Aid Volatility, Human Capital, and Growth

by | Jun 6, 2016 | Papers | 6 comments

Pierre-Richard Agénor

Aid remains an important component of capital flows to low-income countries. Empirical contributions show that there is robust statistical evidence that aid volatility tends to have an adverse effect on economic growth. However, the channels through which such volatility operates have not been fully articulated in endogenous growth models. This paper studies the effect of aid volatility on growth, in a model where the decision to invest in skills is endogenous. The analysis focuses on a low-income economy where the cost of acquiring education benefits from public subsidies, which are partly financed through foreign aid.

Aid Volatility, Wages and Skills Acquisition

Empirical contributions show that there is robust statistical evidence that aid volatility tends to have an adverse effect on economic growth. However, the channels through which such volatility operates have not been fully articulated in endogenous growth models. The author develops a stochastic growth model where skills acquisition is endogenous. Specifically, aid volatility may adversely affect growth (and possibly welfare) when the decision to invest in skills is endogenous. To understand how this can occur, consider a low-income economy where the cost of acquiring education benefits from public subsidies, which are partly financed through domestic taxes and partly through aid. Aid is subject to random shocks.


Using a combination of analytical derivations and numerical experiments, the key insights derived are as follows:

  • By creating uncertainty about the net return to education, a high degree of aid volatility mitigates agents’ incentives to invest in skills.
  • If savings and growth depend on the composition of the labor force, and if more skilled workers are more productive, aid volatility may therefore have an adverse effect on the mean growth rates of investment and output.
  • Aid volatility is also bad for the welfare of skilled households (directly) and unskilled households (indirectly, through output volatility).

Policy Lessons

Aid volatility creates significant macroeconomic management challenges for recipient governments in low-income countries, whose ability to raise resources through domestic taxation and to borrow on domestic and international capital markets is limited. The education incentive-human capital and physical capital channels highlighted in this paper offer an alternative view regarding the potential impact of aid volatility on growth in low-income countries. When promised aid is not provided or when additional aid is disbursed unexpectedly, productive public spending may need to be adjusted abruptly with potentially large social and economic costs.

To improve aid predictability, two approaches can be considered. The first has been to urge recipients to protect themselves from fickle donors by saving aid windfalls in a reserve or stabilization fund. The second approach is to promote more stable donor-recipient relationships, that is, to encourage donors to move away from fragmented, conditionality-based funding and make multi-year pre-commitments, with appropriate safeguards, to ensure a longer time horizon (Eifert and Gelb (2006)).




    Bonjour, ici Emmanuel Pinto Moreira, Lead Economist de la Banque mondiale en poste à Kinshasa, RDC. J’ai une question pour le Professeur Agénor, relative à son étude “Aid Volatility, Human Capital and Growth.” Je trouve l’argument principal de l’étude (un effet adverse de la volatilité sur le rendement de l’éducation) très pertinent sur le plan analytique. Mais comment peut-on quantifier son importance en pratique, compte tenu des autres effets possibles de la volatilité de l’aide (notamment sur les investissements pluriannuels en infrastructure, comme le Professeur l’a montré dans un article antérieur)? Merci.

    • Pierre-Richard Agénor

      Cette question est très pertinente. En effet, l’étude prédit non seulement une relation inverse entre la volatilité de l’aide et la croissance économique, mais également une relation inverse entre la volatilité de l’aide et le niveau de qualification de la population active. En d’autres termes, toutes choses égales par ailleurs, et en tenant compte de l’effet du NIVEAU de l’aide et des autres determinants des niveaux d’éducation (comme par exemple la composition des dépenses d’éducation, la qualité de l’enseignement, etc.) les pays où la volatilité de l’aide productive est la plus élevée devraient avoir les ratios de main d’oeuvre qualifiée-non qualifiée les plus faibles. Cette prédiction pourrait être testée en utilisant des régressions sur données de panel, avec par exemple le ratio entre le taux d’éducation tertiaire et le ratio du taux d’éducation secondaire comme variable expliquée. C’est un travail qui pourrait demander beaucoup d’effort, mais j’espère qu’il sera poursuivi dans un futur proche.

      Voici un résumé de la question et de ma réponse en anglais :

      QUESTION: How can one quantify the adverse effect of aid volatility on education outcomes?

      ANSWER: This is an important question. Indeed, in addition to predicting a negative relationship between (productive) aid volatility and growth, the paper suggests a new testable implication: all else equal, and controlling for the positive effect of the LEVEL of aid and other determinants of education outcomes (such as the composition of public spending on education, the quality of schooling, and so on), countries where the volatility of (productive) aid is the highest should also have the lowest ratio of skilled-unskilled workers in the labour force. This can be tested by using panel data regressions with the ratio of tertiary to primary and secondary enrolment rates as the dependent variable. This would require significant work but I hope that somebody will follow on this idea in the near future.


    This is Issouf Samake, Senior Economist at the International Monetary Fund. Professor Agénor, further to your response to Dr. Pinto Moreira, it seems to me that one would need to carefully disaggregate productive aid into education and other components to test the proposition that aid volatility can be detrimental to growth as a result of its adverse impact on the relative rate of return to skills acquisition. However, I am not sure that this is possible in practice; if aid flows transit through the government budget and productive components are not explicitly tied to specific activities, identification may not be feasible. Even if this is the case, one would need to account for the fact that higher aid dedicated to the education sector may substitute for government education spending based on its own resources. As a result, total spending on (higher) education may not change, implying that identifying a link between aid volatility and educational outcomes may not be possible. While I focus on education for the sake of argument, the above reference applies to (instead of education) health, or any productive spending. I would appreciate your reaction.

    • Pierre-Richard Agénor

      These are very valid points. The formal model assumes, conveniently of course, that there is only one form of aid, which is used in its entirety to subsidize education. In practice, funds are fungible; and aid covers a range of areas, both productive (e.g., infrastructure or health, in addition to education) and unproductive. Even if one were to neglect moral hazard issues (the fact that aid may “displace” public spending that would have taken place any way by using domestic resources), there is a serious problem of getting adequate data to test precisely the implications of the analysis–namely, the fact that the volatility of productive aid (in the form of direct or implicit subsidies to training) may have an adverse effect on education outcomes and growth.
      However, even though building an accurate database is not easy, this is not an impossible task. I would focus on a relatively homogeneous group of countries (e.g., members of the CFA franc zone in Sub-Saharan Africa) and would start by compiling data on aid to (higher) education from the OECD database, which is fairly disaggregated. Aid volatility measures (e.g., moving averages of coefficients of variation) can be calculated from these data. I would then use national statistics to estimate total spending and compile data on education outcomes (namely, the composition of the labour force). Panel data regression methods could then be used to estimate the relationship of interest. This is a lot of work for sure but would be well worth the effort.

  3. Baris Alpaslan

    Hi, this is Dr. Baris Alpaslan, from Ankara Yildirim Beyazit University, Turkey. Professor Agénor, I have read your paper with great interest, my question is–would the same argument about the adverse effect of aid volatility on human capital accumulation and growth apply to wage volatility induced by output shocks?

    Thank you very much for your answer in advance.

    • Professor Pierre-Richard Agénor

      hank you, the answer to the question is a favourite one for economists–it depends. If output shocks–which tend to be quite large in low-income countries, as documented in a number of studies–induce higher volatility of skilled wages relative to unskilled wages, then indeed the increase in the volatility of the expected return from education would tend to have an adverse effect on the decision to acquire skills and thus economic growth. In that sense, yes, the effect of output volatility would be qualitatively similar to the effect of aid volatility in the model, because both affect the volatility of the NET rate of return to education.

      However, it is also possible that output shocks translate into an increase in volatility of BOTH skilled and unskilled wages of the same magnitude; given that it is relative wage volatility that matters, the net effect on the decision to acquire skills may be negligible, with consequently no effect on human capital accumulation and economic growth.


Submit a Comment

Your email address will not be published. Required fields are marked *