vs. Municipalities in Manufacturing Sample Variable (mean values) All sample Excluded from Manufacturing Sample Within Manufacturing Sample Rural Population 9,862.67 6,232.21 19,045.86 Urban Population 27,313.09 3,457.95 87,574.27 Total Population 37,038.08 9,645.72 106,599.10 Working Age Population 31,144.88 7,585.76 85,112.57 Infant Mortality 23.29 24.10 21.23 Municipal GDP 359,164.90 65,235.67 1,058,189.00 Poverty Rates 0.51 0.51 0.52 Individuals Received (IDPs) 347.74 577.70 1,885.52 N of Municipalities 1123 806 317 49 Appendix II: Sample Robustness Table (II.1) Estimates for the Municipalities Available in the Labor Force Surveys: Intensive Margin of Production Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption Time Period 1995-2010 2002-2010 1995-2010 2002-2010 1995-2010 2002-2010 (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.0111** -0.0128** -0.0150*** -0.0182*** -0.0204*** -0.0205*** (0.00480) (0.00518) (0.00501) (0.00554) (0.00586) (0.00598) R-squared 0.950 0.958 0.941 0.949 0.938 0.948 Observations 81,079 67,326 81,068 67,314 80,941 67,194 Panel B. OLS Share of IDP (%Pop in Working Age) -0.000355 -0.00786 -0.00316 -0.0129 -0.00218 -0.0103 (0.00608) (0.00749) (0.00623) (0.00815) (0.00671) (0.00957) R-squared 0.950 0.958 0.941 0.949 0.938 0.948 Observations 81,079 67,326 81,068 67,314 80,941 67,194 Panel C. 2SLS 50 Share of IDP (%Pop in Working Age) -0.0298** -0.0429** -0.0402*** -0.0609*** -0.0545*** -0.0684*** (0.0126) (0.0170) (0.0132) (0.0182) (0.0155) (0.0195) Observations 81,079 67,326 81,068 67,314 80,941 67,194 Panel D. First Stage Dep. Variable: Share of IDP (% Pop in Working Age) Predicted Inflows 0.373*** 0.299*** 0.373*** 0.299*** 0.373*** 0.299*** (0.0181) (0.0151) (0.0181) (0.0151) (0.0181) (0.0151) First Stage F-statistic 425.73 392.41 425.69 392.37 424.46 391.73 Observations 81,083 67,326 81,083 67,326 81,083 67,326 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Yes Yes Table (II.2) Estimates for the Municipalities Available in the Labor Force Surveys: Extensive Margin of Production Dependent Variables (in logs) Entry Firm Exit Firm Time Period 1995-2010 2002-2010 1995-2010 2002-2010 (1) (2) (3) (4) Panel A. Reduced Form Predicted IDP Inflows -0.00413** 0.000425 0.00396* 0.00962*** (0.00199) (0.00236) (0.00226) (0.00218) R-squared 0.611 0.346 0.275 0.230 Observations 119,886 67,326 67,326 119,886 Panel B. OLS Share of IDP (%Pop in Working Age) -0.00327** -0.00105 0.00136 0.00332** (0.00139) (0.00183) (0.00255) (0.00167) R-squared 0.611 0.346 0.275 0.229 Observations 119,886 67,326 67,326 119,886 51 Panel C. 2SLS Share of IDP (%Pop in Working Age) -0.00885** 0.00130 0.0121* 0.0206*** (0.00426) (0.00724) (0.00694) (0.00485) Observations 119,886 67,326 67,326 119,886 Panel D. First Stage Dep. Variable: Share of IDP (% Pop in Working Age) Predicted Inflows 0.467*** 0.326*** 0.326*** 0.467*** (0.0236) (0.0176) (0.0176) (0.0236) First Stage F-statistic 393.01 342.55 342.55 393.01 Observations 119,886 67,326 67,326 119,886 Controls for all Panels Year FE Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Table (II.3) Estimates for the Municipalities Available in the Labor Force Surveys Dependent Variables (in logs) Gross Investment Total Employment Nominal Wages Time Period 1995-2010 2002-2010 1995-2010 2002-2010 1995-2010 2002-2010 (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows -0.00284 -0.00192 -0.00191 -0.00237 -0.00195 -0.00116 (0.0177) (0.0193) (0.00392) (0.00408) (0.00216) (0.00216) R-squared 0.690 0.703 0.932 0.941 0.862 0.860 Observations 47,728 40,452 81,025 67,286 80,604 66,891 Panel B. OLS Share of IDP (%Pop in Working Age) -0.0177 -0.0179 0.00202 -0.00142 0.00203 -0.000124 (0.0175) (0.0229) (0.00420) (0.00533) (0.00236) (0.00271) R-squared 0.690 0.703 0.932 0.941 0.862 0.860 Observations 47,728 40,452 81,025 67,286 80,604 66,891 52 Panel C. 2SLS Share of IDPs (%Pop in Working Age) -0.00725 -0.00575 -0.00512 -0.00793 -0.00521 -0.00389 (0.0452) (0.0578) (0.0105) (0.0136) (0.00576) (0.00719) Observations 47,728 40,452 81,025 67,286 80,604 66,891 Panel D. First Stage Dep. Variable: Share of IDP (% Pop. in Working Age) Predicted Inflows 0.373*** 0.299*** 0.373*** 0.299*** 0.373*** 0.299*** (0.0181) (0.0151) (0.0181) (0.0151) (0.0181) (0.0151) First Stage F-statistic 213.41 296.56 422.22 392 42.69 390.33 Observations 47,728 40,452 81,025 67,286 80,604 66,891 Controls for all Panels Year FE Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Yes Yes Appendix III: Dynamic Analysis of IDP Impacts on Firm Performance Table (III.1) Estimates for a lag of 1 period Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A. Reduced Form Predicted Inflows−1 -0.0118*** -0.0281*** -0.00730 -0.0164*** -0.0281*** -0.00691 -0.0173*** -0.0274*** -0.00926 (0.00410) (0.00661) (0.00458) (0.00480) (0.00763) (0.00487) (0.00531) (0.00957) (0.00645) R-squared 0.937 0.938 0.951 0.926 0.927 0.942 0.923 0.924 0.939 Observations 114,217 114,191 82,666 114,181 114,155 82,655 114,073 114,047 82,527 Panel B. OLS Share of IDP (%Pop in Working Age)−1 -0.0139*** -0.0159** -0.00869 -0.0190*** -0.0140* -0.00916 -0.00451 -0.00460 -0.00691 (0.00429) (0.00622) (0.00542) (0.00483) (0.00718) (0.00560) (0.00708) (0.0126) (0.00702) R-squared 0.937 0.938 0.951 0.926 0.927 0.942 0.923 0.924 0.939 Observations 114,222 114,196 82,670 114,186 114,160 82,659 114,078 114,052 82,531 Panel C. 2SLS Share of IDP (%Pop in Working Age)−1 -0.0269*** -0.0720*** -0.0169 -0.0373*** -0.0720*** -0.0160 -0.0394*** -0.0705*** -0.0215 (0.00923) (0.0171) (0.0105) (0.0108) (0.0197) (0.0111) (0.0123) (0.0252) (0.0148) 53 Observations 114,217 114,191 82,666 114,181 114,155 82,655 114,073 114,047 82,527 Panel D. First Stage Dep. Variable: Share of IDP (% Pop in Working Age)−1 Predicted Inflows−1 0.439*** 0.390*** 0.431*** 0.439*** 0.390*** 0.431*** 0.439*** 0.390*** 0.431*** (0.0155) (0.0212) (0.0180) (0.0155) (0.0212) (0.0180) (0.0155) (0.0212) (0.0180) First Stage F-statistic 802.03 340.59 575.08 802.25 340.53 575.07 809.65 347.7 573.26 Observations 114,217 114,191 82,666 114,181 114,155 82,655 114,073 114,047 82,527 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No No Yes No Homicides Rates No Yes No No Yes No No Yes No Conflict Controls No Yes No No Yes No No Yes No Additional Controls No No Yes No No Yes No No Yes Table (III.2) Estimates for a lag of 2 periods Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A. Reduced Form Predicted Inflows−2 -0.0124*** -0.0249*** -0.000743 -0.0169*** -0.0262*** 0.000404 -0.0167*** -0.0258*** -0.00514 (0.00454) (0.00719) (0.00515) (0.00539) (0.00846) (0.00597) (0.00540) (0.00977) (0.00618) R-squared 0.940 0.940 0.951 0.929 0.930 0.942 0.927 0.928 0.939 Observations 105,776 105,751 82,618 105,740 105,715 82,607 105,632 105,607 82,479 Panel B. OLS Share of IDP (%Pop in Working Age)−2 -0.0121*** -0.0117* -0.00277 -0.0178*** -0.0122* -0.00439 -0.00641 -0.00587 -0.00325 (0.00445) (0.00644) (0.00522) (0.00507) (0.00740) (0.00580) (0.00837) (0.0165) (0.00809) R-squared 0.940 0.940 0.951 0.929 0.930 0.942 0.927 0.928 0.939 Observations 105,780 105,755 82,622 105,744 105,719 82,611 105,636 105,611 82,483 Panel C. 2SLS Share of IDP (%Pop in Working Age)−2 -0.0263*** -0.0564*** -0.00168 -0.0359*** -0.0594*** 0.000914 -0.0355*** -0.0588** -0.0116 54 (0.00962) (0.0166) (0.0117) (0.0114) (0.0194) (0.0135) (0.0117) (0.0228) (0.0140) Observations 105,776 105,751 82,618 105,740 105,715 82,607 105,632 105,607 82,479 Panel D. First Stage Dep. Variable: Share of IDP (% Pop in Working Age)−2 Predicted Inflows−2 0.471*** 0.441*** 0.442*** 0.471*** 0.441*** 0.442*** 0.471*** 0.441*** 0.442*** (0.0148) (0.0231) (0.0175) (0.0148) (0.0231) (0.0175) (0.0148) (0.0231) (0.0175) First Stage F-statistic 1013.79 367.4 640.25 1013.79 367.39 640.17 1027.74 373.68 640.24 Observations 105,776 105,751 82,618 105,740 105,715 82,607 105,632 105,607 82,479 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No No Yes No Homicides Rates No Yes No No Yes No No Yes No Conflict Controls No Yes No No Yes No No Yes No Additional Controls No No Yes No No Yes No No Yes Table (III.3) Full Dynamic Analysis for 1 and 2 periods (Only IV is presented due to space restrictions) Dep. Variables (in logs) Production Intermediate Consumption Energy Consumption (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A. 2SLS Share of IDP (%Pop in Working Age) -0.0212** -0.0368*** -0.0265** -0.0261*** -0.0389** -0.0379*** -0.0304*** -0.0301 -0.0492*** (0.00832) (0.0141) (0.0116) (0.00916) (0.0166) (0.0121) (0.00987) (0.0187) (0.0141) Share of IDP (%Pop in Working Age)−1 -0.00735 -0.0365*** -0.0114 -0.0122* -0.0357*** -0.00847 -0.000222 -0.0288** -0.00701 (0.00580) (0.00907) (0.00780) (0.00712) (0.00985) (0.00877) (0.00725) (0.0140) (0.0110) Share of IDP (%Pop in Working Age)−2 -0.0230** -0.0438*** -0.00227 -0.0307*** -0.0471*** -0.00162 -0.0345*** -0.0491** -0.0158 (0.00960) (0.0153) (0.0113) (0.0117) (0.0180) (0.0135) (0.0109) (0.0204) (0.0130) Observations 0.940 0.940 0.951 0.929 0.930 0.942 0.927 0.927 0.939 Panel B. First Stage Dep. Variable: Share of IDPs (% Pop in Working Age) Predicted IDP Inflows 0.345*** 0.349*** 0.371*** 0.345*** 0.349*** 0.371*** 0.345*** 0.349*** 0.371*** (0.0137) (0.0238) (0.0186) (0.0137) (0.0238) (0.0186) (0.0137) (0.0238) (0.0186) First Stage F-statistic 856.39 305.35 448.22 856.63 305.45 448.12 856.86 302.63 447.09 55 Predicted IDP Inflows−1 -0.173*** -0.167*** -0.0896*** -0.173*** -0.167*** -0.0896*** -0.173*** -0.167*** -0.0896*** (0.00754) (0.0132) (0.0118) (0.00754) (0.0132) (0.0118) (0.00754) (0.0132) (0.0118) First Stage F-statistic 589.42 321.84 529.23 859.64 321.82 529.19 862.58 322.99 527.97 Predicted IDP Inflows−2 -0.0406*** -0.0453*** 0.0148 -0.0406*** -0.0453*** 0.0148 -0.0406*** -0.0453*** 0.0148 (0.00660) (0.0103) (0.0113) (0.00660) (0.0103) (0.0113) (0.00660) (0.0103) (0.0113) First Stage F-statistic 860.45 232.31 618.81 860.44 232.31 618.68 871.11 235.09 619.21 Observations 105,776 105,751 82,618 105,740 105,715 82,607 105,632 105,607 82,479 Controls for all Panels Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Year × Department FE No Yes No No Yes No No Yes No Homicides Rates No Yes No No Yes No No Yes No Conflict Controls No Yes No No Yes No No Yes No Additional Controls No No Yes No No Yes No No Yes Appendix IV: Restricting Sample to Balanced Panel Table (IV.1) Estimates For Firms Which do not Exit Dependent Variables (in logs) Production Intermediate Consumption Energy Consumption Gross Investment Employment Nominal Wages (1) (2) (3) (4) (5) (6) Panel A. Reduced Form Predicted Inflows 0.022 0.016 -0.030 0.371* -0.013 0.026 (0.034) (0.034) (0.038) (0.215) (0.024) (0.017) R-squared 0.982 0.980 0.977 0.877 0.975 0.949 Observations 6,376 6,376 6,376 1,342 6,376 6,376 Panel B. OLS Share of IDP (% Working Age Pop) -0.034 -0.022 0.030 0.136 0.011 -0.021 (0.046) (0.044) (0.053) (0.189) (0.037) (0.036) R-squared 0.982 0.980 0.977 0.876 0.975 0.949 Observations 6,376 6,376 6,376 1,342 6,376 6,376 Panel C. 2SLS Share of IDP (% Working Age Pop) 0.047 -0.003 0.085 -0.616 0.018 -0.023 (0.094) (0.117) (0.122) (0.455) (0.903) (0.049) 56 Observations 6,376 6,376 6,376 1,342 6,376 6,376 Panel D. First Stage Dep. Variable: Share of IDP (% Working Age Pop) Predicted Inflows 0.274*** 0.274*** 0.274*** 0.274*** 0.274*** 0.274*** (0.065) (0.065) (0.065) (0.065) (0.065) (0.065) First Stage F-statistic 19.06 19.06 19.28 28.96 19.06 19.3 Observations 6,376 6,376 6,376 1,342 6,376 6,376 Controls (for all panels) Year FE Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Additional Controls Yes Yes Yes Yes Yes Yes