Measuring capital flight : a case study of Mexico

Policy Reserch WORKING PAPERS L? Country Opersons Country Department II Latin America and the Caribbean The World Bank March 1993 WPS 1121 Measuring Capital Flight A Case Study of Mexico Harald Eggerstedt Rebecca Brideau Hall and Sweder van Wijnbergen The wrong conceptual approach can distort estimates of capital flight. The debt-stock-based estimates widely used in discus- sions of Mexico's debt crisis largely understated capital flight in the early 1980s, but overstated it in the mid-1980s. Other estimates significantly understate the foreign asset accumula- tion in the second half of the 1980s by not including interest earnings. PoIiyResarchWokingPapeaditathcflndinpofwwkinprow= and eae ge dheze fidasmanagank 9fand a1lohedin devdopmtiThcapa papdiaisbetedbythReaurchAdviy StaffcanYthenamesofhe autars.flect aydicirvis.andaou1dbouaeddcitkdaocor&y.hofindI in ,mptadiwiaandCcc1uaionsrathcauthoowiLTboyshould nottihutd tothb WoddBDank.its BoardoD suimn-agcmito-my ofitsmnberotres  Polley Re"arch Country Operations WPS 1121 This paper- a product of the Country Operations I Division, Country Department II, Latin America and the Caribbean - is part of a larger effort in the department to understand the macroeconomic effects of capital flows. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Hedwig Abbey, room D8-099, extension 80512 (March 1993, 52 pages). Eggerstedt, Hall, and van Wijnbergen show how The authors contend that: the various methods commonly used to measure capital flight produce vastly different estimates * Introducing debt stock data into the analysis (with a 100 percent difference between the -instead of the changes in debt recorded lowest and the highest, in Mexico's case). They directly in the balance of payments - requires emphasize the importance of the conceptual many difficult adjustments and should be approach to its measurement. avoided. First, they did not try to Weparate "nonnal" * Foreign asset changes of public corporations capital flows from capital flight. A capital shift must be subtracted. outward because of expected taxation is as much a response to anticipated developments in rate of * Rather than eliminate interest received on return as is a shift out in response to lower foreign assets from the current account, as some interest rates at home. have dorne, eamnings on private assets held abroad should be considered part of the "flight capital" Nor is it satisfactory to directly measure that might have been repatriated, given different capital flight by taking short-term asset changes incentives and macroeconomic conditions. and the balance of errors and omissions from the balance of payments. Neither is necessarily * The effect on capital flight of the faking of related to the unreported private accumulation of trade invoices should be assessed, since import foreign assets. overinvoicing and export underinvoicing can be used to channel capital abroad. They chose the residual approach, which assumes that capital inflows in the fonn of They demonstrate the empirical importance increases in extemal indebtedness and foreign of these choices with a new set of capital flight direct investment should finance either the estimates for Mexico, based on the recommenda- current account or reserve accumulation; any tions they present. They contrast the results of shortfall in reported use can be attributed to this approach with those of other approaches, to capital flight. demonstrate the effect of conceptual choices. Implementing the residual approach requires careful data selection and several adjusuments. The Policy Research Working Paper Series disseminates the findings of workmnder way in theBank. An objective ofthe series is to get these findings out quickly, even if presentations. are less than fully polished. The findings. interpretaticns, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center  Measuring Capital Flight: a Case Study of Mexico by Harald Eggerstedt World Bank, Rebecca Brideau Hall World Bank, and Sweder van Wijnbergen World Bank and CEPR We are indebted to Daniel Oks and Sergio Fadl for many helpful discussions and assistance in interpreting and collecting data used in this paper.  Table of Contents 1. Introduction 2 2. Measuring Capital Flight 4 2.1 Measuring Capital Flight: The Direct Approach .4 2.2 Measuring Capital Flight: The Indirect Approach 7 2.3 Scope of the Concept of Capital Flight 8 2.4 Debt Flows Versus Differences in Measured Debt Stocks 11 2.5 Interest Eaniungs on Flight Capital as Capital Flight 12 2.6 Faked Trade Invoices and Capital Flight 14 3. A New Estimate of Capital Flight 16 3.1 Adjustment for Non-Private Asset Changes 17 3.2 Adjustment for Trade Misinvoicing 17 3.3 Including Returns on Private Assets Held Abroad 20 3.4 Capital Flight from Mexico, 1960-90 22 4. Alternative Approaches 23 4.1 Distortions Resulting from the Use of Debt Stock Data 23 4.2 Consequences of Neglecting Investment Returns 29 4.3 Trade Misinvoicing 30 4.4 Errors- and Omissions-Based Estimates 31 5. Conclusions 32 References 33 Appendix I The Data :Qase 37 Appendix II Alternate Empirical Estimates of Capital Flight 39 Statistical Annex 45  2 1 Introduction Capital flight, loosely defined as unreported private accumulation of foreign assets, is by its very definition difficult to measure. Yet a reasonably accurate estimate of its magnitude is important for a proper diagnosis of and prescription for many macroeconomic ills. Discussions about the extent of capital flight have been particularly important in the debate about debt relief. A $100 USb debt, like Mexico's in 1989, is difficult to present as a national solvency problem when largely offset by private assets held abroad. Large volumes of capital flight are taken as evidence for excessive taxation, economic mismanagement and lack of confidence in announced policies, casting doubt over debt relief as an appropriate response to debt service problems. Tackling the underlying problems first should in such circumstances arguably be made a precondition for any debt relief. Tn spite of its importance, measurement of flight capital has remained a matter of dispute. Table 1 below demonstrates the effect of using one method versus another: it takes the case of Mexico and presents the results from applying various approaches reported in the literature to a common data set and time period. There is almost a 1001 difference between the highest and the lowest estimate of capital flight over the period 1970-1985. Some differences can be traced to differences in definition and carn thus only be settled by precise reference to the questions asked. Others reflect simple oversights, such as the often made mistake of ignoring the impact of cross-currency exchange rate changes on the dollar value of debt stocks. Finally, there are unsettled issues, such as how to treat interest income on assets held abroad but not reported in the current account.  3 ALTERNATE CAPITAL FLIGHT ESTIMATES (US$ billions) Original Original Author Time Period Estimate Duplication Difference Estimate 1970-85 ..... ................. ............................................................... Cuddington 1974-82 32.6 29.6 3.0 39.3 WDR85 I 1979-82 26.5 20.6 5.9 48.5 WDR85 II 1/ 1979-82 21.5 n/a 48.6 Zedillo 1970-85 28.6 30.5 1.9 3.0.5 Morgan I 1976-85 53.0 46.4 6.6 45.7 Morgan II 2/ 1976-85 36.2 n/a 37.6 Alvarez/Guzman 1981-87 22.0 24.2 -2.2 30.3 Gurria/Fadl 1970-90 19.7 17.6 2.1 26.5 .. ............................................................................ 1/ The second estimate adjusts for depreciation of dollar denominated debt. 2/ The second estimate uses debt flow data instead of change in stock data to identify how much of capital flight is attributed to debt data problems. In this paper we set ourselves a modest aim, although one that turned out to be remarkably laborious to achieve. We will use the various methods to demonstrate some common pitfalls in quantifying capital flight. How much of the difference in captal flight estimates in one well studied case, Mexico in the 1980s, can be traced to simple errors, be they of data definition or conceptual approach, and how much to differences in economic definition? Which issues are important in the definition of capital flight, and which can actually be settled? Can one rely on mechanical approaches, or is detailed country knowledge required to avoid major mistakes? In the next section we survey the most important data and conceptual disputes underlying the differences and, where possiblo, argue on theoretical ground which way they should be settled. Section 3 then derives a set of estimates based on the preferred approach that comes out of that discussion. Section 4 uses this approach as a standard of comparison for the various alternative approaches to assess the empirical importance of the various  4 disputes; we highlight the four most important sources of discrepancy and demonstrate their quantitative importance. Section 5 concludes. 2. Measuring Capital Flight. Disputes on measurement can be sorted into three groups; the first one concerns the approach to measurement (directly or residual based); the second issue concerns scope of definition; and the third concerns implementation of any particular definition chosen. We discuss each group in turn. 2.1 Measuring Capital Flight: the Direct Approach. When analyzing the reactions of private investors to macroeconomic instabilities or other policy-induced investment risks it seems straightforward to look directly at the data on foreign asset changes of domestic residents as recorded in the balance of payments. Since capital flight tends to be associated with rapid response, a case is often made to exclude long term investments. Adherents of this approach take all changes in short-term foreign assets, often called 'hot money', end interpret them as predominantly 'speculative', indicating capital flight. One serious problem with this approach is that unrecorded outflows are not captured in this way. Another problem with using short-term asset changes as a proxy for capital flight is that long-term investments cannot be clearly distinguished from short term investments. Long-term bonds issued abroad can be close substitutes to short-term investment, because they can be purchased without significant loss of liquidity; there is after all a secondary market  in most long term instruments. It is also not clear why investment in equity and real estate should per definition be kept out of the definition of capital flight. For the reasons mentioned above the raw data on short-term capital movements are not really suitable for an assessment of capital flight. Several authors, however, used th_se numbers anyhow, but made adjustments to correct for the problems touched upon. The most important asjustment they uindertook was to include 'errors & omissions' in an attempt to capture unrecorded capital flows. The errors and omissions item in the balance of payments statistics accounts for the difference between credit and debit entries of current and capital accounts. A large negative balance has been interpreted as unrecorded capital outflows. In fact, when general conditions in many developing countries were likely to trigger capital flight (debt crisis, overvaluation and subsequent massive devalua'tons) the value of the errors and omissions often increased substantially. But the errors and omissions item is not identical to unrecorded capital flows. It includes true measurement and recording errors, unreported imports (smuggling) and lagged registration. These entries are unrelated to capital flight and could well change the sign and magnitude of the errors and omissions. Unrecorded imports are debited to net errors and omissions, undeclared exports credited. Foreign exchange that has been used to import goods undeclared lack a counter-entry in the current account and thus shows up in the errors and omissions. Foreign exchange earned through illegal exports may  6 escape balance of payments accountirg entirely, if the arnings remain undeclared in the black market. If they are recorded, however, they are credited to errors and omissions. These problems of interpreting 'errors & omissions' do not allow an inclusion into a capital flight estimate without adjustment!, unless one decides to accept the resulting distortions (see for example Cuddington 1986). In a recent survey Sinn (199Q) has tried to eliminate distortions by taking the largest positiv stock estimate for accumulated 'errors & omissions' in the period analyzed (corresponds to a cumulative cap_.tal inflow) and adding it to each stock estimate. The idea is to neutralize the effect of reducing capital flight in cases of a positive balance in errors & omissions, since pos4tive balances would be contrary to the expected trend and thus unrelated to capital flight and "random in nature". The problem with Sinn's adjustment is that it is not sufficient in cases where smuggling accounts for most of the movements in the errors & omissions item. Since smuggling may affect the errors and omissions data in both ways, carrying out a 'one sided' adjustment (focusing on the effect of credit entries) introduces a bias. The effects of smuggling itself on capital flight can only be analyzed in the context of the use of trade data. The method that we are proposing in the next section corrects the current account by using trade data of partner countries. This approach allows not only to incorporate net trade misinvoicing into the capital flight measure, but also to eliminate the distorting effects of smuggling, at least to the extent that smuggled goods were recorded as imports/exports in the partner countries. But overall, the conclusion that attempts to split this  7 balance of payments item into subcomponents and make adjustments are likely to fail is hard to escape. 2.2 Measuring Capital Flight: the Indirect Approach. Given the problems with using short-term changes in foreign assets and the errors & omissions item, the only alternative way of quantifying capital flight is to treat it as a residual of four balance of payments components: Change in foreign debt, foreign direct investment, change in foreign reserves and the current account balance. The basic assumption is that capital inflows in the form of increases in indebtedness and foreign direct investment finance either the current account deficit or official reserve accumulation; any shortfall is indicative of private foreign asset accumulation, which in this approach is associated with capital flight. This approach has been used by most authors. Variations of this method are either related to the use of different sources of data for debt and direct investment or to various further adjustments to the 'basic residual'. In practice, matters are even more complicated. The unadjusted residual, when it is based on balance of payment data only, actually measures all reported and non-reported changes in assets held abroad minus changes in official reserves. However, there may be public entities like state-owned enterprises which hold foreign assets. Since these entities are under public control, their net foreign asset accumulation should be subtracted from the residual just like changes in official reserves are. In the case of Mexico, the most important two examples are net foreign assets of PEMEX, the oil  company, and the foreign assets of the commercial banks after their nationalization in 1982. 2.3 Scope of the concopt of capital flight Capital flight as it is belng discussed here does not necessarily involve illegal transactions. It is seen as a result of private portfolio decisions, reacting to actual or anticipated changes in macroeconomic or general business conditions in a particular country. Some have treated capital flight as a phenomenon separate from "normal" capital flows, with "normal" defined in different ways. These authors tried to distinguish capital flight from 'ordinary' portfolio diversification and business activities of domestic residents. There are two alternatives. One way would be to make further adjustments to the residual obtained from the balance of payments in order to allow for "normal" business activities, which would have to be defined (e.g. portfolio investment, working capital of firms held in foreign currency, trade credits). This idea has been mentioned by most of the authors, although only few were able to tackle this issue in their estimations. Or, alternatively, one could apply a more general definition of "normal" flows of capital, e. g. as those foreign assets that correspond to recorded interes- income. Within this concept, assets that do not generate reported income must originate from circumventing controls and are thus to be considered capital flight as opposed to normal flows. The latter approach has been applied by Dooley (1986) and subsequently by Khan and Ul Hague (1987), DeDpl1r and Williamson (1987).  9 Dooley estimates changes of the "normal" stock of external claims deriving it from reported investment income and an average market yield. Th.s approach implies some methodological problems. Firstly, interest receipts from abroad are insufficiently reported' and could be substantially understated. In the case of Mexico there may also be data problems, since Mexico includes an estimate of interest earned but not remitted, which of course invalidates using the Dooley approach with these data. Secondly, it is problematic to determine an average market yield. It would depend on the kind of assets acquired abroad, their maturity and currency composition. These determinants vary from year to year. Furthermore, it appears from this concept that non-interest bearing assets would automatically fall under the category of "abnormal" flows, hence capital flight. Equally disappointing were attempts to isolate capital flight from "normal" flows by making further a%1ustments to the basic residual. Elements of the normal/abnormal distinctie, are included in Morgan (1986) and Gurria/Fadl (1991) (taking out the banks' assets even in the period before / International Monetary Fund, Report on the World Current Account Discrepancy, Washington 1987, pp. 45 ff. The world current account discrepancy can to a considerable extent be explained by international di3crepancies in reported portfolio investment income data. Reported income debits often exceed corresponding credits by large amounts (e.g. $ -32 billion in 1983, total current account discrepancy in that year: $ -75.1 billion). Interest earnings by foreign residents are apparently more accurately reported than foreign interest receipts by residents. The capital account data confirm this observation by showing a cumulative net capital inflow for the world total, indicating that the countries receiving capital were in a better position to measure the flows than the countries where the creditors resided and were thus better able to record the related investment income flows. The Fund study locates the main source of the discrepancy in insufficiently recorded investment income and capital outflows in developing countries. The Fund study also shows that adjustments can be made by using cross-border liabilities and assets of banks and employ the relevant market interest rates. However, Dgolg does not use this information for his calculations, interest receipts and cumulated stocks of external claims are obtained from the balance of payments.  10 their nationalization) as well as in Cuddington (1986) (excluding long-term portfolio investment). Most authors mention that one would have to subtract in some way trade finance, working capital and even assets held for reasons of portfolio diversification. Only Gasser and Remolona (1987) have tried to roughly quantify these components. However, their approach is too arbitrary, ,assigning 50 per cent of value increases of exports to trade finance and 50 per cent of export related wealth gains2 to portfolio diversification. Varman (1989) has attempted to adjust the residual obtained from the balance of payments data by using a model designed to interprete the resulting time series. Behavioral equations are used to estimate 'normal' balances for transaction purposes3 and 'capital flight' as motivated by certain events4. Consequently, capital flight is explained as a residual by a particular event structure that motivated it. The event structure takes the form of a prior setup of dummy variables. This statistical approach, however, is unlikely to produce precise results. Firstly, it is not convincing that transaction 2! This wealth term is computed by multiplying the increase in export prices by last years quantity of exports. These trade financing and portfolio diversifying adjustments are made only if they do not change the sign of capital flows. I/ Varman's study does not contain an applied methodology to estimate 'normal' portfolio investment. The two case studies presented, India and Philippines, did not warrant such a methodology, since capital controls did not permit this type of transaction. The author proposes to use a portfolio adjustment model that defines the optimal allocation of domestic household's wealth among domestic financial assets, domestic inflation hedges such as land, and foreign financial assets. Such a model would have to be country specific.