Non-published World Bank study: Aid disbursements to highly aid-dependent countries coincide with sharp increases in bank deposits in offshore financial centers known for bank secrecy

Elite Capture of Foreign Aid: Evidence from Offshore Bank Accounts. Jørgen Juel Andersen, Niels Johannesen, Bob Rijkers. Conditionally accepted to the World Bank Working Paper series, December 13, 2019. https://www.nielsjohannesen.net/wp-content/uploads/AidAndHiddenWealth-FullPaper-WEB18Feb2020.pdf

Abstract: Do elites capture foreign aid? We document that aid disbursements to highly aid-dependent countries coincide with sharp increases in bank deposits in offshore financial centers known for bank secrecy and private wealth management, but not in other financial centers. The estimates are not confounded by contemporaneous shocks such as civil conflicts, natural disasters and financial crises, and are robust to instrumenting with predetermined aid commitments. The implied leakage rate is 7.5% at the sample mean and exhibits a strong correlation with the ratio of aid to GDP. Our findings are consistent with aid capture in the most aid-dependent countries.

JEL: D73; F35, P16

Excerpts:

An important feature of all of our empirical specifications is the log-transformation of foreign deposits, which captures the statistical assumption that foreign deposits change exponentially. This assumption has strong economic foundations. First, absent withdrawals and new deposits, compound interest mechanically makes account balances grow exponentially. Second, many theoretical models will predict that changes in deposits in response to changes in the economic environment, e.g. business cycles and policy interventions, are proportional to the stock of deposits. Such considerations have led almost three decades’ of literature on foreign deposits to estimate models in log-levels (Alworth and Andresen, 1992; Huizinga and Nicodeme, 2004; Johannesen, 2014; Johannesen and Zucman, 2014; Menkhoff and Miethe, 2019; OECD, 2019) or log-differences (Andersen et al., 2017).

The main disadvantage of the log-transformation is that the resulting model does not deliver the structural parameter of interest, the leakage rate, directly. It is therefore natural to consider alternatives, for instance to scale deposits by GDP. However, scaling does not preserve the appealing features of the logarithmic transformation when countries are structurally different. For instance, in case two countries exhibit a ratio of haven deposits to GDP of 2% and 10% respectively, compound interest at the rate of 5% increases the ratio of haven deposits to GDP by 0.1% in one country and by 0.5% in the other. Moreover, scaling both deposits (the dependent variable) and aid (the explanatory variable) with GDP may create a mechanical positive correlation. In light of these difficulties, we first estimate the model in log-differences and later retrieve the leakage rate with a simple back-of-the-envelope computation. 

4.1 Main Findings

We present the results from our baseline model in Table 2. Controlling for GDP growth, country fixed effects and time fixed effects, we find that aid disbursements are strongly associated with increases in haven deposits, but do not vary systematically with non-haven deposits. Specifically, as shown in Column (1), an aid disbursement equivalent to one percent of GDP in a given quarter induces a statistically significant increase in haven deposits of around 3.4%. By contrast, as shown in Column (2), the analogous effect on non-haven deposits is a statistically insignificant decrease of around 1.5%. The final specification highlights the difference: an aid disbursement equivalent to one percent of GDP is associated with a statistically significant increase in haven deposits, measured over and above the increase in non-haven deposits, of around 5%, as shown in Column (3).

The results are consistent with aid capture by ruling elites: diversion to secret accounts, either directly or through kickbacks from private sector cronies, can explain the sharp increase in money held in foreign banking centers specializing in concealment and laundering. If the transfers to havens were caused by confounding shocks correlating with aid disbursements, we should expect to see similar transfers to other foreign banking centers; however, there is no evidence of such responses.