VoxEU Column Development Poverty and Income Inequality

IMF loans and offshore financial flows

Recent research suggests that World Bank aid disbursements are associated with outflows from recipient countries to offshore financial centres, indicating elite capture of aid. This column uses 25 years of data to examine whether the same is true for IMF lending. It finds no evidence that IMF loans are diverted to offshore bank accounts. This could be because IMF lending differs in structural respects – such as conditionality, concessionality, and continuity – from World Bank aid.

Are aid packages from international financial institutions captured by local elites and diverted to offshore bank accounts? In a recent paper, Andersen et al. (2020) find that aid disbursements from the World Bank are associated with greater outflows from recipient countries to offshore financial centres. They argue that this provides a plausible measure of aid diversion since offshore financial centres are characterised by bank secrecy and asset protection.

Moreover, the ownership of bank accounts in offshore financial centres is overwhelmingly concentrated among the richest members of society (Alstadstaeter 2019). So a differential increase in outflows to offshore financial centres, relative to other international destinations, is likely to indicate that World Bank aid is being appropriated by local elites and stashed away abroad. 

While it is possible that aid could structurally alter the nature of corruption in the recipient country (Dreher et al. 2010), material leakage from the flow of aid directly prevents the aid from being used for the intended purpose and raises a practical accountability concern. 

The novelty of linking cross-border flows with aid disbursements is that it allows us to quantify the extent of aid diversion, a topic previously understudied due to the lack of high-frequency data.

It is plausible that different types of aid, or aid provided by different institutions, are subject to different degrees of diversion; if this is the case, we would expect corresponding heterogeneity in aid effectiveness. We investigate this question and explore if the same is true for IMF loan disbursements (Aiyar and Patnam 2021).

IMF loans differ in many important respects from World Bank aid. IMF programmes are episodic, not continuous. Rather than financing development projects, IMF assistance provides financing for governments faced with balance-of-payments crises or other macroeconomic shocks (see Dreher et al. 2016 for an analysis of the differences between budget and project aid). 

IMF loans carry conditionality, with the recipient countries expected to meet quantitative targets – such as ceilings on fiscal expenditure or floors on international reserves – or structural benchmarks such as labour market reforms. Finally, IMF programmes are less concessional than World Bank aid (although there is heterogeneity along this dimension). Given these manifold differences, we do not necessarily expect to find identical, or even similar, results for IMF lending. And indeed, we do not.

Baseline results

Our study design is simple. Using data from the Bank of International Settlements, we examine whether countries that receive an IMF loan disbursement show increased outflows to offshore financial centres in that quarter. The baseline results, estimated using OLS, are shown in Figure 1. 

Aid diversion would imply a positive, significant coefficient on outflows to offshore financial centres; instead, we find a negative coefficient, as shown by the first line segment. Flows to other international destinations are not affected, as illustrated by the second line segment (which encompasses zero). The third line segment shows the impact of a loan disbursement on outflows to offshore financial centres relative to outflows to other international destinations; again, the coefficient is negative rather than positive.

Figure 1 Effect of receiving IMF disbursement on growth rate of international flows


Notes: The figure plots the point estimates (white dot) and 95% confidence intervals (in bars) for the effect of IMF disbursements on offshore (blue), non-offshore (red) and differential (green) flows.

We additionally investigate if there are anticipation effects or lags pertaining to IMF loan disbursements. If programme disbursements are anticipated or some aid diversion occurs with a lag, then an increase in offshore flows might occur before or after the disbursement quarter. We re-run our baseline specification after adding lags and leads of up to four quarters and find no evidence of any anticipated or lagged effects, with a negative and statistically significant effect holding only in the quarter of disbursement.

Dealing with endogeneity

By design, the IMF lends to countries facing economic stress. So there is an econometric challenge in separating the impact of the IMF disbursements on offshore flows from the impact of the underlying macroeconomic disturbance on these flows. We therefore supplement our baseline results with two alternative strategies.

First, we restrict the sample to quarters in which the recipient country was under an IMF programme. Loan disbursements do not occur every quarter, but presumably the degree of macroeconomic stress should be similar for disbursement and non-disbursement quarters, conditional on an IMF programme being in place.

Second, we use the ex-ante schedule of loan disbursements as an instrument for the actual loan disbursements. In practice, the two can diverge for a number of reasons, such as non-fulfilment of programme conditionality or new shocks. Since the ex-ante schedule of disbursements should not be correlated with future shocks, the instrument is plausible. Once again, we find no evidence that IMF loan disbursements are associated with greater outflows to offshore financial centres.

Does the effect depend on programme type or loan size?

It is possible that aid diversion could work differently for different types of programmes. The IMF Poverty Reduction and Growth Trust facility provides concessional lending to low-income countries on terms that are much more comparable to World Bank aid than standard loans under the General Resources Account facility.

It is also possible that aid diversion is non-linear, only occurring when the assistance package reaches a large enough size. We find little evidence to support either hypothesis (Figure 2). Although the point estimate is positive for loan amounts over 2.5% of quarterly GDP, it is not significantly different from zero.

Figure 2 Heterogeneous effect of receiving IMF disbursement on growth rate of international flows (differential flows)

a) By programme type


b) By lending amount


Notes: The figures plot the point estimates (white dot) and 95% confidence intervals (in bars) for the effect of IMF disbursements on differential flows, by programme type and lending amount.

The COVID-19 lending surge

In response to the COVID-19 pandemic, IMF lending has recently hit record levels. Moreover, much of this surge has been in the form of emergency assistance, which carries lighter conditionality than other types of IMF lending. Therefore, concerns about governance are elevated; if IMF assistance is diverted to offshore bank accounts, then it cannot be used to fight the pandemic.

Figure 3 IMF lending amounts (quarterly, USD billion)


Notes: The figure shows quarterly IMF disbursements. ‘Emergency Arrangement’ is RCF/RFI/Emergency-Assistance Facilities. Source: IMF Finance Department.

From an econometric point of view, the pandemic provides a quasi-natural experiment. We restrict the sample to the second and third quarters of 2020, when IMF emergency lending peaked in response to the COVID-19 pandemic. As in the case of the longer time series, we are unable to detect any significant impact of IMF loan disbursements on offshore flows (or indeed, to other destinations, or the difference between the two).

Figure 4 Effect of receiving IMF disbursement on growth rate of international flows during COVID-19 (sample restricted to 2020 Q2/Q3)


Notes: The figure plots the point estimates (white dot) and 95% confidence intervals (in bars) for the effect of IMF disbursements on offshore (blue), non-offshore (red) and differential (green) flows.


Our study shows that, unlike World Bank aid, IMF loan disbursements are not associated with greater outflows to offshore bank accounts. We hope that this result will stimulate further research into what features of international assistance are crucial to ‘appropriability’. Does conditionality make aid harder to appropriate? Are funds provided to central banks less prone to corrupt uses than money provided for individual development projects? More broadly, how does multilateral aid stack up against bilateral aid when it comes to elite capture?

In our view, these are first-order questions. While there is ample research on the structural relationship between aid and corruption – or the lack thereof – the question of aid diversion has hardly been studied. If aid is captured by local elites and stashed away in offshore bank accounts, it cannot serve whatever purpose it was intended for. Understanding the mechanisms and features of aid diversion is therefore crucial.

Editors’ note: The views expressed herein are those of the authors and should not be attributed to the IMF, its Executive Board, or its management.


Aiyar, S, and M Patnam (2021), “IMF programs and financial flows to offshore centers”, IMF Working Paper WP/21/146.

Alstadsæter, A, N Johannesen and G Zucman (2019), “Tax evasion and inequality”, American Economic Review 109(6): 2073–103.

Andersen, J J, N Johannesen and B Rijkers (2020), “Elite capture of foreign aid: Evidence from offshore bank accounts”, World Bank Policy Research Working Paper 9150.

Dreher, A, P Nunnenkamp and H Ohler (2010), “Fighting corruption through performance-based aid”,, 26 November.

Dreher, A, S Langlotz and S Marchesi (2016), “Budget versus project aid: A trade-off between control and efficiency”,, 2 December.

1,680 Reads