Volatile Capital Flows and Economic Growth: The Role of Macroprudential Regulation

by | Feb 12, 2016 | Papers | 12 comments

Kyriakos C. Neanidis

Macroprudential policies, their use, implementation and effectiveness, have been at the centre of a heated debate since the onset of the global financial crisis. This note sets the emphasis on the long-run growth effects of financial regulation. It finds that macroprudential regulation promotes economic growth by mitigating the adverse effects of financial volatility.

Macroprudential regulation and economic growth

The global financial and economic crisis of 2007–09 has highlighted weaknesses in macroeconomic and regulatory practices and market failures that contributed to a buildup of systemic risks. At the international level, this led to the setup of macro-prudential oversight frameworks with the aim to contain systemic risks and achieve greater financial stability, and in this way reduce the adverse consequences of financial volatility for the real economy.

Although recent work has examined the effectiveness of macroprudential regulation in reducing systemic risk and financial instability, by focusing on the credit and housing markets, little is known about the effectiveness of these rules on the broader objective of economic growth. To close this gap, in a new paper (Neanidis, 2015) we examine empirically the relationship among macroprudential regulation, financial volatility, and economic growth. In particular, we assess the success of macroprudential policy in reducing systemic risks by dampening the procyclicality and volatility of financial flows, expected to give rise to a growth-promoting effect.

Broadly defining financial volatility as the volatility of international capital flows and using aggregated indicators of macroprudential regulation, we investigate the effect of volatile capital flows on growth in the presence of macroprudential policies. In our panel data framework, the sample covers about 80 countries over the period 1973-2013.


The results indicate that although more variable capital flows reduce economic growth, macroprudential regulation mitigates this negative growth effect. This means that macroprudential policies, by encouraging a greater buildup of buffers, attenuate the adverse growth effects of unstable capital flows and, by so doing, are effective in limiting financial system vulnerabilities.

Further findings are summarised as follows:

  • The outcomes are mainly restricted in the sample of middle-income countries, since it is this group of countries that have relied more on macroprudential policies.
  • In Sub-Saharan Africa and, within it, Francophone SSA, macroprudential regulation has the capacity to attenuate the growth-distorting effect of volatile flows to a much greater degree.
  • The effectiveness of macroprudential regulation diminishes in magnitude in economies that are relatively open, with deeper financial systems, and exposed to greater macroeconomic instability.

Policy implications

Traditionally economists have examined the implications of monetary and fiscal policies for economic growth. With the recent design of regulatory frameworks in the form of macroprudential rules, another set of policies has emerged that could be used to enhance economic growth by simultaneously ensuring financial stability. Our analysis shows that macroprudential regulation can achieve this dual objective making it an important part of a policymaker’s toolkit, especially for countries exposed to large and volatile movements in financial flows. This, in turn, justifies efforts for international cooperation and coordination in setting macroprudential rules and standards as a way of combating and minimizing financial volatility and its consequences on the real economy.



  1. Guillaumont Jeanneney Sylviane

    Ce papier est très intéressant et le résumé remarquable ; c’est un énorme travail empirique
    Quelques questions cependant
    Les taux de croissance sont calculés en moyenne sur trois ans : on ne sait plus bien si on se situe à court terme ou à long terme. Le calcul de la volatilité sur trois ans a-t-elle une pertinence ? Que se passe-t-il si on raisonne sur une moyenne comportant plus d’années ?
    La définition de la volatilité : qu’appelle-t-on les flux normalisés ? (tenant compte du trend ?)
    IL y a sans doute une collinarité entre les flux de capitaux et certaines variables de contrôle, notamment le taux d’investissement et l’ouverture. Cela n’a-t- il pas un effet sur les résultats relatifs aux mouvements de capitaux ?
    Il aurait été intéressant de bien faire apparaître la différence entre les capitaux à long terme et les capitaux à court terme.

    In English
    This paper is very interesting and the summary excellent. It is a huge econometric work!
    However some modest questions:

    The rates of growth are calculated on three years average. Is it short term or long term perspective? Is the calculation of volatility on three years relevant? What would be the results if you use a longer period?
    Explain “normalized flows” (relative to their trend?)

    There is potentially a correlation between capital flows and some control variables (rate of investment, openness) which can impact the capital flows effect?

    It would be interesting to focus on the difference between short term and long term flows.

    • Kyriakos Neanidis

      Dear Sylviane,

      Many thanks for your nice words and for your questions.

      The study, by examining the growth implication of volatile capital flows, focuses on the long run. For this reason, the data are averaged over a number of periods to avoid the usual concerns around business cycles effects. An alternative would be to take a moving period average of the dependent variable only (i.e., growth rate of GDP per capita), but I have decided to average all variables as this is the norm in the related literature.

      Further, the data are averaged over three years because, as I state in a footnote, “The use of three-year averages represents a compromise between the need to focus on long-term
      relationships and the need to maximize the time-series (within-country) variation in the data.” Typically, studies use five year period averages, but four or three year period averages are also common. Thus, I do not view this as a major concern, even for the use of three year averages for the volatility of capital flows.

      All types of a country’s capital flows are normalized by the country’s GDP. Thus, capital flows correspond to their fractions as GDP, while the volatility of capital flows represents the standard deviation of the normalized flows. Again, this is standard practice, although I accept other measures of volatility could be also used (note I also use the coefficient of variation of the normalized capital flows in my robustness testing).

      I do accept that capital flows and some of the other control variables may be correlated, but this is typically the case with most macroeconomic variables in cross-country growth regressions. I think that including other correlates as controls allows for the effect of capital flows to be “net” of the links between the other correlates and growth, thus refining the magnitude of the effect of capital flows on growth.

      Finally, I agree that it is interesting to look at the difference between short term and long term flows. In fact, the literature has been focusing on the former, while the current study represents the first effort to focus on the latter. This is where one of the contributions lies, the other being the explicit exploration of the role of macroprudential regulation on the link between volatile flows and economic growth.


  2. Emmanuel Pinto Moreira

    Hi Kyriakos, this is Emmanuel Pinto Moreira, Lead Economist at the World Bank and member of the Project’s International Advisory Committee, writing from Kinshasa.

    I read your paper with great interest. From the ground here, can you summarize in a few lines what are its key policy implications for low-income countries, where supervisory skills are limited?



    • Kyriakos Neanidis

      Dear Emmanuel,

      Thank you for your question. In responding, I outline the study’s key implications below.

      • Macroprudential regulation tools are effective in curbing the negative growth effects of capital flows that are volatile in nature.

      • This effect applies mainly to middle-income countries and countries located in Sub-Saharan Africa (SSA), with the effect being greater in magnitude for the latter group. This means that these groups of countries have more to gain from the use of macroprudential policies with reference to the particular channel under investigation.

      The data utilized do not include information regarding the intensity of macroprudential regulation instruments but only of their use. As a result, the application of such tools is effective even if it is not taken to its full intensity. This is important when the capacity for supervision from relevant authorities is limited.

      A final point, currently not appearing in the paper but an outcome of further analysis I conducted, has to do with the type of macroprudential regulation instruments most effective in this framework. The mitigating effect unveiled in the analysis, with regard to total capital flows, is especially present when using the following macroprudential tools:

      • Loan-to-Value Ratio Caps which restricts to LTV used as a strictly enforced cap on new loans, as opposed to a supervisory guideline or merely a determinant of risk weights;

      • Leverage Ratio which limits banks from exceeding a fixed minimum leverage ratio;

      • Limits on Interbank Exposures which limits the fraction of liabilities held by the banking sector or by individual banks;

      • Concentration Limits which limits the fraction of assets held by a limited number of borrowers;

      • Limits on Foreign Currency Loans which reduces vulnerability to foreign-currency risks.

      Following from this, then, authorities may be better off enacting these particular instruments if their objective is to attenuate the distortive effects of volatile financial flows on economic growth.


  3. Issouf Samake

    Dear Kyriakos,

    This is Issouf Samake, Senior Economist at the International Monetary Fund. Thank you for the article and I would like to take this opportunity to discuss your well-written paper , “Volatile Capital Flows and Economic Growth: The Role of Macro-prudential Regulation”. Please see below a few comments for your consideration.

    Comment 1 – Appropriateness of the econometric model. A key contribution of the paper is to emphasize on the long-term effects of financial regulation. To that end, you have relied on a dynamic GMM model. However, it is not clear how the dynamic empirical model is linked/fitted to the static long-run model as presented in equation 1. More specifically, given the focus on long-run property of financial regulation, I was wondering why you did not estimate a LSDV and then improve (the LSDV results) with the estimation of a corrected LSDV which, in such a case, could be more efficient than both GMM system and difference GMM.

    Comment 2 – Introduction of MPR in the model. Given that the primary objective of macro-prudential policies is to contain systemic vulnerability, I was wondering whether you could discuss the appropriateness of the model specified by addressing why the current model set up would be better rather than considering: (1) the introduction of (new) MPR (or change in MPR) as opposed to the existing MPR in the model; and (2) output volatility as the dependant variable.

    Comment 3 – Economic consideration of the endogeneity. While econometrically GMM (or LSDV Corrected) addresses endogeneity problem, economically the control (or introduction) of fiscal policy variable would justify how the issues of endogeneity could be addressed. This is because of the well-known direct link between fiscal policy and financial stability (e.g., inconsistent fiscal policy could be associated with unsafe sovereign debt leading to an adverse feedback loop between sovereign risk and systemic vulnerability).

    Hope these would be of help, and I look forward to hearing from you.

    Very best,
    Issouf Samake

    • Kyriakos Neanidis

      Dear Issouf,

      Many thanks for your comments, to which I respond below.

      Comment 1
      Indeed, the key contribution of the paper is to emphasize the long-term effects of macroprudential regulation, although in an indirect manner via its impact on volatile capital flows. I first present OLS regressions as it is standard to show such results for comparison with those of more advanced techniques, such as dynamic GMM regressions. The latter have been used in the empirical growth literature by an increasing number of researchers (e.g., Beck et al., 2000; Roodman 2007). The GMM estimations are the most appropriate since they are based on techniques that control for (i) potential endogeneity of the regressors, (ii) region-specific effects, and (iii) heteroskedasticity and autocorrelation within regions. Further, I compute the finite-sample correction to the two-step covariance matrix derived by Windmeijer (2005) which allows for the estimation of robust standard errors. The above points, so long as the regressions pass the diagnostic tests (which they do), suggest that the dynamic GMM estimator has superior finite sample properties. This is the reason I prefer this estimator, which appears to be delivering a consistent finding supporting the main thesis of the paper.

      Comment 2
      The current regression specification, outlined in equation (1), allows estimating the indirect effect of MPR on economic growth focusing on an interaction term between MPR and the volatility of financial flows. So, the aim is to examine whether MPR policies can attenuate the negative growth effect of volatile capital flows. The positive coefficient estimate of the interaction term supports this rationale.

      Using a different specification that considers a change in MPR instead, as you suggest under (1), would be picking up a different effect: how changes in MPR policies could minimize the distorting growth effect of volatile flows. This would be a different question as the focus would be on the importance of changing MPR rules rather than their levels, which I am considering. Also, keep in mind that many countries have instituted MPR rules only recently, so examining the effects of MPR changes may be premature at this stage. We first need to understand the level effect of these rules before we turn to the implications of their changes.

      Your second suggestion, using output volatility as the dependent variable, could also be pursued as it would give a direct measure of the success, or otherwise, of MPR policies in limiting output volatility. This, in my mind, would represent a short-run analysis of the effects of such policies, i.e., their impact on the business cycle. The focus, however, of the current study is in the long run by examining the growth implications of such policies. Obviously, one study cannot address a multitude of objectives.

      Comment 3
      Although one can accept the rationale you present, this would only represent one channel via which the link between fiscal policy and financial stability takes shape. For instance, in the regressions I only control for government consumption as a fiscal policy variable. There are other aspects of fiscal policy one could also account for in a more systematic way. Due to this, I prefer to explicitly control for endogeneity concerns with the use of the appropriate techniques in this setting, i.e., the dynamic GMM estimator.


  4. Eugenia Vella

    Hi Kyriakos,

    This is Eugenia Vella, Lecturer at the University of Sheffield and Fellow at the Autonomous University of Barcelona.

    I very much enjoyed reading your work. I was wondering whether you could comment further on lessons drawn for European countries (or Euro-area countries).

    Kind regards,

    • Kyriakos Neanidis

      Dear Eugenia,

      Many thanks for your question.

      If one accepts that European, or Euro-area countries, belong in the high-income group of countries, the take away message of the paper is that for these countries neither volatile flows retard growth, nor does financial regulation change the impact of volatile flows. The main explanation for this finding is that advanced economies tend to have more developed financial systems which offer various alternative sources of finance and scope for avoidance, making it possibly harder for macro-prudential policies to be effective.

      Following on your comment, however, I have run some additional regressions to examine explicitly the related effects for the member countries of the European Union and of the Euro-area sub-group. The results point out that both country groups benefit from macroprudential regulation tools in the same way as the average country in the sample, by curbing the negative growth effects of volatile capital flows. Hence, the lesson to be drawn from this result is that country members of these two institutions can only benefit by the application of macroprudential policies, assuming their objectives are limiting financial vulnerabilities and expanding economic growth.


  5. Rangan Gupta

    Dear Kyriakos,

    As always great work!

    Few questions: (1) Is it possible to analyze the results by categorizing countries based on high and low volatility, i.e., say you look at the distribution of the volatility variable and then categorize based on above or below the mean or median? Would that matter? In some sense, this would tell you whether the relationship is actually linear or whether the role of MPR is conditional on the level of volatility;

    (2) You have categorized the countries based on high, middle- and low income. May be you could do the same based on quantile estimation of your model, which would allow you to look at the entire conditional distribution of the dependent variable, without actually categorizing countries;


    Department of Economics
    University of Pretoria

    • Kyriakos Neanidis

      Dear Rangan,

      Many thanks for your kind words and comments. Please see below for responses.

      Comment 1
      It is indeed possible that at the margin the diminishing effect MPR has on the negative growth impact of volatile capital flows may vary at different levels of the distribution of the volatility variable. This, however, would not come as surprise as the effectiveness of MPR policies may as well depend on how volatile financial flows are. Re-running the system-GMM regression with total capital flows (Table 2, column 5) where I now split the sample according to the mean or median of the distribution of the volatility variable, yields once again a significantly positive coefficient in the interaction term (with the volatility coefficient being once again negative). This suggests that the relationship remains intact and does not depend on the level of financial flows volatility. In fact, when using the sample median as the threshold value, the coefficient estimates of the interaction term are similar in magnitude in the two sub-samples (0.5 vs. 0.6). I find similar results when repeating the exercise for the three components of capital flows (Table 2, column 6-8). I believe these findings are reassuring of the importance of MPR policies in this framework.

      Comment 2
      The division of the country sample by income following the categorization to high-, middle-, and low-income countries is natural given the historical tendency of middle-income countries to rely more on MPR policies. This grouping is also consistent with the observation that high-income countries tend to have more developed financial systems, making it possibly harder for macro-prudential policies to be effective in this set of countries. Both these arguments are indicated in the paper.

      A further reason for following the country grouping into high-, middle-, and low-income is that a quantile estimation would not provide much intuition as to why we observe the results we would obtain, whereas with the above categorization the intuition is clear. I think this point is best illustrated when, following your suggestion, I use a 2-quantile regression by splitting the sample at the median observation of GDP per capita (of the regression in Table 2, column 5). The findings indicate that for both sub-samples results are consistent as with the full sample, while the coefficient estimates of the interaction term are exactly the same, 0.5. This suggests that dividing countries by the sample median of GDP per capita would provide a “biased” estimation of the effectiveness of MPR rules, possibly because the estimates would reflect the set of middle-income countries dominating the two sides of the median value. To avoid such concerns, I think the grouping of countries according to income represents a good choice in this case.


  6. Keyra Primus

    Dear Kyriakos,
    This is a very good paper. I enjoyed reading it and I underscore the important contribution of this work to examine whether macroprudential regulation policies can mitigate the adverse effects of volatile capital flows on long-run economic growth.
    I understand that your findings are based on the use of aggregate macroprudential regulation instruments. Would you also be able to comment on which individual macroprudential regulation tools are likely to be more effective in the various country groupings (low-, middle-, and high-income)? Thanks.

    Keyra Primus
    International Monetary Fund

    • Kyriakos Neanidis

      Dear Keyra,

      Many thanks for your kind words and comments. Please see below for my response.

      Indeed, the data on macroprudential regulation I use in the paper are based on aggregate indicators, thus, findings should be viewed in this light. From the three aggregate indices I use, based on Abiad et al. (2008), Cerutti et al. (2015), and Barth et al. (2013), the only index that can be disaggregated in a meaningful way is that by Cerutti et al. (2015). The index by Abiad et al. (2008) cannot be disaggregated as there are no (publicly) available data that allow further decomposition, while that from Barth et al. (2013) can be decomposed further but with only four observations in the time dimension, I do not find this a meaningful exercise.

      Thus, my report below is based on the decomposition of the GMPI, broken down in its twelve individual macroprudential instruments.

      To examine the effectiveness of each of the twelve indicators by income groupings, I had to create interaction terms between each of the indicators with each of the three income groups (low-, middle-, and high-income) as in Table 5, column 1. Running regressions with this objective in mind, by using total capital flows as measure of financial flows, gives rise to the following indicative findings:

      • In high-income countries, the instruments that are effective are:
      o Limits on Foreign Currency Loans which reduces vulnerability to foreign-currency risks;
      o Limits on Domestic Currency Loans which limits credit growth directly.

      • In medium-income countries, the instruments that are effective are:
      o General Countercyclical Capital Buffer/Requirement which requires banks to hold more capital during upturns;
      o Concentration Limits which limit the fraction of assets held by a limited number of borrowers;
      o Foreign Currency and/or Countercyclical Reserve Requirements which restricts to RR when i) imposes a wedge on foreign currency deposits, or ii) is adjusted countercyclically.

      • No disaggregated instrument is effective in low-income countries.

      Following from this, then, authorities in high- and middle-income countries may be better off adopting particular macroprudential instruments as outlined above. One can view this finding as giving to countries a “menu of choices” as to which macroprudential instrument(s) to use contingent on their economic development so long their objective is to attenuate the distortive effects of volatile financial flows on economic growth.



Submit a Comment

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