Policy Research Working Paper 10378 Conscientiousness and Labor Market Returns Evidence from a Field Experiment in West Africa Mathias Allemand Martina Kirchberger Sveta Milusheva Carol Newman Brent Roberts Vincent Thorne Development Economics Development Impact Evaluation Group March 2023 Policy Research Working Paper 10378 Abstract Non-cognitive skills are increasingly recognized as import- intervention designed to affect conscientiousness-related ant determinants of labor market outcomes. To what extent skills. The findings show that treated workers were signifi- specific skills can be affected in adulthood remains an open cantly more likely to stay in their job and have higher wages question. This study conducted a randomized controlled nine months after the intervention. The findings suggest trial with low-skilled employed workers in Senegal, where that non-cognitive skills can be affected even later in the workers were randomly assigned to receive a training life cycle and can have substantial labor market returns. This paper is a product of the Development Impact Evaluation Group, Development Economics. 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 smilusheva@worldbank.org or Martina.Kirchberger@tcd.ie. 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 Conscientiousness and Labor Market Returns: Evidence from a Field Experiment in West Africa Mathias ALLEMAND1 Martina KIRCHBERGER2,∗ Sveta MILUSHEVA3 University of Zurich Trinity College Dublin and CEPR The World Bank Carol NEWMAN2,∗ Brent ROBERTS4 Vincent THORNE2 Trinity College Dublin University of Illinois at Urbana-Champaign Trinity College Dublin Keywords: non-cognitive skills, labor markets, conscientiousness. JEL Classification: J24, M53, O15. We are extremely grateful for the tireless efforts of the field team, which included Mame Diarra Bousso Sarr, Marion Sagot and Paola Elice. Many thanks to Aram Gassama, Meyhar Mohammed, Marine Nelly Cecile Colon de Franciosi and Meriem Fouad for excellent research support and assistance. We are grateful to CETUD and APIX for their support in coordinating and implementing this work. We are indebted to Etienne Thibault, Moustapha Gueye, Birame Olivier Sarr, Ahouseny Sawadogo, Niang Diallo, Gisler Isidore Mandiamy and their teams at Eiffage Rail Senegal (Groupement Eiffage CSE Yapi Merkezi). For helpful comments we thank Achyuta Adhvaryu and Chris Woodruff as well as participants at the 2021 Jobs and Development Conference and Trinity College Dublin. We gratefully acknowledge financial support from RSB and from the ieConnect for Impact program, which is funded by the United Kingdom’s Foreign, Commonwealth & Development Office. Computational reproducibility verified by DIME Analytics. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. All potential errors are our own. The list of authors is in alphabetical order. The trial was registered on the AEA RCT registry (AEARCTR-0004043 Version 2.0). 1 Department of Psychology, University of Zurich. 2 Department of Economics, Trinity College Dublin. 3 Development Impact Evaluation Department, The World Bank. 4 Department of Psychology, University of Illinois at Urbana-Champaign. ∗ Corresponding authors: martina.kirchberger@tcd.ie, cnewman@tcd.ie. 1. Introduction A large body of evidence highlights a link between the Big Five personality traits (agree- ableness, conscientiousness, extroversion, neuroticism and openness) and outcomes in the work place (Roberts et al., 2007; Soto, 2019; Borghans et al., 2008; Bowles et al., 2001; Fletcher, 2013; Nyhus and Pons, 2005).1 One such trait, conscientiousness, which relates to being punctual, hard working and responsible, emerges as a key trait for successful labor market outcomes, such as job performance and productivity (Barrick et al., 2001; Cubel et al., 2016; Dudley et al., 2006).2 It has long been thought that the Big Five personality traits are immutable in adulthood and not amenable to intervention efforts. However, recent theories in psychology propose that mindfulness and behavioral change techniques can help to shift beliefs and behaviors and modify skills to increase traits such as conscientiousness. This can be achieved by creating awareness about personality and its importance and by teaching behavior change skills (Allemand and Flückiger, 2017; Roberts et al., 2017a).3 To what extent interventions that target specific traits can be effective in improving labor market outcomes remains an open question. Between 230 million and 450 million new workers are expected to enter the labor force in Africa by 2030 (World Economic Forum, World Bank and African Development Bank, 2017; World Economic Forum, 2017). This places considerable pressure on young people to find and maintain jobs, particularly in the presence of substantial search and matching frictions and high rates of turnover documented in low-income countries (Abebe et al., 2021, 2020; Banerjee and Sequeira, 2020; Donovan et al., 2021). McKenzie (2021) highlights the potential for non-traditional training incorpo- rating psychology and focusing on essential non-cognitive skills to improve labor market outcomes. A number of recent studies have demonstrated positive impacts of comprehensive skills training programs, for example, on productivity (Adhvaryu et al., forthcoming) or earnings (Chioda et al., 2021). Soft-skills training programs could be important for jobs that require minimal skills but a high level of conscientiousness. This paper examines the impact of a targeted conscientiousness training intervention for low-skilled workers on employment, wages and job retention. To do this, we combine methods and insights from both economics and psychology. The training program is based on a recent conceptual intervention model, the Sociogenomic Trait Intervention Model 1 In these studies, the Big Five personality traits have been found to be predictors of income, long-term unemployment, job stability, job performance, job satisfaction, extrinsic and intrinsic career success and financial security. 2 Conscientiousness is a spectrum of constructs describing individual differences in the propensity to be self-controlled, responsible to others, hardworking, orderly, and rule abiding (Roberts et al. 2009; Roberts et al. 2014). It includes inter-related facets such as industriousness, organization, self-control, responsibility, persistence, decisiveness, conventionality, and punctuality. 3 For example, initial evidence from high-income contexts shows that psychological interventions using smartphone apps can help people to change personality traits in desired directions in adulthood (Stieger et al., 2020, 2021). 2 (STIM) by Roberts et al. (2017a), and a recent intervention program developed by Stieger et al. (2020) for Swiss participants, which we adapted to our particular context. Our sample consists of 386 workers employed at baseline in the construction of a new express train in Dakar, Senegal. The training consisted of an initial two-hour in-person session, conducted on company premises during work hours by a trained professional, external to the company, on how to be more conscientious at work, followed by a series of weekly short phone calls to workers over eight weeks reminding them of different ways to behave more conscientiously at work. Workers were randomly assigned to the treatment and control groups. The control group did not receive any additional training beyond the standard training protocols implemented by the company when new workers take up employment and periodically throughout the duration of their contract. The framework for the intervention builds on the assumption that a trait like conscientious- ness is a system of continuous and varying trait-related behaviors and experiences that can be manifest as a skill (i.e., the momentary, optimal expression of a trait). Hence, the main aim of the intervention is to help people change the behaviors and experiences associated with a domain of conscientiousness in a way that ensures that the change is enduring. This could be by showing certain conscientiousness-related behaviors and experiences more frequently or more intensively (for example, trying to become more punctual at work). To achieve this, the STIM uses behavioral activation theory, a form of cognitive behavior therapy used to treat de- pression (Lejuez et al., 2001; Magidson et al., 2014).4 The intervention aims to motivate and to activate the participants by changing and promoting conscientiousness-related behaviors and experiences. We target four important skill-based facets of conscientiousness that are specifically important in the context of work: industriousness, punctuality, responsibility and orderliness (Roberts et al., 2014). Changes in these skills may have positive effects on labor market outcomes. If participants demonstrate more conscientiousness-related behaviors and experiences, such as punctuality, responsibility, or orderliness, in a work context, and this behavior is observed and valued by their supervisors, they are more likely to be retained in their posts and see their improved performance reflected in higher wages.5 We find that receiving conscientiousness skills training increases job retention and wages nine months after the end of the intervention. Workers in the treatment group were 10% more likely to remain employed by the construction company and their last reported monthly earnings were about 40 USD higher than those of the control group (a 20% increase from average baseline earnings). We interpret these labor market impacts as “hard” evidence that 4 Behavioral activation to treat depression is a method to re-motivate and reactivate depressed patients. The conditions and consequences that trigger and maintain the depressive behavior must be identified and changed. Depressive behavior is replaced step-by-step with potentially rewarding activities. 5 Over the longer-term, the positive reinforcement that conscientiousness-related behaviors by workers are rewarded by supervisors could in turn encourage further conscientious behavior leading to a positive cycle that could lead to longer-term positive effects such as influencing whether someone is promoted or finds alternative employment. 3 our training had an impact. Examining shifts in self-reported measures of conscientiousness- related skills, we find respondents are more likely to find it easier to “work towards reaching one’s goals”, to “live up to your responsibilities”, to “focus on one’s most important goals” and to “fulfill one’s duties and obligations”. We treat these results as suggestive evidence that the training affected conscientiousness-related skills; the reliability of our scales of self-reported measures, while acceptable by conventional standards, is low due to high average reported response scores and/or other response biases. Our study therefore also highlights the difficulty of using standard questions to capture essential skills across very different contexts. The paper contributes to three strands of the literature. First, our paper contributes to a recent literature on the effects of various soft-skills training programs on labor market outcomes that finds substantial effects on individuals’ ability to maintain a job (Barrera-Osorio et al., 2021), returns to the firm with no effect on wages (Adhvaryu et al., forthcoming), effects on both individual earnings and firm profits (Chioda et al., 2021), positive effects on firm profits (Campos et al., 2017) or positive effects only for women (Acevedo et al., 2020). Groh et al. (2016), on the other hand, do not find effects on hours worked, income or employment.6 These programs typically train workers on a combination of skills, such as communication, writing, time management, negotiation, and emotional regulation. We add to this literature by providing evidence on the effectiveness of targeting one of the key traits for labor market outcomes (conscientiousness).7 A recent paper by Bryan et al. (2021) found that a theology education program which promoted evangelical Protestant Christian values had a positive impact on income that appears to be due to increased grit, which can be considered one facet of conscientiousness. Our experiment is the first, to our knowledge, to target different facets of conscientiousness and focuses on the impact on labor market outcomes. Second, we contribute to a recent literature that points out distinctive features of labor markets in low-income countries, such as the high prevalence of non-salaried work (Bandiera et al., 2022), significant search and matching frictions (Abebe et al., 2021, 2020; Banerjee and Sequeira, 2020) or the high turnover for low-earnings jobs (Donovan et al., 2021). Our evidence suggests that improving conscientiousness might play an important role in increasing job tenure. Third, we contribute to an ongoing debate about whether particular traits can be changed in adulthood and answer this question in the context of low-skilled workers in a low-income country (Allemand and Flückiger, 2017; Roberts et al., 2017a). Most of the current evidence 6 Aghion et al. (2019) further highlight the role of complementarities between soft skills of low-skilled workers and a firm’s other assets. 7 We also relate to a large literature on the effectiveness of worker training programs. For examples and overviews, see McKenzie (2017), McKenzie (2021), Alfonsi et al. (2020), Card et al. (2011) or Attanasio et al. (2011). 4 stems from research in high-income countries with high-skilled samples (Stieger et al., 2020, 2021). Our experiment allows us to examine whether conscientiousness-related skills can be activated in this particular setting using a simple and low-threshold intervention and also allows us to test the reliability of psychological interventions in vastly different contexts from which they were originally designed. The rest of the paper is organized as follows. In Section 2 we present the setting for our study, and describe the experiment and the data. Section 3 presents the results and discusses possible mechanisms and potential caveats to the study. Section 4 concludes. 2. Setting and field experiment Setting and background The study took place from April 2019 to May 2020. The setting for our experiment was Dakar, Senegal. Dakar’s workforce is, on average, aged 30 years old, 49% have attended middle school, and 72% are employed in low-skilled occupations or are self-employed (Agence Nationale de la Statistique et de la Démographie Sénégal and ICF, 2020). We worked with a company that forms part of a joint venture involved in the construction of a new 36 km long railway express train connecting the city center of Dakar with Diamniadio, a new city established outside Dakar, and in a second phase with the new international airport. The express train is one of the flagship projects of the government’s five-year strategic plan which highlighted the need for improving transport in the Greater Dakar Area. The first phase of construction started in 2017 and was completed in 2020. Field experiment Workers employed at the company responsible for the construction of the new express train in Dakar were recruited for the study. Having a sample of workers employed by the same company allows us to hold unobservable determinants of conscientiousness constant, such as how motivating supervisors are, company work atmospheres and management styles. To select our sample, the company provided us with a list of workers in low-skilled positions that were working on different portions of the tracks of the express train. There were a total of 386 workers on these lists, which formed our sample. Almost two-thirds of the sample were manual workers, barrier workers, and security agents. The remainder were other types of workers on the construction site that were classified by the company as having a low level of qualifications. Following recruitment to the study, we administered a baseline survey in April 2019 to collect information on worker characteristics and personality traits. After the baseline, half of the workers were randomly selected to receive the behavioral activation intervention 5 focused on conscientiousness. All workers selected for the intervention took part.8 The intervention aims to: (a) present information about the importance and benefits of the four important aspects of conscientiousness at work; (b) increase the motivation to change conscientiousness states; (c) provide instructions to activate conscientiousness states; and (d) prompt behavioral practice using reminders to activate the conscientiousness states.9 The intervention consisted of two components. The first component was a group training session with an average of 24 participants that lasted 2 hours, for eight groups in total. The length of the session and the number of participants were similar to other training sessions that workers received on safety and security. The session was hosted by a professional consultant with experience delivering training of this kind in Senegal.10 The training was conducted in the local language, Wolof, to ensure that workers fully understood all of the material. During the session, the trainer explained different concepts regarding non-cognitive skills to workers in the treatment group, in particular conscientiousness, with an emphasis on how improvements in such skills can lead to long-term benefits for the worker. The control group did not receive any additional training beyond the standard training provided by the company to all workers, which includes an initial security training at the start of the contract and periodic sessions throughout the duration of the project. The second component consisted of weekly reminders for eight weeks via short phone calls (less than 1 minute) to activate non-cognitive skills change. Calls were made by a survey company. Reminders were randomized across weeks and all workers received the same set of reminders each week. During each call, workers were given personal skills reminders, such as, “Make sure not to leave your place of work at any time without replacement”, that target conscientiousness. Phone calls were also conducted in Wolof. A mid-line survey was conducted in January 2020, and an end-line survey was conducted by phone in May 2020. Our sample started with the 386 individuals interviewed at baseline. At mid-line, 344 answered our survey (10.9% attrition), while at end-line we were able to reach 371 respondents (3.1% attrition).11 Despite the fact that this light touch intervention has a relatively low “dose”, it is expected that the repeated reminders function as triggers that initiate the conscientiousness states in the daily life of the workers and instigate change processes. Skill change can be best elicited through repeating behaviors that differ from typical, trait-like behavior (Allemand 8 It is possible that there were spill-over effects between the treatment and control groups (given that workers were assigned to different locations on the construction site depending on the company’s needs). This, however, would bias the results against finding an effect. 9 Full details of the content of the training session are provided in Appendix A. 10 The consultant has previously led several training programs in Senegal in personal development, conflict management, judicial defense and youth support. 11 Appendix Table B.1 compares baseline characteristics of attrition and non-attrition groups. We do not find any systematic differences between these groups after controlling for multiple hypothesis testing. We also do not find differences in attrition between the treatment and control groups. 6 and Flückiger, 2017; Roberts et al., 2017a; Wrzus and Roberts, 2017). The accumulation of conscientiousness-related behaviors and experiences should eventually lead to more habitual behaviors and experiences, and personality change through bottom-up processes. Demon- strating more conscientiousness-related behaviors in the workplace, such as punctuality, responsibility, or orderliness, is desirable from both the employer’s and the individual’s perspective. A more punctual and responsible worker is more likely to be retained by the company and receive a higher wage. This may in turn lead workers to feel that they are living up to the expectations placed on them, promoting even more conscientiousness-related behaviors and experiences in the future. Figure 1: Timeline of intervention Baseline sample: 386 Conscientiousness Baseline survey intervention • Treatment: 193 April 2019 · F2F • Control: 193 August 2019 · F2F End-line survey Mid-line survey Eight weekly phone call (371) (344) reminders May 2020 · Phone January 2020 · F2F August–October 2019 · Phone Note: This figure shows the timeline of the intervention for the treatment group. The number of respondents is in parenthesis. F2F indicates the survey or intervention was done face-to-face, while phone surveys are coded with Phone. See Appendix Figure A.1 for a more detailed version. Figure 1 summarizes the timeline of the intervention and surveys. Appendix Table B.2 shows some basic characteristics of the sample for treatment and control workers. The average age of participants in our study was 36-37 years.12 The vast majority were male and only around one-fifth had completed middle school. For most, this was not their first formal job and about one-tenth of the sample were recent migrants. We do not find any statistically significant difference in baseline traits or household-level variables across the treatment and control groups with the exception of worker age: treatment workers were slightly younger than control workers but this difference is only marginally statistically significant at conventional levels. Following Cumming (2014), we also report Cohen’s d measures and their 95% confidence intervals in the last columns of Appendix Table B.2 for the differences between treatment and control groups. We find that Cohen’s d are all smaller than 0.213 and that the confidence intervals include zero, suggesting that the differences between groups are not significantly 12 Our sample is slightly older than the general workforce in Dakar. This is due to the company recruiting workers with some prior experience. 13 This is the conventional cut-off value for Cohen’s d below which effects are considered small or very small (Sawilowsky, 2009). 7 different from each other. While some of the household-level baseline characteristics are different when looking at them individually, Cohen’s d is always below 0.2. Further, applying a correction for multiple hypothesis testing eliminates any statistical significance in group differences. We therefore conclude that our sample is balanced across the treatment and control groups on these baseline characteristics. Throughout the paper, we report our results with and without baseline controls. 3. Results and discussion This section presents our key results. We start by showing the effects of the intervention on the probability of being employed, the probability of still working at the construction company and on wages. We then discuss possible mechanisms and potential limitations of our study. 3.1. Main results Table 1 shows the results of a simple ordinary least squares regression of the following equation: yi , t =3 = β0 + β1 Ti + β2 X i , t =1 + i , t =3 (1) where yi , t =3 is the outcome of interest for worker i at end-line, Ti is an indicator for whether the worker is in the treatment or the control group, X i , t =1 are baseline individual and household controls which include sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings and household income, and i is a statistical noise term. We show the unconditional results in columns (1) and (3) and add the controls in columns (2) and (4).14 Columns (1) and (2) of Table 1 show that our treatment did not significantly increase the probability of employment at end-line.15 However, workers in the treatment group were about 10% more likely to keep their job in the construction company.16 These results are similar when we add baseline controls in column (4).17 18 14 To account for multiple hypothesis testing we also compute sharpened q-values which control for the false discovery rate. They are computed following Anderson (2008) and correct for multiple hypothesis testing within groups of outcomes. The results are presented in Appendix Table B.3. All of our results on labor market outcomes are robust to this correction. 15 At end-line, 72% of the sample were still employed and 51% were still employed at the construction company. 16 The magnitudes and significance of the treatment coefficients are robust to a logit specification. 17 We do not find any differences in the probability of remaining in the company by baseline occupations (see Appendix Table B.4). 18 About two-thirds of workers who left the company stayed in the construction sector, with an additional 15% of workers employed in wood and metal work, carpentry and craftsmanship (see Appendix Table B.5). 8 Table 1: Labor market outcomes at end-line (1) (2) (3) (4) Employed Employed Still at company Still at company Treated 0.0613 0.0580 0.103∗∗ 0.109∗∗ (0.187) (0.213) (0.048) (0.039) ∗∗∗ ∗∗∗ ∗∗∗ Constant 0.694 0.895 0.456 0.373∗∗ (0.000) (0.000) (0.000) (0.010) Standardized coeff. 0.137 0.130 0.206 0.217 Baseline controls N 371 371 368 368 R-sq 0.00472 0.0428 0.0106 0.0344 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. We next turn to last-month’s earnings at end-line.19 Column (1) in Table 2 shows that the treatment group received significantly higher earnings at end-line. Their monthly earnings rose by a sizeable amount of 22,545 CFA francs (41.51 in 2020 USD) compared to the control group.20 The positive impact on wages is robust to the inclusion of baseline controls in column (2). In column (3) we additionally control for earnings at baseline (baseline outcome) and the result holds, increasing slightly in magnitude. In light of the higher probability of retention within the company shown in Table 1, we next explore to what extent the earnings impacts are driven by keeping a job at the construction company by controlling for whether a worker is still employed by the company in column (4). Table 2 shows that workers employed by the construction company did earn significantly higher wages, which explains part of the result, but there was also an additional effect of the training of 15,759 CFAF (29.02 in 2020 USD) that workers earned, irrespective of where they were working at end-line. 19 Appendix C provides detailed information on how earnings are defined. 20 This represent a 20% increase from average baseline earnings. The average wage of workers at end-line was 84,146 CFAF (155.13 USD). 9 Table 2: Earnings at end-line (1) (2) (3) (4) Last earnings Last earnings Last earnings Last earnings Treated 22544.7∗∗∗ 23816.3∗∗∗ 24408.6∗∗∗ 15759.7∗∗ (0.005) (0.003) (0.001) (0.018) Still at company 73271.0∗∗∗ (0.000) ∗∗∗ ∗∗ Constant 72751.5 44534.9 21716.9 -1626.4 (0.000) (0.045) (0.304) (0.930) Standardized coeff. 0.289 0.306 0.313 0.202 Mean outcome control gr. 112644.92 112644.92 112644.92 112644.92 Baseline controls Baseline outcome N 370 370 370 367 R-sq 0.0210 0.0599 0.169 0.378 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household,total (formal and informal) household debt, total household savings, household income. Baseline outcome: earnings measured at baseline. 3.2. Mechanisms We now turn to potential mechanisms that could explain the improvements in the labor market outcomes that we observe. We consider three possible mechanisms. First, the training could have directly affected conscientiousness-related skills and this in turn affected labor market outcomes by making workers capable of being more responsible and hard-working and thereby more productive. Second, the training could have affected other qualities, such as agreeableness or openness, and these led to participants keeping their jobs. In turn, longer tenure allowed workers to acquire conscientiousness skills and these impacted their earnings. Third, it is possible that the firm kept track of who was trained and on the basis of this, retained workers and paid them higher wages. We discuss each of these potential channels in turn. First, to measure the direct effect of the intervention on conscientiousness, we measured conscientiousness-related skills before the training at baseline and after the training at mid-line using a 32-item questionnaire measuring skill-based versions of four facets of conscientiousness: responsibility, industriousness, organization and punctuality (Soto et al., 10 2022). Using face-to-face interviews, each respondent was asked to answer a set of questions on “how easy or difficult it is to” behave in a particular way relating to each of the scales. Responses were recorded using a five-point scale ranging from one (very difficult) to five (very easy). Details on the conscientiousness-related skill measures used are provided in Appendix D and Appendix Table D.1. While not perfect, the reliability metrics of the psychometric scales (Cronbach’s alpha and the interitem covariance, reported in Appendix Table D.2) are acceptable, and the reliability of the overall conscientiousness skill scale is satisfactory (Taber, 2018). Several factors might explain the imperfect scores of the scales: translation issues,21 cultural differences, social desirability bias or the fact that we asked the questions in an interview format rather than have respondents self-report. More generally, this highlights that psychometric surveys, often designed with university students in Western cultures in mind, may not be easily translatable to different contexts, especially in developing countries and for low-skilled workers. With these caveats in mind, we found that the training had an impact on four of the 32 conscientiousness items and report the results in Table 3. Specifically these are: how easy or difficult is it for you to (1) work towards reaching your goals; (2) focus on your most important goals; (3) live up to your responsibilities; and (4) fulfill your duties and obligations. We did not find a statistically significant effect on any of the other conscientiousness skills or the overall conscientiousness skill scale (see Appendix Tables D.3 to D.11). The high value of the constant term in all specifications indicates very high response scores, saturated close to their maximum value of five. This might be because respondents feel pressured into making positive judgements about themselves to appear socially desirable or signal their willingness to keep their job at the construction company. Such high baseline scores leave little room for change/improvement in these measures as a result of the intervention. Despite this, the results suggest that individuals who received the training report at least some increased levels of conscientiousness skills along some of the facets. Magnitudes of the effects range between a 0.16 point increase on the five-point scale for “focusing on one’s most important goals” and a 0.30 point increase for “living up to one’s responsibilities". These increases are robust to the inclusion of skill scores at baseline and the same set of controls as in Tables 1 and 2.22 We also explored other margins that could have been affected by 21 The measures we used were translated from English into French and then into the local language Wolof, which may affect reliability. 22 It should be noted that we lose statistical significance when we adjust the p-values using the sharpened q-values (see Appendix Table B.3). This is not surprising given the large number of traits considered. As such, these findings should be treated as providing suggestive evidence of the impact of the training on conscientiousness-related behaviors. We also performed a principal component analysis on the four facets of conscientiousness to reduce the dimensionality, extracting either the first principal component of each facet and then examining the impact of the treatment, or combining the first principal component of each facet 11 Table 3: Conscientiousness skills (1) (2) (3) (4) Reach goals Focus on goals Responsible Fulfill duty Treated 0.268∗ 0.162∗ 0.298∗ 0.162∗ (0.059) (0.081) (0.009) (0.011) ∗∗∗ ∗∗∗ ∗∗∗ Constant 3.941 4.666 4.030 4.566∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.202 0.192 0.281 0.274 Mean outcome control gr. 3.47 4.33 4.15 4.53 N 343 343 343 343 R-sq 0.074 0.028 0.084 0.063 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels. Personality skills are measured at baseline and midline. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. All specifications include baseline controls and skill measured at baseline. Baseline controls include sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Reach goals and Focus on goals both measure the industriousness skill, while Responsible and Fulfill duty measure the responsibility skill. Appendix Table D.3 presents specifications excluding the control variables. the treatment such as hours worked per week, having a second occupation or self-reported levels of absenteeism. We do not find that the treatment affected any of these margins.23 Second, an alternative mechanism is that the intervention affected other personality domains. For example, it could be that workers in the treatment group felt that the company had chosen them specifically to invest in. Because of this, they behaved more respectfully towards their supervisors and colleagues. This in turn raised their tenure which allowed them to acquire better levels of conscientiousness which was then reflected in their labor market outcomes. To examine this mechanism, we explored whether any of the other Big Five domains (agreeableness, extroversion, neuroticism and openness) shifted in significant ways. We do not find any clear evidence supporting this alternative mechanism (all coefficients are insignificant when we control for multiple hypothesis testing), although it should be noted that these scores also suffer from high baseline values. Further, we re-estimate the earnings equation (Table 2) and include an interaction term between remaining in the company and treatment (see Appendix Table B.6). The interaction term is not statistically significant into a summary measure. These measures similarly suffer from a high average value at baseline and lack of variation in the measured traits leading to very high p-values. 23 Hours worked are already quite high at baseline, with a mean of 61.4 hours worked and a standard deviation of 11.3 making it difficult for workers to increase hours or take a second job. Further, most workers employed at the company take the company bus to the construction site and back home, so that there is not much scope for an individual worker to adjust their hours worked. 12 suggesting that the effect of training was the same whether an individual is employed by the construction company or not. This suggests that the effects are not working exclusively through longer tenure at the company. Third, we cannot rule out that the company kept track of who we trained and kept workers because they knew they were trained, although this information was not shared with the company. While this could contribute to the employment retention effects, positive wage effects for workers who left the company would not be driven by this. This suggests that our results are not totally driven by the company keeping track of trained workers. Moreover, observations from the field do not support this hypothesis. Supervisors tend to be in close contact with their workers, and closely monitor their behavior. Given how important it is for the company to be able to rely on workers to perform responsibly in these roles, it is unlikely that the company would not take into account the actual performance of workers in their performance evaluations. Overall, the evidence is supportive of a mechanism whereby our intervention affected worker conscientiousness and this in turn affected their labor market outcomes, although more research is needed to better measure conscientiousness skills in such contexts. 3.3. Potential limitations We note two limitations to our study. First, end-line data collection took place after the onset of the COVID-19 pandemic and it is possible that our intervention interacted with the pandemic. For example, given that the training aimed to make individuals more con- scientious, it is possible that they were more careful during the pandemic and lost their jobs due to wanting to stay at home rather than run the risk of exposure to the virus by going to work. On the other hand, more conscientious individuals might have felt a greater responsibility to go to work. To measure whether the intervention affected the impact of COVID-19 on the behavior of the workers, the end-line survey asked a number of questions on how the pandemic impacted their livelihoods. One set of questions asked respondents to compare current incomes and expenditures to before the start of the pandemic, specifically in February 2020. We do not find any significant differences in the extent to which treated and control individuals reported lower incomes, lower transfers and gifts, higher health expenditures, lower savings or more borrowing (see Appendix Tables E.1 and E.2).24 We find that individuals in the control group reported having taken more measures to prevent the spread of the virus but do not find any differences in the rate at which treatment and control individuals displayed symptoms, or lost their main source of income as a result of COVID-19. Overall, this suggests that the intervention did not interact with the pandemic in a significant way. 24 We only find a small statistically significant difference for education expenditures, which is higher for the control group. 13 A second concern relates to the lack of an active control group. For the first part of the intervention, the training session, it is worth noting that general security training sessions were common for all workers at the company. Workers were brought together in similarly sized groups to the conscientiousness training session for these security training sessions. For the phone reminders, providing messages that were different in content to the conscientiousness messages would constitute a different treatment arm rather than an active control group. Moreover, reminders relating to different aspects of the general security training that all workers received would look similar to the phone reminders that the treatment group received. As such, an active control group was not included in the study. Finally, it is worth mentioning that our results capture relatively medium-term impacts. While beyond the scope of this paper, future research should also seek to measure longer-term impacts. 4. Conclusion We used a randomized controlled trial to test the impact of a conscientiousness training intervention for low-skilled workers in an urban developing country setting on employment, wages and job retention. The training program was based on recent work in the field of psychology (Roberts et al., 2017a; Stieger et al., 2020) and was designed to affect conscientiousness traits among workers. We found that providing conscientiousness training significantly affected the probability that the workers were still employed by the company at end-line and their level of earnings. To our knowledge, this is one of the first studies that tests a psychological intervention of this kind in the field. The developing country context for the study is particularly relevant given the importance of job creation and the problem of job turnover in these settings. Our study highlights the potential for psychological training of this kind to improve labor market outcomes for low-skill workers. By making our training materials fully available, we hope to facilitate further research on this topic. Furthermore, while we find some suggestive evidence that the training affected the consci- entiousness skills of the treated workers, our study also underlines the challenges associated with using standard, self-reported measures to capture personality skills across different contexts, with different types of populations, in different languages and across different levels of literacy. Future research that provides guidance on the appropriate instruments to use in settings outside university campuses is vital for being able to pin down the precise channels through which interventions aimed at affecting non-cognitive skills in such contexts work. 14 References ABEBE, G., A. S. CARIA, M. FAFCHAMPS, P. FALCO, S. FRANKLIN, AND S. QUINN (2021): “Anonymity or distance? Job search and labour market exclusion in a growing African city,” The Review of Economic Studies, 88, 1279–1310. ABEBE, G., S. CARIA, M. FAFCHAMPS, P. FALCO, S. FRANKLIN, S. QUINN, AND F. SHILPI (2020): “Matching Frictions and Distorted Beliefs: Evidence from a Job Fair Experiment,” Depart- ment of Economics, Oxford University (mimeo). ACEVEDO, P., G. CRUCES, P. GERTLER, AND S. MARTINEZ (2020): “How job training made women better off and men worse off,” Labour Economics, 65, 101824. ADHVARYU, A., N. KALA, AND A. NYSHADHAM (forthcoming): “Returns to On-the-job Soft Skills Training,” Journal of Political Economy. AGENCE NATIONALE DE LA STATISTIQUE ET DE LA DÉMOGRAPHIE SÉNÉGAL, AND ICF (2020): “Senegal Demographic and Health Survey 2019 [Dataset].” AGHION, P., A. BERGEAUD, R. BLUNDELL, AND R. GRIFFITH (2019): “The innovation premium to soft skills in low-skilled occuptions,” CEPR Discussion Paper 14102, C.E.P.R. Discussion Papers. ALFONSI, L., O. BANDIERA, V. BASSI, R. BURGESS, I. RASUL, M. SULAIMAN, AND A. VITALI (2020): “Tackling youth unemployment: Evidence from a labor market experiment in Uganda,” Econometrica, 88, 2369–2414. ALLEMAND, M., AND C. FLÜCKIGER (2017): “Changing personality traits: Some considerations from psychotherapy process-outcome research for intervention efforts on intentional personality change.,” Journal of Psychotherapy Integration, 27, 476. ANDERSON, M. L. (2008): “Multiple inference and gender differences in the effects of early intervention: A reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects,” Journal of the American Statistical Association, 103, 1481–1495. ATTANASIO, O., A. KUGLER, AND C. MEGHIR (2011): “Subsidizing Vocational Training for Disadvantaged Youth in Colombia: Evidence from a Randomized Trial,” American Economic Journal: Applied Economics, 3, 188–220, 10.1257/app.3.3.188. BANDIERA, O., A. ELSAYED, A. SMURRA, AND C. ZIPFEL (2022): “Young adults and labor markets in Africa,” Journal of Economic Perspectives, 36, 81–100. BANERJEE, A. V., AND S. SEQUEIRA (2020): “Spatial mismatches and imperfect information in the job search,” CEPR Discussion Paper No. DP14414. 15 BARRERA-OSORIO, F., A. D. KUGLER, AND M. I. SILLIMAN (2021): “Hard and soft skills in vocational training: Experimental evidence from Colombia,” NBER Working Paper Number 27548, National Bureau of Economic Research. BARRICK, M. R., M. K. MOUNT, AND T. A. JUDGE (2001): “Personality and performance at the beginning of the new millennium: What do we know and where do we go next?” International Journal of Selection and Assessment, 9, 9–30. B ORGHANS, L., A. L. DUCKWORTH, J. J. HECKMAN, AND B. TER WEEL (2008): “The economics and psychology of personality traits,” Journal of Human Resources, 43, 972–1059. B OWLES, S., H. GINTIS, AND M. OSBORNE (2001): “Incentive-Enhancing Preferences: Per- sonality, Behavior, and Earnings,” American Economic Review, 91, 155–158. BRYAN, G., J. J. CHOI, AND D. KARLAN (2021): “Randomizing religion: the impact of Protestant evangelism on economic outcomes,” The Quarterly Journal of Economics, 136, 293–380. CAMPOS, F., M. FRESE, M. GOLDSTEIN, L. IACOVONE, H. C. JOHNSON, D. MCKENZIE, AND M. MENSMANN (2017): “Teaching personal initiative beats traditional training in boosting small business in West Africa,” Science, 357, 1287–1290. CARD, D., P. IBARRARÁN, F. REGALIA, D. ROSAS-SHADY, AND Y. SOARES (2011): “The labor market impacts of youth training in the Dominican Republic,” Journal of Labor Economics, 29, 267–300. CHIODA, L., D. CONTRERAS-LOYA, P. GERTLER, AND D. CARNEY (2021): “Making entrepreneurs: Returns to training youth in hard versus soft business skills,” Working Paper 28845, National Bureau of Economic Research. CUBEL, M., A. NUEVO -CHIQUERO, S. SANCHEZ-PAGES, AND M. VIDAL-FERNANDEZ (2016): “Do personality traits affect productivity? Evidence from the laboratory,” Economic Journal, 126, 654–681. CUMMING, G. (2014): “The new statistics: Why and how,” Psychological science, 25, 7–29. DONOVAN, K., W. J. LU, T. SCHOELLMAN ET AL . (2021): “Labor market dynamics and devel- opment,” Federal Reserve Bank of Minneapolis Staff Report No. 596. DUDLEY, N. M., K. A. ORVIS, J. E. LEBIECKI, AND J. M. CORTINA (2006): “A meta-analytic investigation of conscientiousness in the prediction of job performance: examining the intercorrelations and the incremental validity of narrow traits.,” Journal of Applied Psy- chology, 91, 40. 16 FLETCHER, J. M. (2013): “The effects of personality traits on adult labor market outcomes: Evidence from siblings,” Journal of Economic Behavior and Organization, 89, 122–135. GROH, M., N. KRISHNAN, D. J. MCKENZIE, AND T. VISHWANATH (2016): “The impact of soft skills training on female youth employment: Evidence from a randomized experiment in Jordan,” IZA Journal of Labor and Development. LEJUEZ, C. W., D. R. HOPKO, AND S. D. HOPKO (2001): “A brief behavioral activation treatment for depression: Treatment manual,” Behavior Modification, 25, 255–286. MAGIDSON, J. F., B. W. ROBERTS, A. COLLADO -RODRIGUEZ, AND C. LEJUEZ (2014): “Theory- driven intervention for changing personality: expectancy value theory, behavioral activa- tion, and conscientiousness.,” Developmental Psychology, 50, 1442. MCKENZIE, D. (2017): “How effective are active labor market policies in developing coun- tries? a critical review of recent evidence,” The World Bank Research Observer, 32, 127–154. MCKENZIE, D. J. (2021): “Small business training to improve management practices in developing countries: Reassessing the evidence for "training doesn’t work",” Oxford Review of Economic Policy, 37, 276–301. NYHUS, E. K., AND E. PONS (2005): “The effects of personality on earnings,” Journal of Economic Psychology, 26, 363–384. ROBERTS, B. W., P. L. HILL, AND J. P. DAVIS (2017a): “How to change conscientiousness: The sociogenomic trait intervention model.,” Personality Disorders: Theory, Research, and Treatment, 8, 199. ROBERTS, B. W., N. R. KUNCEL, R. SHINER, A. CASPI, AND L. R. GOLDBERG (2007): “The power of personality: The comparative validity of personality traits, socioeconomic status, and cognitive ability for predicting important life outcomes,” Perspectives on Psychological science, 2, 313–345. ROBERTS, B. W., C. LEJUEZ, R. F. KRUEGER, J. M. RICHARDS, AND P. L. HILL (2014): “What is conscientiousness and how can it be assessed?” Developmental Psychology, 50, 1315. ROBERTS, B., J. JACKSON, J. FAYARD, G. EDMONDS, AND J. MEINTS (2009): “Conscientious- ness,” in Handbook of individual differences in social behavior ed. by Leary, M., and Hoyle, R. New York: Guilford Press, 369–381. SAWILOWSKY, S. S. (2009): “New effect size rules of thumb,” Journal of Modern Applied Statistical Methods, 8, 26. 17 SOTO, C. J. (2019): “How replicable are links between personality traits and consequential life outcomes? The life outcomes of personality replication project,” Psychological Science, 30, 711–727. SOTO, C. J., C. NAPOLITANO, M. SEWELL, H. YOON, AND B. ROBERTS (2022): “An integrative framework for conceptualizing and assessing social, emotional, and behavioral skills: The BESSI,” Journal of Personality and Social Psychology, DOI: 10.1037/pspp0000401. STIEGER, M., C. FLÜCKIGER, D. RÜEGGER, T. KOWATSCH, B. W. ROBERTS, AND M. ALLE- MAND (2021): “Changing personality traits with the help of a digital personality change intervention,” Proceedings of the National Academy of Sciences, 118. STIEGER, M., S. WEPFER, D. RÜEGGER, T. KOWATSCH, B. W. ROBERTS, AND M. ALLEMAND (2020): “Becoming more conscientious or more open to experience? Effects of a two-week smartphone-based intervention for personality change,” European Journal of Personality, 34, 345–366. TABER, K. S. (2018): “The use of Cronbach’s alpha when developing and reporting research instruments in science education,” Research in Science Education, 48, 1273–1296. WORLD ECONOMIC FORUM (2017): “The future of jobs and skills in Africa: Preparing the region for the fourth industrial revolution,” World Economic Forum Executive Briefing. WORLD ECONOMIC FORUM, WORLD BANK AND AFRICAN DEVELOPMENT BANK (2017): “The Africa Competitiveness Report 2017: Addressing Africa’s demographic dividend,” World Economic Forum Insight Report. WRZUS, C., AND B. W. ROBERTS (2017): “Processes of personality development in adulthood: The TESSERA framework,” Personality and Social Psychology Review, 21, 253–277. 18 Appendix (For Online Publication) A. Details of the Experiment A.1. Curriculum of the psychological intervention 1. Introduction and goals of the group training session Duration: 5 min • Introduction to group training session and warm-up • Present outline and the goals of the session • Present expectations (e.g., willingness to participate in group discussions) 2. Present information about the importance and benefits of conscientiousness at work Duration: max. 25 min 2.1. Present facts about conscientious behaviors at work Present information about factors and skills that can make people more successful or less successful at work. On the one hand, research has shown that skills [“keeping one’s work place clean” “following through with commitments”] and experiences [setting and achieving goals] are important for success at work and satisfaction with the job. On the other hand, research has also shown that a lack of skills or willingness to learn the skills is related to job-related difficulties [being repeatedly late to work]. This group session will introduce and discuss the importance and benefits of four important skills in the context of work and beyond. Present definitions of the four “non-cognitive” skills (i.e., industriousness, punctuality, responsibility, and orderliness (Roberts et al., 2007); [please simplify the definitions to fit the needs of the participants]: • Industriousness captures the tendencies to work hard, aspire to excellence, and persist in the face of challenges. • Punctuality reflects the simple tendency to show up on time to previously scheduled appointments. Being punctual appears important when considering one’s ability to plan, work hard to get somewhere, avoid temptations that might lead one to be late, care enough to meet other people on time, and understand the rules and conventions surrounding one’s social group. 19 • Responsibility: On the high end of the spectrum, responsibility reflects the tendency to follow through with promises to others and follow rules that make social groups work more smoothly. On the low end, it reflects the tendency to be an unreliable partner in achievement settings and to break one’s promises. • Orderliness encompasses the overarching tendency to be “prepared”, which includes tendencies toward neatness, cleanliness, and planfulness on the positive side, or disorderliness, disorganization, and messiness on the negative end of the spectrum. 2.2. Transfer of the definitions to real-life work context Transfer the definitions to the everyday work life contexts of the participants to make them more comprehensible and realistic to participants [provide specific examples and scenarios from the job as barrier workers to motivate the discussion]. • What does it specifically mean to be hardworking at my workplace/in my job? (e.g., invest efforts into work; complete tasks; help site vehicles to cross; observe all rules; secure train passages; remove all stones etc.; monitor equipment; take tours through the allocated area; work as hard as everyone else on the site) • What does it specifically mean to be punctual at my workplace/in my job? (e.g., show up on time; be on time in the morning at the bus pick-up point; be punctual for meetings with supervisor) • What does it specifically mean to be responsible at my workplace/in my job? (e.g., to follow work-related rules; consult the supervisor in case of a delay or absences; inform the supervisor in case of a problem) • What does it specifically mean to be orderly at my workplace/in my job? (e.g., keep the workplace in order; keep the torch clean in the building barracks, workplace or surroundings of the workplace; do not leave your own uniform e.g., helmets, vests, etc. unattended) 3. Increase the motivation to change conscientiousness Duration: max. 20 min 3.1. Discuss possible rewards of being conscientious Discuss real or hypothetical situations/scenarios in the everyday work life of the participants, showing how certain skills lead to better work outcomes. The goal of this task is to motivate participants’ behaviors. It should be made clear what participants would gain when they habitually show conscientious behaviors: 20 • What are possible positive consequences or rewards (for the individual, for others, for the workplace) of being hardworking at the workplace? [getting a raise; helping the company complete the work better or faster] • What are possible consequences or rewards (for the individual, for others, for the workplace) of being punctual at the workplace? [not losing your job; getting another job with the same company] • What are possible consequences or rewards (for the individual, for others, for the workplace) of being responsible at the workplace? [helping the entire organization to be better thus making the company successful leading to keeping your job and future jobs] • What are possible consequences or rewards (for the individual, for others, for the workplace) of being orderly at the workplace? [avoiding accidents and not hurting yourself or others] 3.2. Discuss possible costs of low conscientiousness Discuss real or hypothetical situations/scenarios in the everyday work life of the participants, showing how missing or poorly trained skills leads to poor work outcomes. The goal of this task is to motivate participants’ behaviors by pointing to possible costs [please provide realistic examples and scenarios]: • What are possible negative consequences or costs (for the individual, for others, for the workplace) of being lazy at the workplace? [less likely to receive a recommendation for another job; less likely to be given other opportunities] • What are possible negative consequences or costs (for the individual, for others, for the workplace) of being unpunctual at the workplace? [making other coworkers in a previous shift have to stay late] • What are possible negative consequences or costs (for the individual, for others, for the workplace) of being unreliable at the workplace? [slow-down work for everyone; less likely to be asked to continue working for the company] • What are possible negative consequences or costs (for the individual, for others, for the workplace) of being disorganized and messy at the workplace? [higher likelihood of hurting yourself or someone else in the workplace] 4. Provide instruction on how to activate conscientious behaviors Duration: max. 40 min 21 4.1. Behavioral activation tasks / behavioral “experiments” in daily work life To pursue their specific goals and to perform the four important skills in the context of work, participants have to practice and repeat the goal-related behaviors in the same work context repeatedly so that the contexts may elicit the behaviors [the specific behavioral activation tasks / “behavioral experiments” based on the list in the appendix can be used for the discussion]. 4.2. Possible barriers to perform the four important skills Discuss possible barriers that may hinder participants to perform the skills (e.g., individual barriers, work-related barriers). • What are possible barriers to being hardworking at the workplace? How can these barriers be overcome? • What are possible barriers to being punctual at the workplace? How can these barriers be overcome? • What are possible barriers to being responsible at the workplace? How can these barriers be overcome? • What are possible barriers to being orderly at the workplace? How can these barriers be overcome? 4.3. Possible resources to perform the four important skills Discuss possible resources that may help participants to perform the skills (e.g., social support). • What are possible resources for being hardworking at the workplace? How can these resources be used? • What are possible resources for being punctual at the workplace? How can these resources be used? • What are possible resources for being responsible at the workplace? How can these resources be used? • What are possible resources for being orderly at the workplace? How can these resources be used? 4.4. Identification of goals and goal setting Participants specify one specific goal for each of the four skills (i.e., industriousness, punctu- ality, responsibility, and orderliness) they would like to pursue in the next weeks. For each 22 of the four specific goals, participants have to generate one specific goal in the form of an if-then plan (e.g., “If I see a car right in front of the barrier, then I check all the rules before I let the car pass” [industriousness]; “If I have a meeting with my supervisor, then I will do everything I can to be there on time” [punctuality], “If I want to leave the workplace, then I will ask the supervisor first” [responsibility]; “If I start to work, then I will keep things tidy” [orderliness]) [please provide simple examples from the daily life work context]. This task also may help to identify specific situations in which participants can perform specific goal-related behaviors: What are typical work-related situations (“if”) in which participants perform the behaviors and the four skills, respectively (“then”)? The goal setting process should follow the SMART criteria: • Specific: Goals must be clearly defined (not vague, but as precise as possible). • Measurable: Goals must be measurable. • Attainable/attractive: The goals must be attainable and desirable for the person. • Realistic: The goal set must be possible and feasible. • Timely: It must be possible to set a reasonable time limit to achieve the goal including time limits for smaller steps. Behavior Activation through Reminders Prompt behavioral practice by weekly reminders to activate conscientiousness behav- iors Procedure Every week each participant will get a short phone call with 4 reminders targeted toward each of the four skills. For each phone call, 4 behavioral activation tasks / “behavioral experiments” (1 task per facet of conscientiousness) can be randomly [or, sequentially] selected from the list below. Appendix: Behavioral Activation Tasks / “Behavioral Experiments” Here is a set of broad and specific behavioral activation tasks / “behavioral experiments” [this list can be expanded with more specific tasks depending on the participants’ daily work life]. Industriousness Broad reminders • Try to have high standards and work toward them. 23 • Try to go above and beyond of what is required. • Try to work as hard as the majority of people around you. • Try to give the highest quality in everything you do. • Try to do more than what is required. • It’s important to set goals and achieve them. • Complete the tasks you start. • Persist at tasks after meeting setbacks or failures. • Try to work extra hard on a project to make sure that it is done right. • Complete the projects you start. • Finish what you start. • Put your mind on the task at hand. • Get things done quickly. • Always know what you are doing. • Do not let yourself get distracted. • Do not postpone decisions. • Finish what you start, e.g., checking a vehicle, cleaning a rail, touring through an area for monitoring, unloading a truck. • Get things done quickly. • Do not let yourself get distracted, e.g., by another vehicle while checking one vehicle. Specific reminders • Help site vehicles to cross the rail: – Try to give the highest quality in everything you do, e.g., check all the rules before you let a car pass (headlights on, rear red lights on, no children, no people in the load area, the car has a numbered red badge, etc.) – Put your mind on the task at hand, e.g., when checking a vehicle crossing the rail. 24 • Secure train passages: – Try to give the highest quality in everything you do, e.g., check that there are no stones, sand or holes in the rail at all. – Put your mind on the task at hand, e.g., when a train is arriving, and you prevent pedestrians and vehicles from passing. • Monitoring equipment: – Put your mind on the task at hand, e.g., touring through an area to monitor small railway equipment. • Unloading trucks: – Try to give the highest quality in everything you do, e.g., try to unload the equipment of a truck as properly as you can. – Put your mind on the task at hand, e.g., when you are unloading a truck. Punctuality • Do not forget meetings. • Keep up with required work. • Get to appointments with your supervisor on time. • Do not miss the bus; be at the picking up point on time. • Return phone calls in timely fashion. Responsibility • Try to carry out your obligations to best of your ability. • Go out of your way to keep your promises. • If you are running late, call ahead to notify those who are waiting for you. • If you are running late, call ahead and inform the supervisor. • If you want to leave the workplace, ask the supervisor first. • Unloading trucks: In event of a problem, register the truck number and call the supervisor. 25 • If you need to call in sick, do so before your shift so your supervisor can get a replace- ment. Orderliness • Keep your environment neat in order to work well. • Be organized. • Do not forget to put things back in their proper place. • Clean up the surroundings of your workplace. • Do not leave a mess in your work place. • Tidy up your work place. • Put things back in their proper place. • Do not leave your belongings unattended, e.g., orange project jacket, helmet. • Keep things tidy. • See that rules are observed, e.g., do not make fire around the workstation, do not wear earphones on your ears at the workstation, wear your mandatory equipment. 26 A.2. Detailed timeline Figure A.1: Detailed timeline of the intervention and surveys, including phone call reminders’ contents Baseline sample: 386 Conscientiousness Baseline survey • Treatment: 193 training April 2019 · F2F • Control: 193 August 2019 · F2F Phone call reminder week 2 • "If you are running late, call ahead and Phone call reminder week 1 inform the supervisor" • "Get to appointments with your • "Do not forget meetings" supervisor on time" • "Be organized" • "Try to go above and beyond what is • "Finish what you start" required" • "Go out of your ways to keep your • "Keep things tidy. You need to keep promises" your environment neat in order to work well" Phone call reminder week 4 Phone call reminder week 3 • "It is important to set goals and • "Make sure to wear the mandatory achieve them. Complete the tasks that equipment. it is important that the you start" rules of the workplace are observed" • "Do not miss the bus. Be at the pick-up • "Do not let yourself get distracted" point on time" • "Try to carry out your obligations to • "If you need to call in sick, do so before the best of your ability" your shift so your supervisor can get a replacement" • "Keep up with the required work" • "Do not forget to put things back in their proper place" Phone call reminder week 6 Phone call reminder week 5 • "Persist at tasks after meeting setbacks • "Return phone calls in a timely or failures" fashion" • "Keep up with the required work" • "Try to work extra hard on a task to make sure that it is done right" • "If you are running late, call ahead to notify those who are waiting for you" • "Do not leave a mess in your work place" • "Clean up the surroundings of your work place" • "If you need to leave the workplace, ask the supervisor first" Phone call reminder week 8 Phone call reminder week 7 • "Try to give the highest quality in • "Put things back in their proper place" everything you do" • "Put your mind on the task at hand" • "Get to appointments on time" • "Carefully follow the instructions • "Try to carry out your obligations to given to you by your supervisor" the best of your ability" • "Do not forget meetings" • "Do not leave your belongings unattended, for example your helmet, safety jacket" End-line survey Mid-line survey (371) (344) May 2020 · Phone January 2020 · F2F Note: This Figure expands on Figure 1 and shows the main conscientiousness reminders for each follow-up phone call. The number of respondents is in parenthesis. F2F indicates the survey or intervention was done face-to-face, while phone surveys are coded with Phone. 27 B. Additional tables Table B.1: Sample characteristics at baseline for attrition vs non-attrition groups (1) (2) (3) t-test Cohen’s D Non-attrition Attrition Total Difference Variable N Mean/SE N Mean/SE N Mean/SE (1)-(2) Effect size 95% CI Individual-level variables Last earnings 332 1.11e+05 54 1.16e+05 386 1.12e+05 -4475.682 -0.066 -0.353 0.222 [3228.205] [14796.793] [3453.839] Female 332 0.096 54 0.019 386 0.085 0.078* 0.279 -0.009 0.567 [0.016] [0.019] [0.014] Age 332 36.352 54 37.296 386 36.484 -0.944 -0.105 -0.393 0.183 [0.483] [1.366] [0.457] Middle school 332 0.244 54 0.111 386 0.225 0.133** 0.319 0.030 0.607 [0.024] [0.043] [0.021] First formal job 330 0.124 38 0.053 368 0.117 0.072 0.223 -0.113 0.559 [0.018] [0.037] [0.017] Recent migrant 325 0.098 37 0.054 362 0.094 0.044 0.152 -0.188 0.492 [0.017] [0.038] [0.015] Reach goals 332 3.518 54 3.370 386 3.497 0.148 0.101 -0.187 0.388 [0.079] [0.220] [0.075] Focus on goals 332 4.398 54 4.315 386 4.386 0.083 0.086 -0.202 0.374 [0.052] [0.137] [0.049] Up to responsabilities 332 4.220 54 3.926 386 4.179 0.294* 0.274 -0.015 0.562 [0.057] [0.171] [0.055] Fulfill duty 332 4.503 54 4.593 386 4.516 -0.090 -0.108 -0.396 0.179 [0.046] [0.107] [0.042] Household-level variables # Beds in household 332 4.792 54 5.185 386 4.847 -0.393 -0.145 -0.433 0.143 [0.149] [0.369] [0.138] # Children in household 332 4.452 54 4.500 386 4.459 -0.048 -0.013 -0.301 0.274 [0.203] [0.395] [0.183] # Adults in household 332 6.349 54 6.889 386 6.425 -0.539 -0.141 -0.429 0.146 [0.203] [0.610] [0.194] Household debt 332 26731.928 54 37962.963 386 28303.109 -1.12e+04 -0.193 -0.481 0.095 [3064.672] [9661.929] [2963.524] Household savings 332 83734.940 54 48148.148 386 78756.477 35586.792** 0.350 0.061 0.638 [5752.068] [10876.423] [5209.624] Household income 332 53448.795 54 49351.852 386 52875.648 4096.943 0.140 -0.147 0.428 [1631.964] [3439.215] [1484.070] Note: The Attrition group includes all individuals who missed at least one round (either the mid-line or end-line survey, or both). Recent migrant is defined as living for three years or less in current location and moved from outside Dakar. Middle school indicates whether an individual has completed middle school. Applying multiple hypothesis correction, any statistical significance in group difference disappears. 28 Table B.2: Sample characteristics at baseline (1) (2) (3) t-test Cohen’s D Control Treatment Total Difference Variable N Mean/SE N Mean/SE N Mean/SE (1)-(2) Effect size 95% CI Individual-level variables Last earnings (in th.) 193 112.645 193 111.230 386 111.937 1.415 0.021 -0.179 0.220 (5.618) (4.034) (3.454) Female 193 0.088 193 0.083 386 0.085 0.005 0.018 -0.181 0.218 (0.020) (0.020) (0.014) Age 193 37.326 193 35.642 386 36.484 1.684* 0.188 -0.012 0.388 (0.644) (0.645) (0.457) Middle school 193 0.212 193 0.238 386 0.225 -0.026 -0.062 -0.261 0.138 (0.030) (0.031) (0.021) First formal job 182 0.110 186 0.124 368 0.117 -0.014 -0.043 -0.247 0.162 (0.023) (0.024) (0.017) Recent migrant 179 0.084 183 0.104 362 0.094 -0.020 -0.068 -0.275 0.138 (0.021) (0.023) (0.015) Reach goals 193 3.472 193 3.523 386 3.497 -0.052 -0.035 -0.235 0.164 (0.106) (0.105) (0.075) Focus on goals 193 4.332 193 4.440 386 4.386 -0.109 -0.113 -0.313 0.086 (0.073) (0.066) (0.049) Up to responsibilities 193 4.145 193 4.212 386 4.179 -0.067 -0.062 -0.262 0.137 (0.079) (0.076) (0.055) Fulfill duty 193 4.528 193 4.503 386 4.516 0.026 0.031 -0.168 0.231 (0.057) (0.062) (0.042) Household-level variables # Beds in household 193 4.850 193 4.845 386 4.847 0.005 0.002 -0.198 0.201 (0.206) (0.184) (0.138) # Children in household 193 4.596 193 4.321 386 4.459 0.275 0.076 -0.123 0.276 (0.283) (0.232) (0.183) # Adults in household 193 6.290 193 6.560 386 6.425 -0.269 -0.071 -0.270 0.129 (0.286) (0.264) (0.194) Household debt 193 29274.611 193 27331.606 386 28303.109 1943.005 0.033 -0.166 0.233 (4418.191) (3961.204) (2963.524) Household savings 193 74093.264 193 83419.689 386 78756.477 -9326.425 -0.091 -0.291 0.109 (7054.669) (7671.283) (5209.624) Household income 193 53756.477 193 51994.819 386 52875.648 1761.658 0.060 -0.139 0.260 (2140.630) (2059.713) (1484.070) Note: Last earnings is given in thousand CFA francs, Senegal’s local currency. Recent migrant is defined as living for three years or less in current location and moved from outside Dakar. Middle school indicates whether an individual has completed middle school. Correcting for multiple hypothesis eliminates any statistical significance between group differences. 29 Table B.3: Correction for multiple inference of the treatment effects Bivariate Baseline Baseline controls outcomes Panel A · Labour market outcomes Employed Unadjusted p-value 0.187 0.213 Sharpened q-value 0.078 0.077 Still at company Unadjusted p-value 0.048 0.039 Sharpened q-value 0.051 0.042 Last earnings Unadjusted p-value 0.005 0.003 Sharpened q-value 0.017 0.010 Panel B · Conscientiousness skills Reach goals Unadjusted p-value 0.027 0.048 0.059 Sharpened q-value 0.371 0.916 1.000 Focus on goals Unadjusted p-value 0.096 0.082 0.081 Sharpened q-value 1.000 1.000 1.000 Responsible Unadjusted p-value 0.006 0.008 0.009 Sharpened q-value 0.187 0.268 0.219 Fulfill duty Unadjusted p-value 0.010 0.013 0.011 Sharpened q-value 0.187 0.268 0.219 Note: Each cell contains p- or q-values for the multiple regressions presented in Table 1 to 3. In Panel A, we correct for testing three hypotheses (three outcome variables and one treatment), while in Panel B we correct for 32 hypotheses (all conscientiousness items). The sharpened q-values are calculated using the Stata code from Anderson (2008). 30 Table B.4: Occupations at baseline of workers still employed in the construction company vs workers not employed in the company at end-line (1) (2) T-test Not in company Still in company Difference Variable N /n Mean/SE N /n Mean/SE (1)-(2) Security agent helper 181 0.304 187 0.385 -0.081 [55] (0.034) [72] (0.036) Mason 181 0.127 187 0.176 -0.049 [23] (0.025) [33] (0.028) Iron worker 181 0.105 187 0.053 0.051* [19] (0.023) [10] (0.016) Form setter 181 0.149 187 0.118 0.032 [27] (0.027) [22] (0.024) Carpenter 181 0.017 187 0.000 0.017* [3] (0.010) [0] (0.000) Laborer 181 0.271 187 0.251 0.019 [49] (0.033) [47] (0.032) Driver help 181 0.006 187 0.000 0.006 [1] (0.006) [0] (0.000) Topographer/Topographer helper 181 0.006 187 0.005 0.000 [1] (0.006) [1] (0.005) Flag holder 181 0.000 187 0.005 -0.005 [0] (0.000) [1] (0.005) Specialized laborer/worker 181 0.017 187 0.000 0.017* [3] (0.010) [0] (0.000) Unspecified worker 181 0.000 187 0.005 -0.005 [0] (0.000) [1] (0.005) Note: This table compares the occupations at baseline of those, at end-line, that remained in the construction company with those that did not. In square brackets, the number of individuals per occupation either still in the construction company or not. 31 Table B.5: Activity sector for workers who left the construction company Freq. Pct. Construction 53 64.63 Wood and metal work, carpentry, craftmanship 12 14.63 Dealer/Salesman/Retail sales 8 9.76 Electrician 1 1.22 Driver 1 1.22 Cleaner 1 1.22 Security/Guardian/Soldier 1 1.22 Agriculture, farming, fisherman 3 3.66 Daily worker 1 1.22 ∗ Tyre mechanic 1 1.22 Total 82 100.00 Note: Sector measured at end-line. Out of the 181 respondents not working in the company at end-line, 82 had ∗ other jobs, 97 were unemployed and 2 had left the labour market. The French expression was vulgarisateur. 32 Table B.6: Earnings at end-line with interaction (1) (2) (3) Last earnings Last earnings Last earnings Treated 24408.6∗∗∗ 15759.7∗∗ 23697.7∗∗ (0.001) (0.018) (0.013) ∗∗∗ Still at company 73271.0 80942.9∗∗∗ (0.000) (0.000) Treated × Still at company -15480.5 (0.244) Constant 21716.9 -1626.4 -5341.4 (0.304) (0.930) (0.776) Standardized coeff. 0.313 0.202 0.304 Mean outcome control gr. 112644.92 112644.92 112644.92 Baseline controls Baseline outcome N 370 367 367 R-sq 0.169 0.378 0.378 Note: Columns 1 and 2 are columns 3 and 4 from Table 2 and added to help comparison with main results. p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household,total (formal and informal) household debt, total household savings, household income. Baseline outcome: earnings measured at baseline. 33 C. Details on earnings computation Earnings were measured in different ways in the three survey rounds. At baseline, all workers were asked both their last wage and, as a consistency check, the range where their last wage fell. At mid-line, due to survey constraints, we could only ask whether their wage had risen or fallen compared to baseline, and by how much it rose or fell. At end-line, workers still employed at the construction company were again asked whether their wage varied and by how much, while workers employed elsewhere were asked the amount they earned and the range where that amount fell, just as in baseline. From this, we computed interpolated wages at end-line for workers still at the construction company. First, we checked that the exact wage had been reported at baseline: if that was missing, we used the range to infer the exact wage, using “midpoint estimates”, i.e., the median point in each range category. For example, if a worker did not report the exact wage, but indicated that her wage was in the third category, which corresponds to a wage between 180000−130000 130,000 and 180,000 CFAF, then we infer that her exact wage was 2 = 155,000. Next, using the answers to the items “did your wage rise or fall since baseline?” and “by how much did it rise or fall?” we interpolated mid-line wages. The amount by which wages rose or fell was also given as a range, so we also applied the midpoint estimates to infer an exact amount for the increase or decrease. For instance, if baseline wage was 145,000 CFAF and the worker reported an increase between 5,001 and 10,000 CFAF, the interpolated wage −5001 for mid-line would be 145000 + 100002 = 147,499.5. We applied the same procedure for end-line wages of workers still working in the construction company. For workers who had left the construction company, we proceeded in the same fashion as in baseline: we used the exact wage, and if missing we inferred the amount using the midpoint of the range category. 34 D. Details on conscientiousness measures D.1. Description of conscientiousness questions At baseline and midline, we measure conscientiousness using a 32-item questionnaire. Four conscientiousness scales are defined, and each item measures a specific one of them. These four scales are responsibility, industriousness, organizational and punctuality. Table D.1 describes each item and its associated scale. Each respondent was asked in face-to-face interviews to answer the questions by using a five-point scale: (1) very difficult, (2) fairly difficult, (3) neither easy nor difficult, (4) fairly easy, and (5) very easy. 35 Table D.1: Conscientiousness traits: list of all items Conscientiousness trait How easy or difficult is it to... Organizational Keep things tidy and in order Follow a schedule Use a method to follow-up Organize schedule Clean up after yourself To keep personal spaces organized Keep things well organized Keep things in order Industriousness Make plans Work to reach your goals Fix clear goals Set high standards Change ways of working towards a goal after a setback Work hard to succeed Focus on most important goals Make plans to reach a goal Keep trying after failing Responsibility Be there when others need me That other people depend on me Keep promises Fulfill my promises and engagements Live up to responsibilities Respect engagements That other people are counting on me Fulfill my duties and obligations Manage responsibilities Punctuality Show remorse if late at a meeting Go to meeting early Avoid being late to a meeting or work Avoid procrastination Be punctual Be late to meetings 36 D.2. Reliability of scales Table D.2: Conscientiousness traits: reliability of scales Scale Round Cronbach’s alpha Interitem covariance Number of items All traits Baseline 0.793 0.106 32 Midline 0.785 0.073 32 Organisational Baseline 0.588 0.091 8 Midline 0.495 0.034 8 Industriousness Baseline 0.615 0.21 9 Midline 0.684 0.219 9 Responsibility Baseline 0.707 0.155 9 Midline 0.58 0.079 9 Punctuality Baseline 0.428 0.144 6 Midline 0.343 0.065 6 Note: Each of the 32 conscientiousness question relates to one of the four conscientiousness traits. This table reports the reliability of each of these traits, as measured by Cronbach’s α and interitem covariance, as well as the overall reliability. 37 D.3. Impact of treatment on conscientiousness skills Table D.3: Conscientiousness skills: alternative specifications (1) (2) (3) (4) Reach goals Focus on goals Responsible Fulfill duty Treated 0.316∗∗ 0.152∗ 0.317∗∗∗ 0.165∗∗∗ (0.027) (0.096) (0.006) (0.010) Constant 3.811∗∗∗ 4.509∗∗∗ 4.178∗∗∗ 4.669∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.238 0.180 0.299 0.274 N 343 343 343 343 R-sq 0.0143 0.0081 0.0224 0.0194 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; Specification excludes all baseline controls. 38 Table D.4: Conscientiousness traits: Organisational skill I (1) (2) (3) (4) Tidy Follow schedule Follow-up Organize schedule Treated 0.00316 -0.0131 0.0732 -0.00418 (0.941) (0.739) (0.498) (0.938) [1.000] [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ ∗∗∗ Constant 4.734 4.686 3.874 4.299∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.00811 -0.0366 0.0730 -0.00844 Mean outcome control gr. 4.58 4.77 4.18 4.5 Baseline controls Trait at baseline N 343 343 343 343 R-sq 0.0307 0.0153 0.0578 0.0486 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. Table D.5: Conscientiousness traits: Organisational skill II (1) (2) (3) (4) Clean up Organized personal space Keep organized Keep tidy Treated 0.0393 -0.0217 0.00960 -0.0189 (0.582) (0.620) (0.845) (0.634) [1.000] [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ ∗∗∗ Constant 4.146 4.660 4.737 4.740∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.0596 -0.0547 0.0213 -0.0525 Mean outcome control gr. 4.72 4.77 4.84 4.8 Baseline controls Trait at baseline N 343 343 343 343 R-sq 0.0570 0.0100 0.0461 0.0132 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. 39 Table D.6: Conscientiousness traits: Industriousness skill I (1) (2) (3) (4) (5) Make plans Work goals Clear goals High standards Change after setback Treated -0.149 0.268∗ 0.104 -0.0210 -0.0376 (0.221) (0.059) (0.285) (0.822) (0.777) [1.000] [1.000] [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ Constant 4.312 3.941 3.557 4.478 4.161∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Standardized coeff. -0.131 0.202 0.117 -0.0248 -0.0308 Mean outcome control gr. 3.96 3.47 4.26 4.22 3.93 Baseline controls Trait at baseline N 343 343 343 343 343 R-sq 0.0781 0.0738 0.0404 0.0153 0.0398 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. Table D.7: Conscientiousness traits: Industriousness skill II (1) (2) (3) (4) Work hard Focus on goals Plan to reach goal Keep trying after failure Treated 0.115 0.162∗ 0.0491 -0.155∗ (0.391) (0.081) (0.673) (0.087) [1.000] [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ ∗∗∗ Constant 3.632 4.666 3.818 4.213∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.0931 0.192 0.0461 -0.187 Mean outcome control gr. 4.15 4.33 3.75 4.37 Baseline controls Trait at baseline N 343 343 343 343 R-sq 0.0448 0.0278 0.0463 0.0365 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. 40 Table D.8: Conscientiousness traits: Responsibility skill I (1) (2) (3) (4) (5) Present for others Dependable Keep promises Fullfill engagements Up to responsabilities Treated -0.0104 -0.0901 -0.0139 -0.0181 0.298∗∗∗ (0.935) (0.396) (0.743) (0.694) (0.009) [1.000] [1.000] [1.000] [1.000] [0.219] Constant 3.362∗∗∗ 3.162∗∗∗ 4.348∗∗∗ 4.781∗∗∗ 4.030∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Standardized coeff. -0.00887 -0.0900 -0.0355 -0.0431 0.281 Mean outcome control gr. 3.99 4.27 4.83 4.77 4.15 Baseline controls Trait at baseline N 343 343 343 343 343 R-sq 0.0532 0.0868 0.0514 0.0207 0.0844 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. Table D.9: Conscientiousness traits: Responsibility skill II (1) (2) (3) (4) Respect engagements Others count on me Fullfill duty Manage responsibilities Treated 0.0393 -0.0487 0.162∗∗ 0.0334 (0.361) (0.592) (0.011) (0.578) [1.000] [1.000] [0.219] [1.000] ∗∗∗ ∗∗∗ ∗∗∗ Constant 4.811 4.034 4.566 4.401∗∗∗ (0.000) (0.000) (0.000) (0.000) Standardized coeff. 0.0994 -0.0580 0.274 0.0609 Mean outcome control gr. 4.75 4.3 4.53 4.61 Baseline controls Trait at baseline N 343 343 343 343 R-sq 0.0528 0.0511 0.0629 0.0296 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. 41 Table D.10: Conscientiousness traits: Punctuality skill I (1) (2) (3) Remorse late Meet earlier Avoid being late Treated -0.000120 0.0373 -0.0164 (0.999) (0.540) (0.852) [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ Constant 4.457 4.550 4.661∗∗∗ (0.000) (0.000) (0.000) Standardized coeff. -0.000137 0.0670 -0.0204 Mean outcome control gr. 4.16 4.66 4.41 Baseline controls Trait at baseline N 343 343 343 R-sq 0.0410 0.0314 0.0315 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. Table D.11: Conscientiousness traits: Punctuality skill II (1) (2) (3) Avoid procrastination Punctual On time at meetings Treated 0.0148 0.0135 0.154 (0.919) (0.778) (0.192) [1.000] [1.000] [1.000] ∗∗∗ ∗∗∗ Constant 4.103 3.832 3.914∗∗∗ (0.000) (0.000) (0.000) Standardized coeff. 0.0111 0.0299 0.140 Mean outcome control gr. 3.85 4.79 4.25 Baseline controls Trait at baseline N 343 343 343 R-sq 0.0243 0.0890 0.0614 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Mean outcome control gr.: mean level of the outcome variable for the control group at baseline. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income. Baseline outcome: trait measured at baseline. 42 E. COVID-19 related questions Table E.1: Impact of treatment on COVID-19 outcomes: bivariate regressions Panel A (1) (2) (3) (4) (5) (6) (7) Protective Effective Display Contract Lost Income HH income measures measures symptoms change income decrease decrease Treated -0.404*** -0.414*** 0.0341 0.0224 -0.0694 -0.0443 -0.00855 (0.003) (0.002) (0.509) (0.678) (0.160) (0.374) (0.856) [0.020] [0.020] [1.000] [1.000] [0.787] [1.000] [1.000] Constant 3.809*** 3.749*** 0.412*** 0.520*** 0.367*** 0.667*** 0.721*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Std. coeff. -0.310 -0.325 0.0689 0.0448 -0.147 -0.0925 -0.0189 N 371 371 368 346 365 371 371 43 R-sq 0.0241 0.0265 0.00119 0.000503 0.00543 0.00214 0.0000900 Panel B (8) (9) (10) (11) (12) (13) (14) Transfers received Food expend. Health expend. Educ expend. Transfers sent Savings Borrowing decrease increase increase increase increase decrease increase Treated 0.000756 -0.0343 -0.00924 -0.0391* 0.0249 0.0330 0.0155 (0.988) (0.459) (0.800) (0.088) (0.374) (0.521) (0.723) [1.000] [1.000] [1.000] [0.542] [1.000] [1.000] [1.000] Constant 0.372*** 0.290*** 0.148*** 0.0710*** 0.0656*** 0.557*** 0.219*** (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) Std. coeff. 0.00156 -0.0770 -0.0264 -0.177 0.0925 0.0667 0.0369 N 371 371 371 371 371 371 371 R-sq 0.000000611 0.00148 0.000174 0.00787 0.00214 0.00112 0.000341 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Table E.2: Impact of treatment on COVID-19 outcomes: full-controls regressions Panel A (1) (2) (3) (4) (5) (6) (7) Protective Effective Display Contract Lost Income HH income measures measures symptoms change income decrease decrease Treated -0.412*** -0.425*** 0.0278 0.0217 -0.0916* -0.0500 -0.0137 (0.002) (0.001) (0.594) (0.693) (0.056) (0.314) (0.771) [0.017] [0.017] [1.000] [1.000] [0.287] [1.000] [1.000] Constant 3.882*** 3.831*** 0.550*** 0.715*** 0.441*** 0.897*** 0.871*** (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) Std. coeff. -0.316 -0.335 0.0562 0.0434 -0.194 -0.104 -0.0303 Baseline controls N 371 371 368 346 365 371 371 R-sq 0.0589 0.0615 0.0304 0.0111 0.111 0.0521 0.0417 44 Panel B (8) (9) (10) (11) (12) (13) (14) Transfers received Food expend. Health expend. Educ expend. Transfers sent Savings Borrowing decrease increase increase increase increase decrease increase Treated 0.00368 -0.0462 -0.0188 -0.0361 0.0227 0.0241 0.0131 (0.942) (0.322) (0.600) (0.117) (0.424) (0.640) (0.764) [1.000] [1.000] [1.000] [0.475] [1.000] [1.000] [1.000] Constant 0.292** 0.524*** 0.492*** 0.120* 0.112 0.705*** 0.258** (0.036) (0.000) (0.000) (0.058) (0.151) (0.000) (0.032) Std. coeff. 0.00760 -0.104 -0.0538 -0.164 0.0845 0.0486 0.0312 Baseline controls N 371 371 371 371 371 371 371 R-sq 0.0412 0.0315 0.0695 0.0384 0.0128 0.0445 0.0493 Note: p-values in parentheses, ∗ , ∗∗ , ∗∗∗ denote significance at 10%, 5% and 1% levels; sharpened q-values in square brackets. Baseline controls: sex, age, completed middle school, number of beds in household, number of children in household, number of adults in household, total (formal and informal) household debt, total household savings, household income.