Since its peak in September, 2005, the Selic (basic) interest rate has been falling in every COPOM (Brazilian Central Bank Open Market Committee) meeting until its current 12.5% level. This is the lowest in many years, albeit very high for an inflation forecast below 4%. In the same period, the BRL appreciated from 2.30 to 1.95 BRL/USD. The fact that the currency appreciation occurred simultaneously to the fall in the Selic rate has been used by many pundits as evidence that the interest rate differential plays no significant role in the appreciation of the BRL. According to this line of reasoning, the great improvement in trade and current accounts balances explain entirely the appreciation, and further cuts in the Selic should not help to stop or reverse the appreciation trend.

Data of exchange rate transactions, compiled by the Brazilian Central Bank (BCB), are shown as additional evidence that the interest rate differential that attracts carry trade does not affect the exchange rate. According to those statistitics, exchange rate transactions originated by trade are in surplus (i.e., bring USD on net); while those originated by finance, in deficit. In the first quarter of 2007, trade related exchange rate transactions brought 20.6 USD billion (net flow), while finance related exchange rate transactions were responsible for a net outflow of 3.2 USD billion.

These statistics, however, cannot be used as proof that the interest rate differential does not matter for the BRL exchange rate. First, because the way the CB compiles the data does not allow one to appropriately separate the trade related from the finance related exchange rate transactions. For example, an import transaction that is financed through a loan with maturity longer than one year, when settled, will enter the BCB statistics as a financial outflow.

Moreover, even if the BCB decides to compile the statistics accounting for cases as the aforementioned one, no statistics of exchange rate quantities may shed light on the question of whether or not the interest rate differential affects the exchange rate. Suppose an exporter that receives USD 10 million for exports and converts those in BRL (as required by Brazilian export surrender requirements). Furthermore, after paying the costs, he/she is left with the equivalent in BRL of USD 1 million. The decision to keep the profit in BRL or USD will depend, in great measure, on the interest rate differential; the higher the differential is, the higher the share of profits kept in BRL. There is no easy way that statistics of exchange rate quantities can capture this. They will always show that the inflow was trade related, although it stayed in the country due to the interest rate differential.

One additional complication has to do with derivatives trading. The traditional way to profit from the interest rate differential is to borrow in USD, convert the USD into BRL, and invest those BRL in fixed income. This is the plain vanilla carry trade. Foreign investors may replicate the aforesaid financial strategy by purchasing NDFs (non-deliverable forward contracts) of BRL. Those are sold by banks outside of Brazil. An equivalent way is to sell USD futures at BM&F (Brazilian Mercantile and Futures Exchange). Any of these financial strategies will provide the investor with a return equal to the interest rate differential plus the BRL appreciation. The risk is a BRL depreciation greater than the interest rate differential. In such case, investors will lose money with the carry trade. So far, I have only been able to gather data on quantities traded at BM&F. Therefore, the empirical evidence I henceforth analyze has to do with strategies involving derivatives trading at BM&F.

Chart 1 shows two series. The first are (gray area) is the foreign investorsâopen interest positions in USD futures at BM&F, in USD billion. (Keep in mind that to sell USD futures is akin to borrow in USD and invest in BRL, as previously mentioned.) The second series (blue line) is the exchange rate, quoted in BRL/USD. Chart 1 shows that both series are positively correlated (simple correlation coefficient of 0.60). When the open interest increases, the BRL appreciates; and vice-versa.

In spite of the fact that the foreign investorsâ open position in USD futures at BM&F is only part of the foreign investorsâ investments in BRL fixed income, it is reasonable to assume that the open interest is a good proxy of the total investment. If this is the case, one possible interpretation of the abovementioned positive correlation is that it represents movements of a âsupply of foreign fundsâ

Chart 2 exemplifies what I have in mind. Suppose we start in point A, when a negative shock hits, increasing the uncertainty over the US economy future path, therefore increasing foreign investors` risk aversion. This happened in May, 2006. Chart 1 shows that the exchange rate depreciated (increased) abruptly as the open interest decreased (in absolute terms), even turning positive. Such movement corresponds to the movement of the supply curve from A to B (Chart 2). At B, the exchange rate is more depreciated, and the open interest smaller. The reverse situation occurs when good news make the Brazilian country risk fall, as it has been happening since March, 2007. In those instances, the supply of funds increases, and, assuming stable demand, the exchange rate falls (appreciates), as in the movement from A to C (Chart 2).

Therefore, the interest rate differential, albeit decreasing, seems to impact the exchange rate. Chart 1 shows that there are three periods when the open interest are strongly reduced: May-June, 2006, October-November, 2006, and February-March, 2007. In the first two periods, foreign investors completely offset their (short) open interest in USD futures. If we are willing to adopt the zero open interest as a âneutralâ fall in Brazilian interest rates.

In May, 2006, the exchange rate that corresponds to point D in Chart 2 is 2.30 BRL/USD. Incidentally, 2.30 is also the linear coefficient of an OLS regression of the exchange rate on the foreign investorsâ open interest. In November, 2006, the exchange rate climbed only up to 2.20 BRL/USD. In March, 2007, it stopped short of 2.15 BRL/USD.

Of course, these figures are preliminary estimates of how much the exchange rate could depreciate as the interest rate reduction process proceeds. A full-blown analysis must consider explicitly other factors, as the effect of sterilized interventions, foreign investorsâ risk aversion, current account performance, etc. Nevertheless, I think these are clear indications that: 1) the interest rate differential contributes to the current appreciation of the BRL; 2) if and when the BCB completes the current monetary policy easing, the BRL will not climb anywhere near its levels in 2002/2003.

**CHART 1**

**CHART 2**

Dr. Garcia – Very interesting analysis. The open interest position graph really tells us everything: Interest differentials are indeed an important determinant of BRL appreciation. I believe the only risk to a depreciation of the BRL is in case the US economy decelerates more than expected creating a double interruption of USD flows into the Brazilian economy. In one side, market expectations deterioration would reduce open interest positions and on the other side (more medium-term) world demand for EM exports would fall, reducing also the USD flows coming from the current account side of the balance of payments. But this would be a deeper crash than the one seen in May-2006 – and is quite related to the issue of the ‘decoupling of US growth’ vis-à-vis the rest of the world. I tend to believe the chances of such a scenario materializing are quite reduced.

Very interesting analysis. A few questions: how fast should the central bank reduce rates? how much of the Real appreciation is a fundamental phenomenon and how much are we getting into overvaluation territory? What if more aggressive monetary policy easing does not stop the capital inflow and leads to further appreciation? Should other policy options – such as controls on hot money capital inflows – be considered?

Marcio, You provide very clear insights. We, ourselves, follow the BMF open USD interest very closely on daily basis. I completely agree with your theoretical conclusions. However there are some caveats in your approach. First of all, I agree that the data of exchange rate transactions is very poor to track the BRL and interest rate dynamics, for the reasons you have mentioned. Secondly, I would add that simple analysis between interest rate differential (DI – Libor) produce, at first sight, counter intuitive results, i.e., there is a positive relationship between BRL depreciation and rate differential. In fact, this happens if one fails to control for risk premiums and external shocks. Over the past years, the same external shocks that induced BRL depreciation also induced local rates hikes. Hence, one might “find” that interest rate differential do not matter. Now turning back to the BMF data. There is, indeed, a positive relationship between BRL and BMF open interest, however this relationship is non stationary. Nor the correlation coefficients neither the OLS regression are robust. In fact, both of them are spurious. Hence, one has to be careful with the 2.30 “neutral FX rate”. I have run a similar regression. The linear coefficient is very close to the one you have found. But, unfortunately the regression is spurious, i.e. the residuals are non stationary (unit-root). Variable Coefficient Std. Error t-Statistic Prob. BMF 1.91E-06 1.54E-07 12.39253 0.0000 C 2.35 0.009981 236.2537 0.0000 R-squared 0.198270 Mean dependent var 2.266567 Durbin-Watson stat 0.017543 Prob(F-statistic) 0.000000 Null Hypothesis: RESID has a unit root t-Statistic Prob. Augmented Dickey-Fuller test statistic -2.241330 0.1920 Moreover, the causality goes from BRL to Open Interest, not the other way around as suggested. On top of that, there is a day lag in the statistic published by BMF. Null Hypothesis: Obs F-Statistic Probability DBRL does not Granger Cause DBMF 619 2.43316 0.06399 DBMF does not Granger Cause DBRL 1.39417 0.24351 We need to take this into account in the model specification. Moreover, is unlikely that the BMF data would be any good to FX forecast. At the end of the day, FX modeling remains the via-crucis of economists. In fact, Messe-rogoff result remains robust: random walk models over perform fundamentals in FX forecasting. The hope comes from microstructure modeling that has shown promising results. The BMF open interest plays a very important role in the BRL dynamics and has its roots on market microstructure, but unfortunately it does not tell us everything. But as you have put a “full-blown analysis must consider explicitly other factors, as the effect of sterilized interventions, foreign investors’ risk aversion, current account performance, etc.”. I would add necessity of a dynamic model specification taking in account non stationary variables. Paulo F. Hermanny Vice President Chief Fixed Income Strategist Itaú Corretora de Valores S.A.

Brazil Survey: Analysts Cut Interest Rate Forecast Again Mon, May 21 2007, 14:54 GMT http://www.djnewswires.com/eu Brazil Survey: Analysts Cut Interest Rate Forecast Again SAO PAULO (Dow Jones)–Brazilian financial market analysts and economists have once again reduced their estimates for the 2007 year-end Selic base interest rate, according to the central bank’s weekly market survey published Monday. Analysts cut their 2007 year-end Selic rate forecast to 10.75% from 11% in last week’s survey. It was the second consecutive reduction in Selic forecasts. Currently, the base rate stands at 12.50%. The weekly central bank survey tracks the opinions of 100 analysts and economists from banks and brokerages, reporting the average of their expectations. Earlier Monday, the central bank said this week’s survey would be canceled because of a strike by central bank workers. Later Monday, however, the bank said the survey was completed by non-striking supervisory personnel. The central bank next meets on June 6 to review monetary policy. At its last meeting in April, the bank reduced the rate for the 15th consecutive time. In the meantime, survey respondents maintained their forecast for the 2008 year-end Selic rate at 10%. In Monday’s survey, analysts again reduced their estimates for 2007 year-end inflation as measured by the official consumer price index, or IPCA, to 3.6% from 3.62% in last week’s survey. The inflation projection is below the central bank’s inflation target of 4.5% for the year. Inflation has trended lower in 2007, with the IPCA 12-month rate through April at 3.0%. In 2006, IPCA inflation was 3.14%, down from 5.69% in 2005. Experts kept their 2008 year-end IPCA outlook at 4.0%. In addition, analysts kept their forecasts unchanged for Brazilian gross domestic product growth in 2007. Experts maintained their 2007 GDP estimate at 4.10% and their projection for 2008 GDP growth at 4.0%. In 2006, Brazil’s GDP grew 3.7%. The estimate for the 2007 year-end debt-to-GDP ratio was increased to 43.9% from 43.76% in the previous survey. Experts also boosted their estimate for the 2007 trade surplus to $41.1 billion from $40.55 billion in the previous survey. In 2006, Brazil posted a record trade surplus of $46.07 billion. The Brazilian real is expected to end 2007 at BRL2.00 to the dollar, analysts forecast. On Monday, the real was trading at BRL1.95 per dollar.