Health and economic benefits of secondary education in the context of poverty: Evidence from Burkina Faso.

Even though formal education is considered a key determinant of individual well-being globally, enrollment in secondary schooling remains low in many low- and middle-income countries, suggesting that the perceived returns to such schooling may be low. We jointly estimate survival and monetary benefits of secondary schooling using detailed demographic and surveillance data from the Boucle du Mouhoun region, Burkina Faso, where national upper secondary schooling completion rates are among the lowest globally (<10%). We first explore surveillance data from the Nouna Health and Demographic Surveillance System from 1992 to 2016 to determine long-term differences in survival outcomes between secondary and higher and primary schooling using Cox proportional hazards models. To estimate average increases in asset holdings associated with secondary schooling, we use regionally representative data from the Burkina Faso Demographic Health Surveys (2003, 2010, 2014, 2017-18; N = 3,924). Survival was tracked for 14,892 individuals. Each year of schooling was associated with a mortality reduction of up to 16% (95% CI 0.75-0.94), implying an additional 1.9 years of life expectancy for men and 5.1 years for women for secondary schooling compared to individuals completing only primary school. Relative to individuals with primary education, individuals with secondary or higher education held 26% more assets (SE 0.02; CI 0.22-0.30). Economic returns for women were 3% points higher than male returns with 10% (SE 0.03; CI 0.04-0.16) vs. 7% (SE 0.02; CI 0.02-0.012) and in rural areas 20% points higher than in urban areas with 30% (SE 0.06; CI 0.19-0.41) vs. 4% (SE 0.01; CI 0.02-0.07). Our results suggest that secondary education is associated with substantial health and economic benefits in the study area and should therefore be considered by researchers, governments, and other major stakeholders to create for example school promotion programs.

Introduction

Education is widely considered as one of the main determinants of socioeconomic status yielding large benefits for the individual [1] as well as on a community level [2, 3]. Education is not only a key predictor of income [4] but has also been linked to many aspects of health. In low-resource settings, increased secondary schooling has been shown to lower HIV-infection risk [5], reduce child mortality [6, 7], and increase adult life expectancy [6-8]. In high-income settings, increases in compulsory education reforms have been shown to reduce adult mortality [9], and to increase life expectancy [10].

While substantial improvements in primary schooling have been achieved during the MDG era [11], the ambitious secondary school enrolment targets specified within the Sustainable Development Goals (SDGs) seem hard to achieve in some parts of the world [12]. Secondary schooling participation rates remain particularly low in sub-Saharan Africa, where only 42% enter last grade of lower secondary education [13]. School dropout may occur due to economic [14, 15] and social reasons [16]; as well as “external”—e.g. security issues, distance to school–and/or “internal” factors—e.g. poor school quality [17]. Even though the health and economic returns to secondary education in low-income settings are generally considered high [6-8], perceptions of low or negative returns to schooling, particular in poor rural areas, may be an additional reason why adolescents discontinue formal education [18-20].

In this paper, we estimate economic and survival benefits of secondary education vs. primary education in poor rural areas of sub-Saharan Africa using detailed demographic surveillance data from Burkina Faso. Rather than just focusing on economic returns to schooling, we assess both the links between schooling and life expectancy, and the links between schooling and asset holdings. We study this question in a particular context, where upper secondary schooling completion rates are among the lowest globally (< 10% in 2013/14) [21]. To do so, we explore almost 25 years of surveillance data from the Nouna Health and Demographic Surveillance System (HDSS), in rural Burkina Faso, to determine long-run differences in survival outcomes across educational attainment levels. In a second analysis, we estimate economic benefits from schooling using multiple rounds from the Burkina Faso Demographic Health Surveys (DHS) data, and subsequently compare the health and wealth benefits of increased education in the region.

Study context

This study was conducted in Boucle du Mouhoun region, north-west of Burkina Faso. The school system in Burkina Faso consists of a “6-4-3-system” with 6 years of primary, 4 years of lower secondary and 3 years of upper secondary schooling. Burkina’s upper secondary school completion rate (<10% in 2013/14) is among the lowest in the world, with particularly low secondary schooling rates in rural areas and among girls [13, 21–23]. Burkina Faso’s gross domestic product per capita (GDPPC) was USD 822 in 2019 (constant 2010 USD) [24]. The GDP is mainly created in the service sector (43.6%), one third comes from agriculture and one fifth from industry [25, 26]. In the Nouna HDSS in the region Boucle du Mouhoun most people live in rural areas and work in agriculture and animal husbandry [27]. The semi-urban town of Nouna has better access to the education and health system than the surrounding villages [27]. Life expectancy at birth in Burkina Faso rose from around 50 years in 2000 to 61 years in 2018 [28].

Data description

Nouna health and demographic surveillance survey data

The Nouna HDSS covers a population of 105,000 habitants with a mainly rural population [27, 29]. The Nouna HDSS is managed by the Health Research Centre of Nouna (Centre de Recherche en Santé de Nouna—CRSN) which has collected data since 1992 (initial census) and control census’ in 1994, 1998 and 2009 [29]. The surveillance site was extended in 2000 and in 2004, now counting 58 villages vs. 39 in the initial census and the city of Nouna [29, 30]. Vital events were collected up to 3 times (every 120 days) per year besides other information such as verbal autopsies and household questionnaires among others [29]. The sampling strategy for the collection of information on educational status could not be reconstructed; information on educational status was given for almost one third of the total sample. See for missingness. Additional details on the Nouna HDSS, data collection, and cohort are available elsewhere [27, 29–31]. For the current analysis, we extracted data on sex, age, education, survival status, time of visit, and migration status. Individuals entered the study when first observed and exited in case of death, migration out of the Nouna HDSS or other reasons. All individuals observed for at least 1 year (at least 2 observations) with complete data on variables of interest (age, sex, educational attainment, self-reported survival status) who were a (former) resident of the Nouna HDSS in the period 1992 to 2016, and born in or before 1980 were included in our analysis. We limited our sample to those born before 1980 to study old-age survival. Our final sample included 14,892 individuals with a mean observation time of 11.7 years (range: 1.0–24.6 years).

Demographic and Health Survey (DHS) data

To estimate the economic benefits of education, we used multiple Burkina Faso DHS surveys with data on sub-national administrative units (2003, 2010, 2014, and 2017–18). DHS are nationally representative household surveys and cover all regions of Burkina Faso. Additional details on the DHS surveys have been described elsewhere [32]. To improve comparability with the Nouna HDSS data, we only included individuals from the Boucle du Mouhoun region, where the Nouna HDSS is located [29]. Data were extracted on sex, age, education, household wealth assets, geographical region, sampling weights, date of interview, number of household members, type of place of residence (rural/urban), and marital status, using information from the DHS household recode, personal recode and individual recode files. The DHS did not collect information on income because a majority of the population in this low-resource setting engages in subsistence farming or depend on informal jobs with irregular incomes. We included all working-age individuals (defined as 25 to 64 years), with complete data on educational attainment in the household roster. Information on educational status was given for 3,924 out of 5,043 working age individuals. The last two Burkina Faso survey rounds were Malaria Indicator Surveys (MIS) which do not collect educational information for men. See for a study participant flow diagram. Sampling weights were used for the creation of asset quintiles as outlined in DHS (2021).

Statistical analysis

Our analysis proceeded in two steps. First, we conducted a survival analysis using Cox proportional hazards regression models and the Nouna HDSS dataset. Second, we estimated wealth regression using a Mincer equation and DHS data. displays a conceptual framework underlying our analysis.

Survival analysis

To determine the relationship between education and survival status, we employed a multivariable Cox proportional hazards model. The Cox proportional hazards model is a survival model that allows the hazard of a certain event (here: death) to change over time under the condition that the hazard ratios in between the different groups stay the same (proportional- hazards assumption) [33]. In our application, the Cox proportional hazards model describes the hazard of “all-cause mortality” (outcome) while assuming that the hazards in between the different schooling groups remain the same. In the HDSS, information on educational attainment was collected for a subsample of 14,892 individuals. Most individuals entered the HDSS between 1990 and 2000 –this entry in the HDSS constitutes the beginning of the survival analysis. Observation periods end (censoring occurs) when individuals migrated out of the HDSS or exited the study for other reasons. Our primary exposure of interest was educational attainment. Due to a limited number of individuals in the higher education categories we used education as continuous variable instead of categorizing educational attainment as in subsequent analyses. We also predicted life expectancy for three levels of educational attainment defined as either no schooling, at least some primary schooling (1 to 6 years of schooling), or at least some secondary schooling and higher (7+ years of schooling). Our analysis focused on the full sample (‘basic model’, with years of birth between 1900 and 1980) as well as three birth cohorts, including individuals born between (i) 1st January 1940 and 31st December 1949, (ii) 1st January 1950 and 31st December 1959, and (iii) 1st January 1960 and 31st December 1969. In all analyses, we controlled for sex. Additionally, we controlled for year of birth in our ‘basic model’. We assessed survival benefits separately by urban (city of Nouna) vs. rural (surrounding villages).

Wealth regression

We use asset ownership as a proxy for permanent income and estimate average increases in asset holdings associated with additional education. Asset quintiles were used as outcome in standard Mincer regression models to estimate returns to education [34-36]. The Mincer equation expresses log earnings as a function of years of schooling and a quadratic labor market experience term. We abstained from including more controls in the equation as this might exacerbate bias already existent in the data [34, 37]. We categorized educational attainment into three groups: no schooling (<1 year), primary schooling (1 to 6 years of schooling), and secondary schooling and higher (7 and more years of schooling). Individuals having at least started primary school were chosen as reference group to compare income benefits associated with secondary schooling to those associated with primary schooling. We provide results for a continuous education variable (years) in . Wealth regressions were performed for all working age individuals (25 to 64 years of age) with complete information on education. 1,119 working age individuals had no data on education leaving an analytical sample of 3,924 individuals for the Boucle du Mouhoun region. We adjusted for age, as a proxy for labor market experience, age squared as part of the Mincer equation and DHS survey year. For methodology and estimation of predicted incomes and relative health and financial returns see .

Results

Sample description HDSS data

As summarized in , a total of 14,892 individuals were included in the survival analysis. 51.2% were men and 47.9% were women. Median age of women was around one year higher than that of men when first enrolled in the HDSS (33 compared to 32 years). Median years of schooling was highest for men (3 years) ranging from 0–22 years compared to a median of 0 years for women (range: 0–20 years). Most individuals had no formal education (45.6% of men and 68.6% of women), with less than one fifth of individuals attaining 7 or more years of schooling (at least some secondary education). Characteristics of the 1940s, 1950s and 1960s cohorts are summarized in .

Survival benefits: Survival analysis

shows the results of our survival analysis using Cox proportional hazards models. Controlling for sex and year of birth we found that each year of school was associated with a 3.6% reduction in the mortality hazard (aHR 0.96, 95% Confidence Interval [CI] 0.95–0.98). Female mortality was generally lower than male mortality (‘Basic model’: aHR 0.73; CI 0.66–0.80). When we stratified results by decade of birth, largest mortality reductions were observed for the 1950–59 cohort (aHR 0.94; CI 0.91–0.98).We also found larger survival benefits for the rural compared to the urban population with mortality reductions of 13.3% and 16.2% in the 1940–49 and 1960–69 cohorts (‘Rural 1940–49’: aHR 0.87, CI 0.75–1.01 and ‘Rural 1960–69’: aHR 0.84, CI 0.75–0.94; ‘Urban 1940–49’: aHR 0.95, CI 0.91–0.99 and ‘Urban 1960–69’: aHR 0.96, CI 0.92–1.00). shows exemplarily with Kaplan Meier curves that individuals with at least some schooling born in year 1940 and 1950 get older than individuals having received no schooling. shows median life expectancies for individuals aged 23 to 114 years by educational attainment level. Secondary- and higher-schooled men and women lived on average 1.9 and 5.1 years longer than primary-schooled men and women, respectively.

Sample description DHS data

Sample characteristics for individuals from the Boucle du Mouhoun region surveyed in the DHS are shown in . 3,924 individuals were analyzed. 61.8% of respondents were female, with a slightly higher median age for men than for women (38 and 36 years respectively). Median schooling was 0 years with almost 80% of women and two thirds of men not having attended school. About one fifth of men and less than 10% of women went to secondary school. Median estimated yearly income was 883 USD with about USD 300 higher incomes for men than for women (USD 1,066 vs. USD 757).

Economic benefits: Wealth regressions

shows the basic relationship between schooling and increase in asset holdings. shows the results for the wealth regressions. On average, secondary or higher schooling was associated with an increase in asset holdings of 26% (SE 0.02; CI 0.22–0.30) in the pooled sample. Returns for women were 3% points higher than male returns: 10% (SE 0.03; CI 0.04–0.16) vs. 7% (SE 0.02; CI 0.02–0.012). The difference between rural and urban residents was even higher with more than 20% points difference: 30% (SE 0.06; CI 0.19–0.41) vs. 4% (SE 0.01; CI 0.02–0.07). Results were similar when analyzing the nationally representative dataset (). For wealth regressions by year of educational attainment instead of educational attainment categories, see . For results for Burkina Faso see and .

Economic analysis.

Estimated increase in asset ownership by educational attainment level in the Boucle du Mouhoun region, Burkina Faso. Notes: Asset ownership by year of educational attainment based on OLS regression results from Mincer Earnings regressions, controlling for age, age squared and DHS survey year. Source: data for the Boucle du Mouhoun region using the Burkina Faso Demographic and Health Surveys (DHS) of 2003, 2010, 2014, and 2017–18 (N = 3,924). S4 Fig in the appendix shows results for the full (nationally representative) dataset of Burkina Faso.

Discussion

In this paper, we have used surveillance data from the Nouna HDSS and repeated cross-sectional data from the DHS to estimate the associations between secondary schooling, health, and incomes. We show that on average life expectancy for women with secondary or higher education is 5 years higher than life expectancy for women with only primary school; for men, the difference is about 2 years. We also show that the financial returns to secondary school are high, with an increase in asset quintiles of 26%. Benefits seemed to be generally higher for women and rural areas than those for men and urban areas. The higher returns to secondary education could be interpreted as evidence for education being particularly productive in groups where such education is scarce. It is also possible, that the additional educational opportunities after secondary schooling–such as vocational training–may be particularly attractive for women [38]. The relatively high returns to secondary education may of course also represent marriage outcomes, with more highly educated women much more likely to marry somebody with high socioeconomic status.

Our results suggest that both–wealth and health benefits—clearly make a major contribution to overall improvements in livelihoods. Only focusing on wealth gains will clearly underestimate the total benefits of staying in school.

The results presented here build on recent research showing that average rates of return are high in general and especially in sub-Saharan Africa (e.g., 12,5% in Montenegro et al, 2014),and that adolescents enrolled in school have a wide range of improved health outcomes compared to their peers in this setting [4, 39, 40]. Concerning the magnitude of health benefits, Gathmann, Jürges et al. (2015) estimate a 2.0% reduction in 20-year-mortality rate for only men and Lleras-Muney (2005) [41] a 10-year mortality rate reduction by 1.3 to 3.6% per additional year of education. Those results are similar to our estimations finding a 3.6% reduction in the mortality hazard. Perceived returns to education are high among other key stakeholders in the demographic surveillance area, such as parents, teachers, as well as health workers [42]. There appears to be a general consensus in literature that women receive higher economic benefits than men (for example an average rate of return of 14,6% for women vs, 11,3% for men) [4, 8, 43, 44] in line with our findings. Peet et al. (2015) found lower benefits for rural areas contrary to our findings. Concerning returns to schooling according to schooling level, Barouni and Broecke (2014) found higher returns for higher (secondary and tertiary) schooling than for primary schooling similar to the current paper, while Psacharopoulos (1985) found highest returns for primary education in the past. This might show a tendency towards higher returns for higher schooling levels. Concerning the methodology, most studies on returns to schooling use the discounting method or Mincer regression similar to us [39].

Literature on returns to education is large–estimates from this setting are however rare [45, 46]. Most of these studies, however, focus on either survival [3] or economic benefits [47, 48]. To our knowledge, only one study to date examines both monetary and non-pecuniary returns to schooling, but in a high-resource setting [49].

A key strength of the analysis is the combination of 25 years of longitudinal data on adult survival with detailed cross-sectional data from the region. This link allows us to comprehensively analyze the long-term outcomes related to education. Future research should take into consideration different dimensions of returns to education, and not concentrate on economic returns alone. Our results suggest that secondary schooling holds significant potential for increasing general life expectancy and earnings and should therefore be considered by researchers, governments, and other major stakeholders to create for example school promotion programs with different incentives. Formal schooling as part of human capital empowers individuals and thereby societies and helps to translate knowledge into progress. We suggest replicating similar analyses across multiple sites (e.g., HDSS sites) to compare and corroborate our results. Further research will be needed to understand the main barriers to secondary schooling access in this low-resource setting.

Study limitations

Our analysis has several limitations. First, data on educational attainment was not collected for all individuals, which means that the results presented may not necessarily be representative of the entire Boucle du Mouhoun community. Second, our life expectancy estimates do not represent life expectancy at birth but rather conditional life expectancies as adults and are thus not directly comparable to national life expectancy estimates. Given that secondary education becomes only relevant in adolescence, we believe that the focus on adult survival outcomes is well justified in our analysis. Third, we only had data on years of schooling, while other aspects of schooling, and schooling quality in particular, are likely also important [50]. Similarly, limited data was available in the HDSS data on other health outcomes beyond survival. Education may have additional health benefits, such as improved mental health, which may not be fully captured in our survival data. Conversely, our definition of “return” does not account for the short-term cost faced by adolescents and families as no data was available on expenditures linked to schooling. Fourth, our research design is observational by construction and does not allow causal inference. Fifth, endogenous selection bias cannot be excluded: Wealthy individuals for example might select into education and bias health and income benefits positively, or very intelligent individuals who might bias stronger income than mortality results.

Conclusions

Using almost 25 years of surveillance data from the Nouna HDSS and cross-sectional data from the DHS, we found that the returns to secondary schooling in rural Burkina Faso are high both in terms of lifetime income and in terms of life expectancy. Women and men with secondary or higher schooling experienced a 26% increase in wealth quintiles and lived 2 to 5 years longer than individuals completing only primary education. Despite these large benefits, less than 10% of adolescents currently finish secondary schooling in this setting [21]. Further research is urgently needed to identify the most effective strategies for reaching the ambitious SDG goals for secondary schooling.

Predicted incomes and relative health and financial returns to education.

(ZIP)

Click here for additional data file.

Study participant flow diagram DHS sample Boucle du Mouhoun region.

(DOCX)

Click here for additional data file.

Framework of statistical analysis.

(DOCX)

Click here for additional data file.

Survival according to schooling—Kaplan Meier graphs.

(DOCX)

Click here for additional data file.

Wealth analysis: Estimated income increase by educational attainment for entire Burkina Faso.

(DOCX)

Click here for additional data file.

Missingness table Nouna HDSS.

(DOCX)

Click here for additional data file.

Wealth regressions for Boucle du Mouhoun, Burkina Faso with education as continuous variable.

(DOCX)

Click here for additional data file.

Sample characteristics of cohorts followed over time.

(DOCX)

Click here for additional data file.

Wealth regressions for entire Burkina Faso with education as categorized variable.

(DOCX)

Click here for additional data file.

Earnings regressions for entire Burkina Faso with education as continuous variable.

(DOCX)

Click here for additional data file.

15 Dec 2021

Health and economic benefits of secondary education in the context of poverty: Evidence from Burkina Faso

PLOS ONE

Dear Dr. Werner,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The two reviewers raised overlapping concerns about the study rationale and its background. Please address all reviewer comments, in particular requests for clarification regarding the data analysis.

Please submit your revised manuscript by Jan 28 2022 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Yann Benetreau, PhD

Senior Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. In ethics statement in the manuscript and in the online submission form, please provide additional information about the patient records used in your retrospective study. Specifically, please ensure that you have discussed whether all data were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

3. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

4. Your ethics statement should only appear in the Methods section of your manuscript. If your ethics statement is written in any section besides the Methods, please move it to the Methods section and delete it from any other section. Please ensure that your ethics statement is included in your manuscript, as the ethics statement entered into the online submission form will not be published alongside your manuscript.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: No

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I think this is a very interesting paper on a very important topic. It is evident that there's a lot pf hard work behind this manuscript however, in some places, it seems that the authors assumed that the crucial aspects of their analysis are common knowledge. For instance, Cox methodology is widely used but this paper should at least provide a brief description about it. Similarly, datasets and their sampling methodologies are not explained in detail. I think it is very important to include this information.

From estimations it appears as if the variables reported in the tables are the only ones included in the regressions i.e there are no controls. This is potentially problematic and all confounding variables should be included. There are are several demographic variables in DHS that could be used.

Mincers equation also seems to be incomplete. Age squared in missing and authors should also attempt education squared in continuous form. See schooling locus in George Borjas book of labor economics.

Citations seem a little unusual to me. Authors mention study number for direct citations and mention the sane number in parentheses as well. Please check the convention.

It would've been awesome if authors explained the results in context of Burkina Faso. It would've been easier to grasp the meaning of 1000 USD in context of the country being studied. Storytelling is the weakest aspect of this paper. Too much emphasis is on estimations and even the review section seems to be rushed. However, with little polishing, it can be accepted.

Reviewer #2: This paper provides evidence about the health and economic benefits of secondary education in rural Burkina Faso, using two sources of evidence: long-run health surveillance data, and the DHS. They find evidence that attending secondary school is associated with higher income as well as higher life expectancy.

My overriding comment about this paper is that it provides only correlational evidence about the relationship of interest, when we know that correlational evidence is probably severely limited, and causal evidence is widely available. The authors' acknowledgment of this fact (absence of causal evidence) is limited to a single sentence in the limitations section. Particularly for the analysis of the returns to education, there is an absolutely enormous economics literature analyzing causal effects of education on earnings in the developing world already published, and so the bar for an additional contribution is high. The authors fail to cite the majority of this literature, I won't aim to summarize it myself but will refer the authors to two recent literature reviews: Peet et al. in the Economics of Education Review 2020, and Patrinos and Psacharopoulos in the Economics of Education (also 2020). I am (furthermore) additionally skeptical of this analysis because the authors don't have actual income data, but are imputing income based on assets as reported in the DHS.

My suggestion is that this entire analysis be dropped from the paper and that the authors focus only on the health benefits of secondary education. Here the authors seem to be on stronger ground in reporting only a correlation for several reasons: they are using a relatively novel data source with long-term longitudinal data; and the literature is smaller. However, the authors again need to do a lot more to situate their findings in this literature, however large it is. There are a number of papers cited (citations 3 to 10, broadly), but in the discussion section the authors should more clearly elaborate on what they find relative to the existing papers and what contribution these findings make. Are the existing estimates from comparable settings or using similar research designs, or are they different? Is the magnitude of the relationship between education and mortality they estimate similar to existing estimates, or different? Is the potential bias due to endogenous selection into education similar in this analysis, or worse? (In general - in a context such as rural Burkina Faso where almost no one attends secondary school, we would expect this bias to be particularly acute: only highly motivated and intelligent students from families with adequate resources and/or a high interest in education would attend secondary school.)

Having dropped the analysis of income returns to education, the authors would then have more space for a longer discussion section that might fruitfully grapple with some of these questions around situating the mortality analysis in the literature and unpacking how to interpret this relationship.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PLOS_One_D_21_01735_Review.pdf

Click here for additional data file.

6 Mar 2022

Response to Reviewers

Review Comments to the Author

Reviewer #1:

1) I think this is a very interesting paper on a very important topic. It is evident that there's a lot of hard work behind this manuscript however, in some places, it seems that the authors assumed that the crucial aspects of their analysis are common knowledge. For instance, Cox methodology is widely used but this paper should at least provide a brief description about it. Similarly, datasets and their sampling methodologies are not explained in detail. I think it is very important to include this information.

We thank the reviewer for his/her valuation of the work provided for this research. More detailed explanations on methodology and datasets were included, for example:

“To determine the relationship between education and survival status, we employed a multivariable Cox proportional hazards model. Our outcome was all-cause mortality. The Cox proportional hazards model is a survival model that allows the hazard of a certain event (here: death) to change over time under the condition that the hazard ratios in between the different groups stay the same (proportional-hazards assumption) (34). In our application, the Cox proportional hazards model describes the hazard of “all-cause mortality” (outcome) while assuming that the hazards in between the different schooling groups remain the same.”

2) From estimations it appears as if the variables reported in the tables are the only ones included in the regressions i.e there are no controls. This is potentially problematic and all confounding variables should be included. There are several demographic variables in DHS that could be used.

In the survival regressions, we controlled for sex and year of birth in the basic model. The Nouna HDSS dataset does not include too many other relevant controls for the survival analysis.

With regard to the earnings regressions, all models included controls for age squared. We tried to include the most important pre-determined variables by stratifying for sex and type of location (urban/rural). Those stratified models additionally included controls for DHS survey round. We considered other pre-determined variables such as place of birth or birth order, but these are only available for younger household members in the DHS. We also considered other factors such as marital status or fertility but decided to exclude them from the model because these factors are likely directly determined by education and thus partially explain the effects of interest.

“We abstained from including more controls in the [Mincer]equation as this might exacerbate bias already existent in the data (38, 41.)”

3) Mincer’s equation also seems to be incomplete. Age squared in missing and authors should also attempt education squared in continuous form. See schooling locus in George Borjas book of labor economics.

We apologize for the lack of clarity. Age squared is one of the variables we had included in the Mincer regression, as recommended by the Reviewer, but was not shown in the main tables for visualization purposes. We have now further clarified this point throughout the paper:

“We adjusted for age, as a proxy for labor market experience, age squared as part of the Mincer equation and DHS survey year.”

“Table 5. Income analysis: Results from OLS earnings regression models for the Boucle du Mouhoun region, Burkina Faso.

[Table]

Notes: […] In all models we controlled for age squared; in the stratified models we additionally controlled for survey round. OLS: ordinary least squares.”

Additionally, we have now included regression results when adding education squared in continuous form in the supplementary information (shown in Table S3).

4) Citations seem a little unusual to me. Authors mention study number for direct citations and mention the sane number in parentheses as well. Please check the convention.

We thank the Reviewer for bringing this to our attention. We have now carefully reviewed the reference list and have corrected all corresponding citations.

5) It would've been awesome if authors explained the results in context of Burkina Faso. It would've been easier to grasp the meaning of 1000 USD in context of the country being studied. Storytelling is the weakest aspect of this paper. Too much emphasis is on estimations and even the review section seems to be rushed. However, with little polishing, it can be accepted.

We have now provided additional context for the results – e.g., see:

“We also show that the financial returns to secondary school are high, with an estimated additional USD 8,000 of lifetime income which equals almost 1,740,000 CFA representing around 3 to 4 annual salaries based on GDP per capita (44, 45). Combining health and economic benefits suggests a total return between approx. USD 10,000 and USD 33,000 (> 7,000,000 CFA) in this region.”

We decided that it would be difficult, however, to add more daily life comparisons - such as the rent of an apartment - as there is high price variability and few reliable sources to cite such comparisons.

Reviewer #2:

1) This paper provides evidence about the health and economic benefits of secondary education in rural Burkina Faso, using two sources of evidence: long-run health surveillance data, and the DHS. They find evidence that attending secondary school is associated with higher income as well as higher life expectancy. My overriding comment about this paper is that it provides only correlational evidence about the relationship of interest, when we know that correlational evidence is probably severely limited, and causal evidence is widely available. The authors' acknowledgment of this fact (absence of causal evidence) is limited to a single sentence in the limitations section. Particularly for the analysis of the returns to education, there is an absolutely enormous economics literature analyzing causal effects of education on earnings in the developing world already published, and so the bar for an additional contribution is high. The authors fail to cite the majority of this literature, I won't aim to summarize it myself but will refer the authors to two recent literature reviews: Peet et al. in the Economics of Education Review 2020, and Patrinos and Psacharopoulos in the Economics of Education (also 2020).

We thank the reviewer for this helpful comment on the pre-existing literature. We included the contribution in the Economics of Education Review 2020 from Patrinos and Psacharopoulos and also added some additional references (Peet et al 2015, Patrinos and Psacharopolous 2018).

“The results presented here build on recent research showing that average rates of return are high in general and especially in sub-Saharan Africa (e.g., 12,5% in Montenegro et al, 2014), and that adolescents enrolled in school have a wide range of improved health outcomes compared to their peers in this setting (4, 48, 49).”

2) I am (furthermore) additionally skeptical of this analysis because the authors don't have actual income data but are imputing income based on assets as reported in the DHS.

We acknowledge the fact that only imputed income is used in our research. Income data is not available in the DHS, unfortunately, and we have now further clarified this in the paper:

“Data were extracted on sex, age, education, household wealth assets, geographical region, sampling weights, date of interview, number of household members, type of place of residence (rural/urban), and marital status, using information from the DHS household recode, personal recode and individual recode files. The DHS did not collect information on income because a majority of the population in this low-resource setting engages in subsistence farming or depend on informal jobs with irregular incomes.”

3) My suggestion is that this entire analysis be dropped from the paper and that the authors focus only on the health benefits of secondary education. Here the authors seem to be on stronger ground in reporting only a correlation for several reasons: they are using a relatively novel data source with long-term longitudinal data; and the literature is smaller. Having dropped the analysis of income returns to education, the authors would then have more space for a longer discussion section that might fruitfully grapple with some of these questions around situating the mortality analysis in the literature and unpacking how to interpret this relationship.

We agree with the Reviewer that the economic returns to education literature is large and that we only report correlational data. There are other papers that look at survival benefits as well, but to our knowledge there is no paper that has tried to combine economic and health returns in this context, which is in our view the novelty of this paper.

4) However, the authors again need to do a lot more to situate their findings in this literature, however large it is. There are a number of papers cited (citations 3 to 10, broadly), but in the discussion section the authors should more clearly elaborate on what they find relative to the existing papers and what contribution these findings make. Are the existing estimates from comparable settings or using similar research designs, or are they different? Is the magnitude of the relationship between education and mortality they estimate similar to existing estimates, or different? Is the potential bias due to endogenous selection into education similar in this analysis, or worse? (In general - in a context such as rural Burkina Faso where almost no one attends secondary school, we would expect this bias to be particularly acute: only highly motivated and intelligent students from families with adequate resources and/or a high interest in education would attend secondary school.)

We agree that the literature on returns to education is large – estimates from this setting are however rare. There are other papers that look at survival benefits as well, but to our knowledge there is no paper that has tried to combine them, which is in our view the novelty of this paper. We would be most grateful for any additional references if there are any.

We concur that endogenous selection into education is possible, which would affect health and economic outcomes in similar ways and therefore not falsify our comparisons between health and economic returns. Literature on returns to education, however, suggests that instrumental variables estimates are generally similar to - or slightly larger - than ordinary least squares based Mincerian regressions (1). This is why we consider selection biases not being large. We have now tried to include more literature to better situate our findings:

“The results presented here build on recent research showing that average rates of return are high in general and especially in sub-Saharan Africa (e.g. 12,5% in Montenegro et al, 2014), and that adolescents enrolled in school have a wide range of improved health outcomes compared to their peers in this setting (4, 48, 49). Perceived returns to education are high among other key stakeholders in the demographic surveillance area, such as parents, teachers, as well as health workers (50). There appears to be a general consensus in literature that women receive higher economic benefits than men (for example an average rate of return of 14,6% for women vs, 11,3% for men) (4, 8, 50, 51) in line with our findings. Peet et al, (2015) found lower benefits for rural areas contrary to our findings. Concerning returns to schooling according to schooling level, Barouni and Broecke (2014) found higher returns for higher (secondary and tertiary) schooling than for primary schooling similar to the current paper, while Psacharopoulos (1985) found highest returns for primary education in the past. This might show a tendency towards higher returns for higher schooling levels. Concerning the methodology, most studies on returns to schooling use the discounting method or Mincer regression similar to us (49).

Literature on returns to education is large – estimates from this setting are however rare (54, 55). Most of these studies, however, focus on either survival (3) or economic benefits (56, 57). To our knowledge, only one study to date examines both monetary and non-pecuniary returns to schooling, but in a high-resource setting (58).”

Literature Cited

1. Patrinos HA, Psacharopoulos G. Returns to education in developing countries. In: The Economics of Education: Elsevier; 2020. p. 53–64.

Submitted filename: Response to Reviewers_26Jan.docx

Click here for additional data file.

14 Apr 2022

24 May 2022

1) I have now been asked to serve as a guest editor for this submission; in the interests of transparency, I should note I was previously serving as a reviewer (reviewer #2). I reviewed the updated materials you provided and also invited reviewer #1 to review the manuscript again; s/he declined to do so. My judgment is that you have responded thoroughly to this reviewer's comments. Following up on my own comments, I would like to request some minor additional revisions.

We thank the reviewer for his/her judgment of this paper and the work invested in a second revision.

2) I am uncomfortable with the use of imputed income in an analysis of returns to education, though of course you are right that the DHS does not report income. I suggest that you conduct this analysis primarily using an asset index drawing on the asset information that is actually reported in the DHS (i.e., estimate the correlation between education and assets.) Since this is not in any case a causal analysis that would fall into the core returns to education literature, it does not seem necessary to use an artificial income measure. If you wish to also include the analysis using imputed income as an extension in the paper, of course you can, but I suggest this be a secondary analysis.

We replaced the income estimates with the original asset score in form of asset quintiles and used ln(asset quintile) as dependent variable to calculate increase in asset holdings and described health and economic benefits separately.

“We use asset ownership as a proxy for permanent income and estimate average increases in asset holdings associated with additional education. Asset quintiles were used as outcome in standard Mincer regression models to estimate returns to education (38–40).”

We moved the predicted income table to the supporting information (Supplementary Analysis S1) as suggested.

“For methodology and estimation of predicted incomes and relative health and financial returns see Supplementary Analysis S1 in the supporting information.”

3) In the discussion section, you focus primarily on comparing your estimate of the returns to education to other estimates. What about other estimates of mortality benefits of education - are there any such estimates? What is the magnitude of those estimates compared to yours? It may be that the literature here is minimal; if so, you can just note this.

There is indeed not nearly as much literature on the survival returns to education - we have now added the two main references to the revised Discussion section, where we write:

“Concerning the magnitude of health benefits, Gathmann, Jürges et al. (2015) estimate a 2.0% reduction in 20-year-mortality rate for only men (9) and Lleras-Muney (2005) a 10-year mortality rate reduction by 1.3 to 3.6% per additional year of education (47). Those results are similar to our estimations finding a 3.6% reduction in the mortality hazard.”

4) Following up on your response to the previous referee report: it is not necessarily the case that bias in the estimated returns to education due to selection into education is the same across the two dimensions examined (income and mortality). For example, if very intelligent individuals (who are not necessarily healthy or rich) select into education, the selection bias may be larger for income vis-à-vis mortality. If individuals from wealthy families select into education, then the bias may be inverted (larger for mortality, since wealthier families live longer, ceteris paribus). It is up to you whether you wish to engage with this point in the discussion section, but it may be useful context for your interpretation.

We thank the reviewer for bringing this to our attention; we have now added the following point to the revised Discussion section:

“Fifth, endogenous selection bias cannot be excluded: Wealthy individuals for example might select into education and bias health and income benefits positively, or very intelligent individuals who might bias stronger income than mortality results.”

Submitted filename: Response to Reviewers_24May2022.docx

Click here for additional data file.

8 Jun 2022

Health and economic benefits of secondary education in the context of poverty: Evidence from Burkina Faso

PONE-D-21-01735R2

Dear Dr. Werner,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Jessica Leight, PhD

Guest Editor

PLOS ONE

Additional Editor Comments (optional):

Thanks so much for submitting the revised manuscript, Health and economic benefits of secondary education in the context of poverty: Evidence from Burkina Faso. I'm happy to accept your manuscript for publication in PloS One and believe it has the potential to make a significant contribution to the literature.

14 Jun 2022

PONE-D-21-01735R2

Health and economic benefits of secondary education in the context of poverty: Evidence from Burkina Faso

Dear Dr. Werner:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Jessica Leight

Guest Editor

PLOS ONE

Table 1

Baseline characteristics of the HDSS complete case dataset.

Subsample	Female			Male			Both sexes
	N or median	(IQR)	(% or range)	N or median	(IQR)	(% or range)	N or median	(IQR)	(% or range)
Characteristic
Number of subsequent deaths	843	-	(11.8%)	991	-	(12.8%)	1,834	-	(12.3%)
Age at first visit (years)	33	(25–46)	(11–99)	32	(24–42)	(11–99)	32	(25–44)	(11–99)
Schooling (years)	0	(0–4)	(0–20)	3	(0–6)	(0–22)	0	(0–6)	(0–22)
Highest schooling attainment
None (0 years)	4,889	-	(68.6%)	3,541	-	(45.6%)	8,430	-	(56.6%)
Primary (1–6 years)	1,451	-	(20.3%)	2,560	-	(33.0%)	4,011	-	(26.9%)
Secondary (7+ years)	791	-	(11.1%)	1,660	-	(21.4%)	2,451	-	(16.4%)
Observations	7,131	-	(47.9%)	7,761	-	(52.1%)	14,892	-	(100.0%)

Notes: All 14,892 individuals were followed over the years 1992–2016. Data are from the Nouna Health and Demographic and Surveillance (HDSS), Burkina Faso. IQR: Interquartile range.

Table 2

Survival analysis: Adjusted hazard ratios (95% CI) from multivariate Cox regression models with death as outcome variable.

Birth Cohort	All	1940–49			1950–59			1960–69
	Basic model	Total	Urban	Rural	Total	Urban	Rural	Total	Urban	Rural
Variable
Educational Attainment	0.96***	0.96**	0.95**	0.87*	0.94***	0.92***	0.96	0.96*	0.96*	0.84***
	(0.95–0.98)	(0.92–1.00)	(0.91–0.99)	(0.75–1.01)	(0.91–0.98)	(0.88–0.97)	(0.89–1.03)	(0.93–1.00)	(0.92–1.00)	(0.75–0.94)
Female sex	0.73***	0.59***	0.43***	0.8	0.60***	0.68**	0.48***	0.64***	0.56***	0.7
	(0.66–0.80)	(0.48–0.73)	(0.33–0.57)	(0.58–1.11)	(0.47–0.78)	(0.49–0.95)	(0.32–0.73)	(0.48–0.86)	(0.38–0.82)	(0.45–1.10)
Observations	14,892	1,372	819	553	1,980	1,034	946	3,282	1,788	1,494

Notes: The ‘basic model’ includes years of birth 1900–1980. The sample of the three cohorts included 6,634 individuals. All individuals were followed over the years 1992–2016. Adjusted Hazard Ratios (aHR) from multivariable Cox regressions for the relationship between educational attainment (per additional year of schooling) and death (outcome variable). Confidence intervals in parentheses.

*** p<0.01

** p<0.05

* p<0.1.

In the basic model we controlled for year of birth. Data are from the Nouna Health and Demographic and Surveillance (HDSS), Burkina Faso. Sample sizes were low for rural areas in two cohorts.

Table 3

Conditional life expectancy for individuals aged 23 to 114 years in years by sex and educational group.

			Male	Female
Highest schooling attainment
none (0 years)			70.8	74.7
primary (1–6 years)			72.6	79.5
secondary and higher (7+ years)			74.5	84.6
Mean			71.6	76.8
Median			70.8	74.7

Notes: Median life expectancy predicted from a pooled sample of individuals born between 1902 and 1969 from the Nouna Health and Demographic and Surveillance (HDSS) data, separately for men and women. Life expectancies reflect conditional probabilities among adults aged 23 to 114 in the Nouna HDSS data.

Table 4

Selected characteristics of the DHS sample, Boucle du Mouhoun region.

Subsample	Male			Female			Both Sexes
	N or median	(IQR)	(% or range)	N or median	(IQR)	(% or range)	N or median	(IQR)	(% or range)
Characteristic
Age (years)	38	(30–47)	(25–64)	36	(30–45)	(25–64)	37	(30–45)	(25–64)
Schooling (years)	0	(0–6)	(0–18)	0	(0–0)	(0–18)	0	(0–3)	(0–18)
Highest schooling attainment
none (0 years)	900	-	(60.0%)	1,901	-	(78.4%)	2,801	-	(71.4%)
primary (1–6 years)	292	-	(19.5%)	295	-	(12.2%)	587	-	(15.0%)
secondary (7+ years)	307	-	(20.5%)	229	-	(9.4%)	536	-	(13.7%)
Income estimate (USD)	1,066	(475–1,506)	(166–2,860)	757	(475–1,461)	(166–2,939)	883	(475–1,506)	(166–2,939)
Observations	1,499	-	(38.2%)	2,425	-	(61.8%)	3,924	-	(100.0%)

Notes: Individuals were surveyed in the Burkina Faso Demographic and Health Surveys (DHS) of 2003, 2010, 2014, and 2017–18. Income was estimated based on each household’s relative position in the wealth distribution of the country. No sampling weights were used for descriptive statistics. IQR: interquartile range. USD: Constant 2011 international US Dollar.

Table 5

Economic analysis: Results from OLS wealth regression models for the Boucle du Mouhoun region, Burkina Faso.

		Basic Model	Stratified by Sex		Stratified by Type of Place of Residence
Variable			Males	Females	Urban	Rural
Age		0.01*	0.02***	0.01*	0.00	0.02***
		(0.01; 0.00–0.02)	(0.01; 0.01–0.04)	(0.01; -0.00–0.03)	(0.01; -0.01–0.01)	(0.01; 0.01–0.04)
Educational Attainment
No schooling		-0.28***	-0.15***	-0.17***	-0.06***	-0.12***
		(0.02; -0.32 - -0.24)	(0.03; -0.21 - -0.10)	(0.03; -0.22 - -0.12)	(0.02; -0.09 - -0.02)	(0.03; -0.18 - -0.06)
Primary schooling		Reference group
Secondary or higher		0.26***	0.07***	0.10***	0.04***	0.30***
		(0.02; 0.22–0.30)	(0.02; 0.02–0.12)	(0.03; 0.04–0.16)	(0.01; 0.02–0.07)	(0.06; 0.19–0.41)
Observations		3,924	1,499	2,425	1,449	2,475
R-squared		0.15	0.38	0.27	0.25	0.04

Notes: A total of 3,924 individuals were surveyed in 2003, 2010, 2014, and 2017–18. The dependent variable was Ln (asset quintile). Coefficients represent the yearly increase in asset ownership on a natural logarithmic scale. Robust standard errors (SE) and 95% Confidence Interval (CI) in parentheses (SE; CI).

*** p<0.01

** p<0.05

* p<0.1. In all models we controlled for age squared; in the stratified models we additionally controlled for survey round. OLS: ordinary least squares.