Treasury Loans to the BNDES: Cornucopia?

If the Treasury’s loans to the Brazilian Development Bank (BNDES) generated net fiscal gains the perpetuum mobile would finally have been discovered.

The Institute for Applied Economic Research (IPEA) recently published discussion paper nº 1665 entitled “Measuring the fiscal result of loans from the Treasury to the BNDES: net cost or gain for the Union?”. The views set out in this instigating study are in sharp contrast with those of various analysts, including those put forward in a column I wrote a little over a year ago, which express a strong opposition to the massive transfers of heavily subsidized funds from the Treasury to the BNDES that have been undertaken since 2009, and which currently total nearly a quarter of a trillion BRL.  The study reaches the surprising conclusion that, when the benefits derived from the short and long-term generation of additional investment and income are computed, the Treasury’s loans to the BNDES in 2009 and 2010, which totaled BRL 180 billion, generated a net fiscal gain of BRL 100 million in present value.

In order to arrive  at a gain of  BRL 100 billion, the authors first of all estimated the loans’ implicit subsidy costs, given that the Treasury charges the TJLP (the official Long-Term Interest Rate, currently at 6%) on its loans to the BNDES but  generally raises funds  at the Selic rate (which currently stands at 11%). They then estimate a short-term fiscal gain resulting from the increase in production and income that supposedly occurred on account of the BNDES loans made possible by the Federal funds.  They add a long-term component to this fiscal gain due to the supposed effects of the investments financed by the BNDES on economic growth. The estimated fiscal cost is BRL 50 billion, while the gains amount to BRL 150 billion, thus resulting in the alleged net gain of BRL 100 billion.

The first part of the study estimates the direct costs of the implicit subsidy in the BRL 180 billion loan. This kind of calculation involves many details and suppositions. Unfortunately the authors have not published appendices containing the figures on which the study is based in order to assess whether the hypotheses adopted are reasonable or not.  The final results of the estimates indicate that the direct fiscal cost of the Union’s loan to the BNDES (the loan’s implicit subsidy) amounts to 28% of the BRL 180 billion, or BRL 50 billion.

In order to assess whether these values are reasonable, without having access to the figures used in the study, one can perform a simple calculation, assuming that that the cost of the public debt remained equal to the 10.25% Selic rate prevailing when the study was performed (7/7/2010) and that the TJLP also remained constant at 6%.  Obviously this calculation is merely a rough approximation but it is an effective way, in principle, of assessing the article’s results. The table shows the subsidy (loss) of the Union’s loan to the BNDES. Given that the loans have maturities of 30 to 40 years, one can only conclude that the authors’ calculations seem to considerably underestimate the loans’ fiscal cost. I repeat that, in order to be able to perform a more thorough analysis, the authors must reveal their figures.



Maturity of the loan (years) Loss (% of the loan) Loss (BRL Billion)
1 4% 7
5 18% 32
10 33% 59
15 45% 80
20 54% 98
30 69% 125
40 79% 143

Source: author’s calculations. The percentage loss is calculated by the following formula [1-(1,06/1,1025)^n], where n is the loan’s maturity. The loss in BRL billion is the percentage loss multiplied by the total value of BRL 180 billion.

But the most problematic part of the article lies in the estimation of the supposed fiscal revenue gains resulting from the effects of BNDES loans on the increase in investment. The authors recognize that part of the investments financed by the BNDES would have been financed by other sources, if the subsidized financing had not existed. Therefore, it is necessary to estimate the volume of investment financed by the BNDES that would not have been undertaken if the subsidized loans had not been made possible by Treasury funds. However, to do this they use inadequate econometric tools. The techniques used to estimate the coefficients used by the authors in the simulations do not deal effectively with the econometric endogeneity problems that permeate regressions in macroeconomics.  Techniques that take these problems into account, developed largely by this year’s Nobel prize winners, Sargent and Sims, have been available or some time and should be used. Microeconomic evidence has shown that “… BNDE´s allocations do not seem to affect firm-level operational performance and investment decisions, although they reduce firm-level cost of capital due to the governmental subsidies accompanying those loans.”[1] After all, how can one explain that, despite the great increase in BNDES disbursements, investment’s share of GDP has not even managed to attain a modest 20% in recent year?

But these are not the study’s only shortcomings. Having estimated (albeit  erroneously) the effects of BNDES loans on the quantum of investment, the authors continue the article with analyses based on antiquated Keynesian models that implicitly assume that the economy is not at full employment  (a good hypothesis in 2009 but not as from 2010). Some of the coefficients are also very optimistic, thus helping to inflate the fiscal gains. As they currently stand, the estimates made are not valid and cannot be used as a basis for economic policy prescriptions.

The study also neglects important themes that no bank should fail to consider, such as the quantification of non-performing loans. Nor does it mention that part of the BNDES loans were used to acquire firms, in Brazil and abroad, without any fiscal gain, even in their flawed methodology. Finally the study considers that the dividends paid by the BNDES to the Treasury constitute a fiscal gain.   This practice constitutes a bogus way of generating fake fiscal surpluses through increases in the public debt, distorting fiscal statistics and compromising the fiscal accounts reliability.

A final thought:  if the BRL 180 billion worth of loans from the Union to the BNDES really generated a fiscal gain of BRL 100 billion, as well as leading to higher investment, employment and income, the obvious economic policy prescription should be to increase loans to the BNDES as much as possible. The fact that the authors do not dare to put such a prescription in writing is probably a strong indication that they are somewhat wary of their results. In this case they are right. They certainly should be.

[1] Lazzarini, Musacchio, Bandeira de Mello and Marcon (2011), “What do development banks do? Evidence from Brazil, 2002-2009”, mimeo.