The local human capital cost of oil exploitation

Balza, Lenin H.; de los Rios Rueda, Camilo; Mori, Raul Jimenez; Manzano, Osmel Working Paper The local human capital cost of oil exploitation IDB Working Paper Series, No. IDB-WP-1258 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Balza, Lenin H.; de los Rios Rueda, Camilo; Mori, Raul Jimenez; Manzano, Osmel (2021) : The local human capital cost of oil exploitation, IDB Working Paper Series, No. IDBWP-1258, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0003382 This Version is available at: https://hdl.handle.net/10419/237513 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode The Local Human Capital Cost of Oil Exploitation Lenin H. Balza Camilo De Los Rios Raul Jimenez Mori Osmel Manzano IDB WORKING PAPER SERIES No IDB-WP-1258 Inter-American Development Bank Infrastructure and Energy Sector July 2021 The Local Human Capital Cost of Oil Exploitation Lenin H. Balza Camilo De Los Rios Raul Jimenez Mori Osmel Manzano Inter-American Development Bank Infrastructure and Energy Sector July 2021 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library The local human capital costs of oil exploitation / Lenin H. Balza, Camilo De Los Rios, Raul Jimenez Mori, Osmel Manzano. p. cm. — (IDB Working Paper Series; 1258) Includes bibliographic references. 1. Petroleum industry and trade-Social aspects-Colombia-Econometric models. 2. Human capital-Colombia-Econometric models. 3. College attendance-Colombia- Econometric models. I. Balza, Lenin. II. De Los Rios, Camilo. III. Jimenez Mori, Raul. IV. Manzano, Osmel, 1971- V. Inter-American Development Bank. Infrastructure and Energy Sector. VI. Series. IDB-WP-1258 http://www.iadb.org Copyright © 2021 Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 Attribution- NonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc- nd/3.0/igo/legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB’s name for any purpose other than for attribution, and the use of IDB’s logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association’s EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication’s author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. [email protected]; c[email protected]; rjim[email protected]; [email protected] The Local Human Capital Costs of Oil Exploitation∗ Lenin H. Balza†Camilo De Los Rios‡Raul Jimenez Mori§ . Osmel Manzano¶ July, 2021 Abstract This paper explores the impacts of oil exploitation on human capital accumulation at the local level in Colombia, a resource-rich developing country. We provide evidence based on detailed spatial and temporal data on oil exploitation and education, using the number of wells drilled as an intensity treatment at the school level. To find causal estimates we rely on an instrumental variable approach that exploits the exogeneity of international oil prices and a proxy of oil endowments at the local level. Our results indicate that oil has a negative impact on human capital since it reduces enrollment in higher education. Furthermore, it generates a delay in the decision to enroll in higher education and leads students to prefer technical areas of study and programs in social science, business, and law. However, we do not find any effects on quality or tertiary education completion. Our results are robust to a number of relevant specification changes and we stress the role of local markets and spillovers as the main transmission channel. In particular, we find that higher oil production causes an increase in formal wages but that there is no premium to tertiary education enrollment. Keywords: Natural Resource Exploitation, Human Capital, Colombia. JEL classification: I23, Q32, Q35. ∗The views expressed herein are entirely those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. This paper is a joint effort of the Infrastructure and Energy Sector and the Country Department Andean Group at the Inter-American Development Bank, together with the Development Effectiveness Division at IDB Invest. We thank Luis Omar Herrera-Prada for kindly sharing data on education in Colombia. We also express our gratitude to Stephanie Majerowicz and one anonymous reviewer for carefully reading our first manuscript and for providing insightful feedback. All remaining errors are our own responsibility. †Inter-American Development Bank. Infrastructure and Energy Sector.  :lenin[email protected] ‡Inter-American Development Bank. Infrastructure and Energy Sector.  :camilo[email protected] §IDB Invest. Development Effectiveness Division.  :[email protected] ¶Inter-American Development Bank. Country Department Andean Group.  :[email protected] 1 Introduction Natural resource exploitation and its effect on the economic performance of countries has been extensively studied. Some scholars argue that human capital accumulation is one of the mechanisms through which the natural resource curse might operate. However, consensus has yet to be reached as to whether natural resource exploitation hinders or fosters human capital accumulation. Certain studies find a negative relationship between natural resource exploitation and human capital accumulation (Birdsall et al.,2001;Gylfason,2001; Papyrakis and Gerlagh,2004), while others provide compelling evidence of a positive relationship (Smith,2015;Stijns,2006). The quality of the institutions (Ebeke et al.,2015) and the type of natural resource measures used (Stijns,2006) may drive these differences. While much research has been done at the country level, we still know surprisingly little about the effects of natural resources at the subnational level (Arag´on et al.,2015). Findings at this level indicate that natural resource exploitation causes lower school attendance (Santos, 2018), educational attainment (Ahlerup et al.,2020), education quality (Bonilla,2020), promotion rates (Cascio and Narayan,2020), and tertiary education enrollment (Bonilla,2020). However, results on this front are also mixed as there is evidence of natural resource exploitation increasing promotion rates, high school enrollment (Bonilla,2020) and education quality (Ag¨uero et al.,2021). In this paper we explore the impacts of oil exploitation on human capital accumulation at the local level. Specifically, we study the case of Colombia, a developing and highly oildependent country, using administrative data from 2002 to 2014 on all high school students in the country. The data captures education quality (as measured with performance in a national exam), tertiary education decisions and the exact location of schools. We also use data from the National Hydrocarbons Agency (ANH), which contains the location and spud date of every oil well ever drilled in Colombia.1 We exploit the exogeneity of international oil prices and the location of oil endowments, relying on an instrumental variable approach to find causal estimates. As a proxy of schoollevel oil production, we use the number of wells drilled within a certain distance of the school. As this measure is potentially biased due to endogeneity issues, we instrument it with the interaction between international oil prices and an oil suitability measure at the school level. The latter is measured as the total number of wells drilled in the buffer around the school for a given year before the first year of our analysis. The basic intuition behind our instrument is that international oil prices and the oil endowments of the school should affect human capital accumulation decisions only as they change the year-to-year oil production at the school level. In our context, the results suggest that oil operates as a curse in terms of human capital accumulation at the local level. First, we find that an increase in oil exploitation reduced the enrollment rate in higher education programs. An increase of one standard deviation in the number of wells generated a decrease of 2.81 pp in the enrollment rate in our benchmark model. Second, our results show that oil exploitation affects the talent allocation decisions of students. An increase in the number of nearby wells causes a higher proportion of students to 1The spud date indicates the day the drilling process began. 1 enroll in technical degree programs. This is consistent with the structure of oil production in Colombia, which employs a larger amount of non-trained workers and those with a technical training than professional workers, which we term the ”talent allocation effect”. Likewise, oil activity increases the proportion of students and the probability of enrolling in higher education programs in business, economics, law, social sciences and related programs, as compared to STEM. We also find evidence of what we call a ”delay effect”. In line with Emery et al. (2012), our findings indicate that oil exploitation changes the timing of tertiary education decisions, generating a delay in students’ enrollment of about three months. These findings are robust to a number of relevant transformations, including changing the definition of school vicinity, adjusting for the life cycle of oil wells, and using different thresholds to measure our instrument. Moreover, our results are not driven by selection, as there is no effect on the number of students, or on education quality. This is in line with our identification strategy, which stresses the fact that schools are unable to directly benefit from rents derived from oil exploitation in their vicinity. Focusing our analysis at the school and individual level allows us to disentangle the higher rent effect from a local economic activity and labor market effect. Using data on formal wages for a subset of the students in our sample, we show that the latter mechanism is most likely operating in our setting. We find that a one standard deviation increase in the number of wells rises annual income by approximately 4 current minimum wages. In conducting this analysis, we contribute to the literature on the dynamics of natural resource exploitation. Very little research has explored the effects of oil exploitation on educational outcomes at the subnational level. Cascio and Narayan (2020) estimate that fracking increased high school dropout rates in production areas while James (2017) finds lower graduation rates during oil booms in resource-rich states in the US. Tangential evidence shows that public expenditures on education are higher in oil-producing states or municipalities (Caselli and Michaels,2013;James,2017), though this does not necessarily translate into higher welfare (Caselli and Michaels,2013), perhaps due to patronage (Robinson et al.,2006). Other studies focus on the effects on corruption (Brollo et al.,2013;James and Rivera,2019;Vicente,2010), civil conflict and democracy (Br¨uckner and Ciccone,2010; Dube and Vargas,2013;Tsui,2011), wages (Kearney and Wilson,2018), and general local economic indicators (Bartik et al.,2019). We build on this body of work by carrying out our assessment at a non-administrative local level (schools and individuals). This allows us to identify one of the main mechanisms of transmission of the resource curse (or blessing), namely the local market spillover effect. We also join a nascent strand of literature that uses resource endowments and exploitation measures to overcome certain challenges in identifying the causal effects of natural resources on different scenarios. Most such studies rely on an instrumental variable approach to deal with selection, measurement error, or omitted variable biases. Examples include using some measures of oil or gas shale formations and gas well drilling (Maniloff and Mastromonaco, 2017), production (Brown,2014), employment (Feyrer et al.,2017), gas property taxes (Weber et al.,2016), and gold deposits and gold mining titles (Bonilla,2020). Yet, to the best of our knowledge, our paper is the first to use conventional oil wells as a treatment measure at 2 a non-administrative local level and fully exploit the temporal and spatial variation in well drilling. In addition, our study is tangentially related to the conflict literature, in particular that of armed conflict in Colombia. Conflict in this country has been measured with war-related episodes (e.g. Dube and Vargas,2013;Prem et al.,2020) or self-reported violence victims (e.g. Ib´a˜nez and V´elez,2008). We contribute to this work by proposing a new, local measure of violence using detailed data on the location of all accidents involving landmines. The data, collected by the High Commissioner for Peace (OACP), allows us to control for violence shocks at a very granular level, that of schools and individuals. The rest of the paper is organized as follows. Section 2 gives an overview of oil exploitation in the Colombian context while Section 3 describes the data. Section 4 then presents our empirical strategy. Section 5 shares the results of the study and Section 6 discusses the robustness checks. Section 7 presents evidence of our main mechanism and, finally, Section 8concludes with a discussion and a reflection on the policy implications of our findings. 2 Oil Exploitation in Colombia Natural resource exploitation has played a central role in the economic performance of many countries in Latin American and the Caribbean (LAC), constituting one of the key pillars of regional economies for many decades. Oil, gas, and mining rents alone accounted, on average, for 5% of LAC countries’ GDP and nearly 27% of the region’s total exports in the last decade. The extractive sector is particularly important in Colombia. Oil, gas, and mining represented nearly 60% of the total exports and the rents accounted for nearly 6% of the national GDP in that same period.2Furthermore, economic growth in Colombia and the region is highly correlated with the international oil price (see Figure 1). During the global financial crisis of 2008, oil prices plummeted but quickly returned to record-high levels in 2010. 2Authors’ calculations based on data from the World Bank and the Atlas of Economic Complexity at Harvard University. 3 Figure 1: Dependence on Oil Performance 20.00 40.00 60.00 80.00 100.00 $/bbl -2 0 2 4 6 8 GDP growth (%) 2000 2003 2006 2009 2012 2015 2018 Year Colombia LAC Oil Price (right axis) The oil price shown is an equally weighted measure of the Brent, Dubai and WTI spot prices. Source: Authors’ calculations based on the World Bank’s pink sheet and development indicators. The Colombian oil sector, once dominated by the public sector, has seen major efforts to increase private participation since 2000. For instance, State participation in oil exploration and exploitation activities was mandatory until 2003, when the requirement was dropped. In return, private firms must pay royalties based on reported production.3Nevertheless, the State is the majority shareholder of Ecopetrol, one of Colombia’s biggest oil companies. Since 2003, the ANH has been in charge of regulating and managing the State participation in the Colombia oil sector. Changes to Ecopetrol and the creation of the ANH were an additional component of the policies seeking to ensure efficiency in the oil sector Balza and Espinasa (2015); G´omez (2013). Exploration activities in Colombia’s offshore fields have been largely unsuccessful and, to this day, the vast majority of the country’s production occurs in onshore oil fields. Oil production is scattered throughout the country, with particularly high levels of production in some regions. Figure 2 shows the mean oil production between 2011 and 2016 for all municipalities and the location of every oil well ever drilled in the country. While oil production is concentrated in the country’s western, production sites are also present along the northern coast, the central region and the south. The overlap between oil-producing regions 3Before 2012 only oil-producing municipalities and departments received those royalties. Since then, all municipalities are eligible to receive part of the royalties derived from natural resource exploitation. See Bonet-Mor´on et al. (2018) for a full account on the history of the fiscal regime in Colombia. 4 time in any area of the country. Measures at the municipality level are, by nature, less flexible and usually show little spatial variation over time. Figure 5: Landmines and FARC Presence Landmines accidents after 2010. Source: FARC presence data made available by Prem et al. (2020) and collected and edited by Restrepo et al. (2004) and the Universidad del Rosario. Landmine accidents reports were collected by the Oficina del Alto Comisionado para la Paz (OACP) Other In addition to the above-described information, we also have data on income for a subset of the students in our sample. Specifically, we draw from the PILA database, which records all payments to social security in the formal sector and the number of days worked per year, as well as the main ISIC code of the industry in which the student was employed. It thus covers the salaries and worked days from 2008 until 2014 for all students who joined the formal labor force. Annual income is simply the sum of the reported contributions to social security and is normalized by the legal minimum wage in a given year. 11 4 Identification Strategy To estimate the impact of oil activity on different dimensions of human capital accumulation, we propose a school-level measure of oil production using information on the number of wells drilled in the vicinity of the educational institution. As discussed in section 3, the number of oil wells is the best proxy for oil production at the local level. Measuring the number of oil wells drilled within a buffer around the school allows us to use the most detailed available information on oil production in Colombia and its distribution over space and time. Using school coordinates and detailed information on oil wells locations for every year in our study, we can count the number of wells that are drilled around each school in the country in a given year, allowing us to capture the effect of higher oil activity. In our benchmark model we count wells that are within a 10km buffer around the school, though our results are also robust to buffers ranging from 5km to 30km. The average school has 12.9 wells within the 10km buffer, but this figure is quite variable, with a standard deviation of 105.05 wells (see Table 4). Table 5 shows the number of schools that had at least one oil well drilled within these buffers by the year 2000. Naturally, as the buffer size increases, the number of schools with at least one oil well within the buffer area rises as well. Table 4: Wells and Landmines at the School Level Variable Mean Std. Dev. Max Obs. Wells 12.9 105.05 1,993 69,050 Landmines 0.25 1.16 26 69,050 Note: Authors’ calculations based on data from ANH, Ministry of Education, and OCAP. Values are calculated using a buffer of 10km around every school, which corresponds to our benchmark model. Table 5: Schools with Nearby Wells Prior to 2000 Buffer 5 km 10 km 20 km 30 km Schools 851 1,946 3,450 4,394 Note: Authors’ calculations based on data from ANH and Ministry of Education. Number of schools that have at least one well drilled nearby prior to the year 2000 using different buffers. We estimate the impact of oil activity on educational outcomes by using the number of wells drilled around the school as an intensity treatment. Since our data allows for aggregation at the school level, the model specification depends on the unit of analysis. For educational outcomes at the school level (e.g., enrollment rate), we estimate a fixed effects panel as in Equation 1.yst measures a dimension of human capital in school s and year t.W ellsk st is our variable of interest, measuring the total number of wells that have ever been drilled around the school sby year twithin a buffer of kkilometers. γs are school level fixed effects, λdt captures department-time trends, and Xk st controls for the 12 total number of landmine accidents in the vicinity of sin year t. When using individual measures of human capital accumulation (e.g., test scores), we estimate a repeated crosssection as in Equation 2. The model also includes school fixed effects, department-time trends, and controls for landmine accidents and student’s gender and age. Since the high school graduation year and the year of enrollment in tertiary education are both available in our dataset, we assign a student to the graduation year when estimating the effects on enrollment decisions and assign the student to the enrollment year when estimating completion decisions. In our benchmark model, we count wells within 10 km, but present robustness checks at different buffer ranges. The decision to employ a 10km buffer is also based on the relatively low number of schools with wells within a smaller buffer. yst =βWellsk st +δXk st +γs+λdt +st (1) yist =βWellsk ist +δXk ist +γs+λdt +ist (2) The school fixed effects swept out any time-invariant characteristics of the schools in our dataset, while the department-time trend captures all common shocks over time that schools in the same department might face. We use quadratic trends as they fit better with the oil price dynamics during the study period, but our results are robust to different polynomial specifications. An ordinary least squares (OLS) estimation of Equation 2 and Equation 1 has two econometric concerns in the setting of this study: omitted variables and reverse causality. There are two obvious omitted variables that can affect tertiary education decisions and also be correlated with oil drilling activity. First, an oil company might take public education spending into account when locating oil wells, as the latter could influence the local labor supply and as well as the costs of drilling. The fact that we use schools and individuals as our unit of analysis mitigates this concern as schools do not directly receive any rents derived from oil activities. Second, violence and conflict have been shown to affect education decisions in several different contexts (Barrera and Iba˜nez,2004;Miguel and Roland,2011;Shemyakina,2011), including Colombia (Rodriguez and S´anchez,2012). Using the number of landmine accidents as a control for violence alleviates this concern, as landmines are the best available proxy of violence at the school level. However, studies have also shown that oil activity can generate increases in civil conflict in Colombia (Dube and Vargas,2013), as terrorist groups try to capture the rents generated by oil exploitation. This may raise concerns over bad controls in our estimation. However, the main mechanism identified by Dube and Vargas (2013) is not binding in our estimation as schools do not directly receive rents derived from oil exploitation. Furthermore, landmines have been used in Colombia by terrorist groups to attack the military or to protect their coca plantations (Ruiz and Pinto,2017), but not in operations related to oil wells. To account for reverse causality and other potentially omitted variables, we use an instrumental variable (IV) approach to estimate the two-stage least squares (2SLS) estimator. We 13 begin by instrumenting our oil activity variable with the interaction between international oil prices and the total number of wells drilled around school swithin a buffer of kkm until year T, prior to 2001. Since we use all the information available on oil well drilling and wells are the best available predictor of production at the local level, our instrument can be interpreted as an oil suitability proxy at the school level. In our benchmark model we use T= 2000, but we present robustness checks using 1970,1980 and 1990 as thresholds. Equation 3 and Equation 4 present the first stage of the estimations at the school and individual level, respectively. Wellsk st =βPricet·W ellsk s;T+δXst +γs+λdt +st (3) Wellsk ist =βPricet·W ellsk is;T+δXist +γs+λdt +ist (4) We argue that both the exogeneity and the exclusion restriction of our instrument are fully satisfied. First, Colombia is a price taker in the international oil market. Thus, international oil prices are not correlated with the error term as they are set in the international market. The second part of our instrument reflects geological characteristics that cannot have endogenous responses in the model. Moreover, wells that have already been drilled cannot disappear, and the intensity of the measure provides information on the possible oil endowments of the school’s immediate surroundings. Together these facts ensure the exogeneity of our instrument. However, our instrument is subject to concerns relating to path dependence, since oil wells are often drilled repeatedly in the same areas. We show that our findings are robust to using 1970, 1980, and 1990 as thresholds to create our instrument, which alleviates this concern. Measuring the total number of oil wells in a year prior to the first year of our study period ensures that the exclusion restriction of our instrument is satisfied. Since schools do not directly receive rents from oil exploitation, oil endowments and international prices only affect human capital decisions through their impact on economic activity. The basic intuition behind our instrument is that international oil prices should disproportionately affect the number of wells drilled in schools where oil production is more likely to be successful, namely sites with previous evidence of oil presence.7 Though a reduced form of our estimation has been used in related studies to estimate causal effects (e.g.,Black et al. (2005); Bonilla (2020); Dube and Vargas (2013); Michaels (2011)), our paper is able to use much more detailed data to capture current spatial and temporal variation in oil activity. Since we are interested in estimating the impact of increases in oil activity on human capital accumulation measures, our instrumental variable approach gives us the precise local average treatment effect (LATE) that we are looking for and satisfies the exclusion restriction. Finally, our instrument is the best available predictor of oil well drilling at the local level. As discussed in section 3, oil well drilling is very responsive to the international oil price. Moreover, the general uncertainty in the literature and the public debate on the future 7Bonilla (2020) uses a similar strategy in some exercises for the case of gold mining. 14 production of oil stems from poor quality data. The modeling of the natural distribution of oil fits fairly well with past oil production (Laherr`ere,2003) and our instrument accounts for granular spatial variation in oil activity by measuring the number of oil wells that have been drilled around the school. For ease of interpretation, we standardize both our instrument and instrumented variables. To give a more intuitive interpretation of the first stage of our IV, we use the logarithm of the international oil price. Our estimates use the Brent price, which is the benchmark for Colombian oi, though our results remain unchanged when using WTI or a composite crude price. The results of this study can be interpreted as the effect of the number of wells changing by one standard deviation, which is the best proxy available of oil production at the school level. 5 Results Enrollment Decisions We begin our analysis by presenting the impact of oil activity on enrollment decisions. Table 6 shows how oil activity affects tertiary enrollment decisions at the school level. Panel A contains the results from the OLS, estimating Equation 1 without instrumenting our wells variable, while Panel B presents the results of the reduced form and Panel C provides the IV estimates. Each column shows the results for the outcomes of interest, which in this case are the enrollment rate and program selection. The first stage of our IV estimates indicates the relevance of our instrument. Following Lee et al. (2020), the Kleibergen-Paap F statistic is well above the 104.7 threshold. Furthermore, all of the estimates go in the same direction, with the IV estimates larger than the OLS results. Our results suggest that an increase of one standard deviation in the number of well drilled decreases the tertiary enrollment rate by 2.81pp. This magnitude represents more than one eighth of a standard deviation of the enrollment rate in our sample. Likewise, it has a negative effect on the number of students enrolling in professional programs relative to technical programs. 15 Table 6: Effect of Oil Exploitation on School Level Human Capital Measures Enrolment Rate Professional Program Intensity A. OLS Wells -0.938 -1.603 (0.62) (1.55) Observations 68,701 66,635 B. Reduced Form Price ·Wells2000 -0.351** -0.859*** (0.15) (0.31) Observations 68,701 66,635 C. IV Estimates Wells -2.813** -7.004** (1.21) (2.79) Observations 68,701 66,635 Kleibergen-Paap F 675 703 School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. For a better sense of the interpretation of these results, Table 7 provides the estimations at the individual level. In addition to presenting the probability of enrollment and program selection, we also measure the impact on the number of years between high school graduation and tertiary education enrollment. Consistent with the school-level estimates, we find that increased oil exploitation lowers tertiary education enrollment and the relative probability of enrolling in a professional degree. Specifically, an increase of one standard deviation in the number of nearby oil wells decreases tertiary enrollment by 1.8 pp and professional degree enrollment by 3.6 pp, while also delaying students’ decisions to pursue tertiary education by an average of half a semester. 16 Table 7: Effect of Oil Exploitation on Individual-Level Human Capital Measures Enroled Professional Degree Semesters to Enrolment A. OLS Wells -0.007** -0.006 0.185*** (0.00) (0.00) (0.04) Observations 2,794,699 1,164,185 1,225,191 B. Reduced Form Price ·Wells2000 -0.002*** -0.004*** 0.053*** (0.00) (0.00) (0.01) Observations 2,794,699 1,164,185 1,225,191 C. IV Estimates Wells -0.018*** -0.036*** 0.484*** (0.01) (0.01) (0.10) Observations 2,794,699 1,164,185 1,225,191 Kleibergen-Paap F 44,148 20,216 25,040 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. We now turn to the specific programs that individuals choose to pursue, examining the probability of enrollment in STEM programs, business, economics, law and related programs, and all other programs (Table 8). These outcomes, are of interest for two reasons. First, STEM programs have a higher wage premium (Saavedra et al.,2017). Second, the literature has found that positive oil shocks generate higher enrollment in business, economics, law and related programs (Ebeke et al.,2015), which might in itself be pervasive since those programs can be rent-seeking behavior, while an oversupply of labor in those subjects might hinder economic growth (Murphy et al.,1991). In line with previous studies, our results suggest that oil activity decreases the probability of enrolling in STEM programs (Bonilla, 2020) and increases the probability of entering programs in the social sciences, business, law and related areas (Ebeke et al.,2015). 17 Table 8: Effect of Oil Exploitation on Individual-Level Human Capital Measures STEM Soc. Sci, Business, Law Others A. OLS Wells -0.039*** 0.034*** 0.007** (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 B. Reduced Form Price ·Wells2000 -0.009*** 0.008*** 0.000 (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 C. IV Estimates Wells -0.076*** 0.074*** 0.001 (0.01) (0.01) (0.01) Observations 1,293,718 1,293,718 1,293,718 Kleibergen-Paap F 25,807 25,807 25,807 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Completion Decisions We now turn to analyze completion decisions among those who decided to enroll in tertiary education. Since a professional program in Colombia takes between 4 and 5 years to complete, we limit our sample to students who enrolled in higher education prior to 2010. Our data includes information on whether the students in the sample completed or deserted the degree program. Table 9 shows the effect of completion and desertion decisions both at the school and the individual level. We do not find any impact of oil activity on these decisions at either level. However, these results are not unsurprising in that it is plausible that students across the country accurately consider the costs of completing their tertiary education programs and that oil activity does not play a role once they are already enrolled. 18 Table 9: Effect of Oil Exploitation on Individual-Level Human Capital Measures Completion Desertion Individual School Individual School A. OLS Wells 0.011 2.293 -0.022* -3.015 (0.01) (2.22) (0.01) (2.20) Observations 536,312 40,011 536,312 40,011 B. Reduced Form Price ·Wells2000 -0.003 -0.367 0.002 0.460 (0.00) (0.35) (0.00) (0.36) Observations 536,312 40,011 536,312 40,011 C. IV Estimates Wells -0.038 -3.988 0.018 5.000 (0.03) (3.90) (0.03) (3.99) Observations 536,312 40,011 536,312 40,011 Kleibergen-Paap F 8,444 473 8,444 473 School fixed effects; Time-Department quadratic trends; Landmines, age and gender included as control. Wells are standardized and Price is in logs. Shock assigned in year of enrolment in tertiary education. Sample trimmed to 2010. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Quality and Selection Since the impact of oil exploitation on educational outcomes could be driven by the general quality of education in schools near oil-producing areas, we test whether increased oil activity has an impact on students’ test scores. Table 10 provides the results of this analysis indicating no impact on the percentile of test scores. his finding makes sense as schools do not directly receive any rents from oil exploitation, . Selection represents a potential barrier to the identification of our results. We consequently test whether there is any impact of oil activity on the number of students in each school. Our results consistently show that this form of selection is not present. Unfortunately, we cannot check whether a migration-type self-selection occurs since we lack information on each student’s place of origin. However, as we previously discussed, migration-related selfselection bias naturally attenuates over time. Moreover, as we show in the next subsection, our results are robust to using different thresholds for constructing our instrument, making this is a minor concern. 19 Table 10: Effect of Oil Exploitation on Individual-Level Human Capital Measures Test Scores Number of Students A. OLS Wells -0.311** -1.315* (0.16) (0.75) Observations 2,771,137 68,701 B. Reduced Form Price ·Wells2000 -0.049 -0.323 (0.03) (0.24) Observations 2,771,137 68,701 C. IV Estimates Wells -0.417 -2.585 (0.30) (1.89) Observations 2,771,137 68,701 Kleibergen-Paap F 43,875 675 School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Another source of selection might stem from schools opening or closing depending on local oil endowments. We performed two exercises to test whether such selection is present our context and if it might be biasing our results. First, we replicated our results using only schools with information for at least the first two years of our study period. We then dropped schools with less than six years (half the length of the study period) of information from our sample. Appendix Table A1,Table A2,Table A3, and Table A4 present the results of these exercises. Both the qualitative and quantitative characteristics of our main estimations remain fairly unchanged, indicating that this kind of selection bias does not affect our results. Deviations from Benchmark Model Oil Suitability Threshold To further support our claims regarding the exogeneity of the instrument, we repeat our analysis using a more restrictive measure of our instrument. For instance, we measure our oil suitability proxy using 1970, 1980, and 1990 as cutoff years. While measuring our instrument starting further back in time weakens potential bias from path dependence, doing so also causes us to lose information as the number of wells naturally increases over time. Appendix Table A5,Table A6,Table A7,Table A8, and Table A9 show that all of our estimations are robust to this modification. Note that, by construction, the OLS estimation is the same as in the benchmark model. 20 Br¨uckner, M. and Ciccone, A. (2010). International commodity prices, growth and the outbreak of civil war in sub-saharan africa. The Economic Journal, 120(544):519–534. [page 2.] Cascio, E. and Narayan, A. (2020). Who needs a fracking education? the educational response to low-skill biased technological change. ILR Review. [pages 1and 2.] Caselli, F. and Michaels, G. (2013). Do oil windfalls improve living standards? evidence from brazil. American Economic Journal: Applied Economics, 5(1):208–38. [page 2.] Dube, O. and Vargas, J. F. (2013). Commodity price shocks and civil conflict: Evidence from colombia. The review of economic studies, 80(4):1384–1421. [pages 2,3,13, and 14.] Ebeke, C., Omgba, L. D., and Laajaj, R. (2015). Oil, governance and the (mis) allocation of talent in developing countries. Journal of Development Economics, 114:126–141. [pages 1, 9,10,17, and 23.] Emery, J. H., Ferrer, A., and Green, D. (2012). Long-term consequences of natural resource booms for human capital accumulation. ILR Review, 65(3):708–734. [pages 2and 9.] Feyrer, J., Mansur, E. T., and Sacerdote, B. (2017). Geographic dispersion of economic shocks: Evidence from the fracking revolution. American Economic Review, 107(4):1313– 34. [pages 2,21, and 22.] Gylfason, T. (2001). Natural resources, education, and economic development. European economic review, 45(4-6):847–859. [page 1.] G´omez, A. (2013). Columna vertebral del sector hidrocarburos. Technical report, SURA. [pages 4,5, and 6.] Ib´a˜nez, A. M. and V´elez, C. E. (2008). Civil conflict and forced migration: The micro determinants and welfare losses of displacement in colombia. World Development, 36(4):659– 676. [page 3.] Jacobsen, G. D. and Parker, D. P. (2016). The economic aftermath of resource booms: evidence from boomtowns in the american west. The Economic Journal, 126(593):1092– 1128. [pages 21 and 24.] James, A. (2017). Natural resources and education outcomes in the united states. Resource and Energy Economics, 49:150–164. [page 2.] James, A. and Rivera, N. M. (2019). Oil, Politics, and “Corrupt Bastards”. Working Papers 2019-04, University of Alaska Anchorage, Department of Economics. [page 2.] Kearney, M. S. and Wilson, R. (2018). Male earnings, marriageable men, and nonmarital fertility: Evidence from the fracking boom. Review of Economics and Statistics, 100(4):678–690. [page 2.] Laherr`ere, J. (2003). Future of oil supplies. Energy exploration & exploitation, 21(3):227–267. [pages 6and 15.] 27 Lee, D. L., McCrary, J., Moreira, M. J., and Porter, J. (2020). Valid t-ratio inference for iv. arXiv preprint arXiv:2010.05058. [pages 15 and 21.] Londo˜no-V´elez, J., Rodr´ıguez, C., and S´anchez, F. (2020). Upstream and downstream impacts of college merit-based financial aid for low-income students: Ser pilo paga in colombia. American Economic Journal: Economic Policy, 12(2):193–227. [page 9.] Maniloff, P. and Mastromonaco, R. (2017). The local employment impacts of fracking: A national study. Resource and Energy Economics, 49:62–85. [page 2.] Marchand, J. (2012). Local labor market impacts of energy boom-bust-boom in western canada. Journal of Urban Economics, 71(1):165–174. [pages 22 and 24.] Mehlum, H., Moene, K., and Torvik, R. (2006). Institutions and the resource curse. The economic journal, 116(508):1–20. [page 9.] Michaels, G. (2011). The long term consequences of resource-based specialisation. The Economic Journal, 121(551):31–57. [page 14.] Miguel, E. and Roland, G. (2011). The long-run impact of bombing vietnam. Journal of development Economics, 96(1):1–15. [page 13.] Murphy, K. M., Shleifer, A., and Vishny, R. W. (1991). The allocation of talent: Implications for growth. The quarterly journal of economics, 106(2):503–530. [pages 17 and 24.] Papyrakis, E. and Gerlagh, R. (2004). The resource curse hypothesis and its transmission channels. Journal of Comparative Economics, 32(1):181–193. [page 1.] Prem, M., Saavedra, S., and Vargas, J. F. (2020). End-of-conflict deforestation: Evidence from colombia’s peace agreement. World Development, 129:104852. [pages 3,10,11, and 30.] Restrepo, J. A., Spagat, M., and Vargas, J. F. (2004). The dynamics of the colombian civil conflict: A new data set. Homo Oeconomicus, 21:396–429. [pages 11 and 30.] Robinson, J. A., Torvik, R., and Verdier, T. (2006). Political foundations of the resource curse. Journal of development Economics, 79(2):447–468. [pages 2and 23.] Rodriguez, C. and S´anchez, F. (2012). Armed conflict exposure, human capital investments, and child labor: evidence from colombia. Defence and peace economics, 23(2):161–184. [page 13.] Ross, M. L. (2015). What have we learned about the resource curse? Annual Review of Political Science, 18:239–259. [pages 23 and 24.] Ruiz, G. and Pinto, M. E. (2017). La guerra escondida: Minas antipersonal y remanentes explosivos en Colombia. Centro Nacional de Memoria Hist´orica. [pages 10 and 13.] 28 Saavedra, J., Maldonado, D., Santibanez, L., and Prada, L. O. H. (2017). Premium or penalty? labor market returns to novice public sector teachers. Technical report, National Bureau of Economic Research. [page 17.] Santos, R. J. (2018). Blessing and curse. the gold boom and local development in colombia. World Development, 106:337–355. [pages 1and 24.] Shemyakina, O. (2011). The effect of armed conflict on accumulation of schooling: Results from tajikistan. Journal of Development Economics, 95(2):186–200. [page 13.] Smith, B. (2015). The resource curse exorcised: Evidence from a panel of countries. Journal of Development Economics, 116:57–73. [page 1.] Stijns, J.-P. (2006). Natural resource abundance and human capital accumulation. World development, 34(6):1060–1083. [page 1.] Tsui, K. K. (2011). More oil, less democracy: Evidence from worldwide crude oil discoveries. The Economic Journal, 121(551):89–115. [page 2.] Vicente, P. (2010). Does oil corrupt? evidence from a natural experiment in west africa. Journal of Development Economics, 92(1):28–38. [pages 2and 23.] Weber, J. G. (2012). The effects of a natural gas boom on employment and income in colorado, texas, and wyoming. Energy Economics, 34(5):1580–1588. [pages 21 and 24.] Weber, J. G., Burnett, J. W., and Xiarchos, I. M. (2016). Broadening benefits from natural resource extraction: Housing values and taxation of natural gas wells as property. Journal of Policy Analysis and Management, 35(3):587–614. [page 2.] 29 8 Appendix Figure A1: Landmines and FARC presence All mines included. Source: FARC presence made available by Prem et al. (2020), and collected and edited by Restrepo et al. (2004) and Universidad del Rosario. Landmine accidents reports were collected by the Oficina del Alto Comisionado para la Paz (OACP) 30 Table A1: Effect of Oil Exploitation on Individual Level Human Capital Measures Enrolment Rate Professional Program Intensity A. OLS Wells -0.793 -0.708 -0.358 -0.844 (0.61) (0.62) (1.58) (1.57) Observations 49,092 63,001 48,176 61,617 B. Reduced Form Price ·W ells2000 -0.425*** -0.326** -0.755** -0.886*** (0.14) (0.14) (0.30) (0.30) Observations 49,092 63,001 48,176 61,617 C. IV Estimates Wells -3.406*** -2.605** -6.142** -7.196*** (1.20) (1.18) (2.79) (2.77) Observations 49,092 63,001 48,176 61,617 Kleibergen-Paap F 656 684 688 717 Schools Included A B A B This table replicates the main results but using only either schools with date in at least the first two year of the study period (A) or schools with at least 6 years of data (B). School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 31 Table A2: Effect of Oil Exploitation on Individual Level Human Capital Measures Enroled Professional Degree Semesters to Enrolment A. OLS Wells -0.007** -0.007** -0.005 -0.005 0.174*** 0.187*** (0.00) (0.00) (0.00) (0.00) (0.04) (0.04) Observations 2,298,158 2,678,268 978,648 1,126,420 1,036,225 1,188,003 B. Reduced Form Price ·W ells2000 -0.002*** -0.002*** -0.004*** -0.004*** 0.051*** 0.053*** (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) Observations 2,298,158 2,678,268 978,648 1,126,420 1,036,225 1,188,003 C. IV Estimates Wells -0.020*** -0.019*** -0.035*** -0.036*** 0.462*** 0.483*** (0.01) (0.01) (0.01) (0.01) (0.10) (0.10) Observations 2,298,158 2,678,268 978,648 1,126,420 1,036,225 1,188,003 Kleibergen-Paap F 42,917 44,244 19,694 20,279 24,452 25,123 Schools Included A B A B A B This table replicates the main results but using only either schools with date in at least the first two year of the study period (A) or schools with at least 6 years of data (B). School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 32 Table A3: Effect of Oil Exploitation on Individual Level Human Capital Measures STEM Soc. Sci, Business, Law Others A. OLS Wells -0.040*** -0.039*** 0.034*** 0.035*** 0.007* 0.007* (0.00) (0.00) (0.01) (0.00) (0.00) (0.00) Observations 1,089,794 1,252,396 1,089,794 1,252,396 1,089,794 1,252,396 B. Reduced Form Price ·W ells2000 -0.009*** -0.009*** 0.008*** 0.008*** 0.000 0.000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Observations 1,089,794 1,252,396 1,089,794 1,252,396 1,089,794 1,252,396 C. IV Estimates Wells -0.079*** -0.076*** 0.075*** 0.074*** 0.002 0.001 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Observations 1,089,794 1,252,396 1,089,794 1,252,396 1,089,794 1,252,396 Kleibergen-Paap F 25,185 25,893 25185 25,893 25,185 25,893 Schools Included A B A B A B This table replicates the main results but using only either schools with date in at least the first two year of the study period (A) or schools with at least 6 years of data (B). School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 33 Table A4: Effect of Oil Exploitation on Individual Level Human Capital Measures Completion Desertion A. OLS Wells 0.010 0.009 -0.022 -0.021 (0.01) (0.01) (0.01) (0.01) Observations 490,821 532,689 490,821 532,689 B. Reduced Form Price ·W ells2000 -0.003 -0.004 0.001 0.002 (0.00) (0.00) (0.00) (0.00) Observations 490,821 532,689 490,821 532,689 C. IV Estimates Wells -0.037 -0.042 0.015 0.022 (0.03) (0.03) (0.03) (0.03) Observations 490,821 532,689 490,821 532,689 Kleibergen-Paap F 8311 8319 8311 8319 Schools Included A B A B This table replicates the main results but using only either schools with date in at least the first two year of the study period (A) or schools with at least 6 years of data (B). School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 34 Oil Suitability Robustness Tables Table A5: Effect of Oil Exploitation on Individual Level Human Capital Measures Enrolment Rate Professional Program Intensity t 1970 1980 1990 1970 1980 1990 A. OLS Wells -0.938 -0.938 -0.938 -1.603 -1.603 -1.603 (0.62) (0.62) (0.62) (1.55) (1.55) (1.55) Observations 68,701 68,701 68,701 66,635 66,635 66,635 B. Reduced Form Price ·W ells2000 -0.427*** -0.419*** -0.364** -0.854*** -0.836*** -0.892*** (0.14) (0.15) (0.15) (0.31) (0.31) (0.31) Observations 68,701 68,701 68,701 66,635 66,635 66,635 C. IV Estimates Wells -3.749*** -3.626*** -2.978** -7.578*** -7.313*** -7.403*** (1.33) (1.31) (1.24) (2.76) (2.74) (2.81) Observations 68,701 68,701 68,701 66,635 66,635 66,635 Kleibergen-Paap F 484 609 753 486 617 781 School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Table A6: Effect of Oil Exploitation on Individual Level Human Capital Measures Enroled Professional Degree Semesters to Enrolment t 1970 1980 1990 1970 1980 1990 1970 1980 1990 A. OLS Wells -0.007** -0.007** -0.007** -0.006 -0.006 -0.006 0.185*** 0.185*** 0.185*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.04) (0.04) (0.04) Observations 2,794,699 2,794,699 2,794,699 1,164,185 1,164,185 1,164,185 1,225,191 1,225,191 1,225,191 B. Reduced Form Price ·W ellst-0.002*** -0.002*** -0.002*** -0.004*** -0.004*** -0.004*** 0.054*** 0.052*** 0.055*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) Observations 2,794,699 2,794,699 2,794,699 1,164,185 1,164,185 1,164,185 1,225,191 1,225,191 1,225,191 C. IV Estimates Wells (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.10) (0.10) (0.10) Observations 2,794,699 2,794,699 2,794,699 1,164,185 1,164,185 1,164,185 1,225,191 1,225,191 1,225,191 Kleibergen-Paap F 39,601 47,438 49,656 15,468 19,478 22,627 19,053 24,080 28,006 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 35 Table A7: Effect of Oil Exploitation on Individual Level Human Capital Measures STEM Soc. Sci, Business, Law Others t 1970 1980 1990 1970 1980 1990 1970 1980 1990 A. OLS Wells -0.039*** -0.039*** -0.039*** 0.034*** 0.034*** 0.034*** 0.007** 0.007** 0.007** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 B. Reduced Form Price ·W ellst-0.008*** -0.008*** -0.008*** 0.008*** 0.008*** 0.008*** -0.000 0.000 0.000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 C. IV Estimates Wells -0.072*** -0.074*** -0.077*** 0.072*** 0.073*** 0.075*** -0.000 0.000 0.000 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 Kleibergen-Paap F 19,881 25,007 28,860 19,881 25,007 28,860 19,881 25,007 28,860 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Table A8: Effect of Oil Exploitation on Individual Level Human Capital Measures Completion Desertion Individual School Individual School t 1970 1980 1990 1970 1980 1990 1970 1980 1990 1970 1980 1990 A. OLS Wells 0.011 0.011 0.011 2.293 2.293 2.293 -0.022* -0.022* -0.022* -3.015 -3.015 -3.015 (0.01) (0.01) (0.01) (2.22) (2.22) (2.22) (0.01) (0.01) (0.01) (2.20) (2.20) (2.20) Observations 536,312 536,312 536,312 40,011 40,011 40,011 536,312 536,312 536,312 40,011 40,011 40,011 B. Reduced Form Price ·W ellst-0.003 -0.003 -0.003 -0.379 -0.366 -0.394 0.001 0.001 0.002 0.449 0.428 0.477 (0.00) (0.00) (0.00) (0.35) (0.35) (0.35) (0.00) (0.00) (0.00) (0.36) (0.36) (0.36) Observations 536,312 536,312 536,312 40,011 40,011 40,011 536,312 536,312 536,312 40,011 40,011 40,011 C. IV Estimates Wells -0.032 -0.032 -0.039 -4.334 -4.154 -4.347 0.015 0.014 0.019 5.137 4.863 5.269 (0.03) (0.03) (0.03) (3.92) (3.92) (3.94) (0.03) (0.03) (0.03) (4.08) (4.08) (4.04) Observations 536,312 536,312 536,312 40,011 40,011 40,011 536,312 536,312 536,312 40,011 40,011 40,011 Kleibergen-Paap F 5,962 7,463 9,331 461 540 530 5,962 7,463 9,331 461 540 530 School fixed effects; Time-Department quadratic trends; Landmines, age and gender included as control. Wells are standardized and Price is in logs. Shock assigned in year of enrolment in tertiary education. Sample trimmed to 2010. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 36 Instrument Threshold and Oil Wells life Cycle Robustness Tables Table A15: Effect of Oil Exploitation on Individual Level Human Capital Measures Enrolment Rate Professional Program Intensity t 1970 2000 1970 2000 A. OLS Wells -0.321 -0.321 -0.804 -0.804 (0.37) (0.37) (1.01) (1.01) Observations 68701 68701 66635 66635 B. Reduced Form Price ·W ellst-0.427*** -0.351** -0.854*** -0.859*** (0.14) (0.15) (0.31) (0.31) Observations 68701 68701 66635 66635 C. IV Estimates Wells -3.546*** -2.705** -7.205*** -6.811** (1.24) (1.17) (2.59) (2.74) Observations 68701 68701 66635 66635 Kleibergen-Paap F 396 254 388 255 School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Table A16: Effect of Oil Exploitation on Individual Level Human Capital Measures Enroled Professional Degree Semesters to Enrolment t 1970 2000 1970 2000 1970 2000 A. OLS Wells -0.002 -0.002 -0.003 -0.003 0.114*** 0.114*** (0.00) (0.00) (0.00) (0.00) (0.03) (0.03) Observations 2,794,699 2,794,699 1,164,185 1,164,185 1,225,191 1,225,191 B. Reduced Form Price ·Wellst-0.002*** -0.002*** -0.004*** -0.004*** 0.054*** 0.053*** (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) Observations 2,794,699 2,794,699 1,164,185 1,164,185 1,225,191 1,225,191 C. IV Estimates Wells -0.020*** -0.019*** -0.037*** -0.037*** 0.513*** 0.495*** (0.01) (0.01) (0.01) (0.01) (0.10) (0.10) Observations 2,794,699 2,794,699 1,164,185 1,164,185 1,225,191 1,225,191 Kleibergen-Paap F 23,870 14,556 9,909 6,610 12,587 8,414 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 43 Table A17: Effect of Oil Exploitation on Individual Level Human Capital Measures STEM Soc. Sci, Business, Law Others t 1970 2000 1970 2000 1970 2000 A. OLS Wells -0.016*** -0.016*** 0.013*** 0.013*** 0.005** 0.005** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 B. Reduced Form Price ·Wellst-0.008*** -0.009*** 0.008*** 0.008*** -0.000 0.000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 C. IV Estimates Wells -0.073*** -0.079*** 0.073*** 0.077*** -0.000 0.001 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Observations 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 1,293,718 Kleibergen-Paap F 12,628 8,429 12,628 8,429 12,628 8,429 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Table A18: Effect of Oil Exploitation on Individual Level Human Capital Measures Completion Desertion Individual School Individual School t 1970 2000 1970 2000 1970 2000 1970 2000 A. OLS Wells 0.007 0.007 1.264 1.264 -0.015* -0.015* -1.886 -1.886 (0.01) (0.01) (1.44) (1.44) (0.01) (0.01) (1.41) (1.41) Observations 536,312 536,312 40,011 40,011 536,312 536,312 40,011 40,011 B. Reduced Form Price ·Wellst-0.003 -0.003 -0.379 -0.367 0.001 0.002 0.449 0.460 (0.00) (0.00) (0.35) (0.35) (0.00) (0.00) (0.36) (0.36) Observations 536,312 536,312 40,011 40,011 536,312 536,312 40,011 40,011 C. IV Estimates Wells -0.025 -0.030 -3.279 -3.027 0.012 0.014 3.887 3.794 (0.02) (0.02) (2.96) (2.96) (0.02) (0.02) (3.08) (3.03) Observations 536,312 536,312 40,011 40,011 536,312 536,312 40,011 40,011 Kleibergen-Paap F 4,890 4,169 363 245 4,890 4,169 363 245 School fixed effects; Time-Department quadratic trends; Landmines, age and gender included as control. Wells are standardized and Price is in logs. Shock assigned in year of enrolment in tertiary education. Sample trimmed to 2010. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 44 Table A19: Effect of Oil Exploitation on Individual Level Human Capital Measures Test Scores Number of Students t 1970 2000 1970 2000 A. OLS Wells -0.522*** -0.522*** -0.369 -0.369 (0.10) (0.10) (0.41) (0.41) Observations 2771137 2771137 68701 68701 B. Reduced Reform v brent price -0.049 -0.049 -0.286 -0.323 (0.03) (0.03) (0.21) (0.24) Observations 2771137 2771137 68701 68701 C. IV Estimates Wells -0.461 -0.440 -2.373 -2.486 (0.33) (0.32) (1.77) (1.83) Observations 2771137 2771137 68701 68701 Kleibergen-Paap F 23728 14421 396 254 School fixed effects; Time-Department quadratic trends; Landmines included as control. Oil suitability and Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. Table A20: Formal Sector Wages OLS Reduced Form IV Estimates Wells 1.767*** . 11.945*** (0.32) . (1.94) Price ·Wells2000 . 0.487*** . . (0.08) . Enroled 1.646*** 1.645*** 1.650*** (0.04) (0.04) (0.04) Enroled ×Wells 0.082 . -0.034 (0.05) . (0.05) Enroled×Price ·Wells2000 . 0.014 . . (0.01) . Observations 1,044,133 1,044,133 1,044,133 Kleibergen-Paap F . . 1,437 School fixed effects; Time-Department quadratic trends; Landmines included as control. Wages are adjusted for the minimum wage. Wells are standardized and Price is in logs. Buffer of 10Km. Robust standard errors in parentheses. ***p<0.01, **p<0.05, *p<0.1. 45