The Impact of International Capital Inflows on Brazilian Fixed-Rate Public Bonds Márcia S. Leon September 2012

Preliminary Version

1. Introduction In Brazil, before 1990, foreign capital inflows did not reach more than 10 billion dollars annually as a result of tight official controls on financial transactions between residents and non residents. This situation changes significantly in the 1990’s, when initiated a gradual liberalization of the Brazilian economy aimed at basically reducing transaction costs of foreign investment inflows. This process took further steps in terms of deepening the previous changes during the period 2001 to 2005, but at this time the liberalization measures were more oriented to outward transactions of local individuals and firms (Van Der Laan, 2007). In response to the financial opening to foreign capital, Brazil experienced since the second half of the 1990’s a rise in foreign direct investment and significant inflows of portfolio investments from abroad1. These flows declined during the 2001-2002 episode of high investors’ risk aversion in international capital markets and also, as a result of the September 2008 collapse of the Lehman Brothers. Portfolio inflows had a quick recovery in 2009 and 2010, but contracted in the following year because of the local government restrictions in the form of a rise in the tax rate for financial transactions (IOF) charged on fixed-income investments by non-residents. After a fall in 2009, foreign direct investment resumed its upward trend that started after 2006. Throughout the 2000’s, equities have represented a major part of the foreign investors’ portfolio, but the public debt bonds and investment funds have also been significant. However, since 2006, the share of equities and investment funds has decreased and given room to a higher preference towards investment on public debt bonds. Between 2006 and 2011, the share of public debts bonds in foreign investors’ total assets has increased almost three times. Part of the reason for this is the exemption from income tax that this group of investors received on the yields of investment on internal federal public debt bonds. In relation to the groups of institutional investors and other non-financial investors, foreign investors have played a significant part of the total assets of the Brazilian economy during the period between 2000 and 2011. In particular, in the market for domestic federal public debt (DFPD) bonds, its share has risen and has shifted almost completely its position, in 2011, with the position of the group of other non-financial investors had held in 2000. When the banking sector is included as holders of DFPD, the share of non-resident investors is reduced, but still has an impressive upward trend from 2 percent in 2000 to 16 percent in 2011. The public debt bond that has had a major part in their portfolio of Brazilian DFPD are the fixed rate bonds, represented

1

The balance of payments methodology classifies foreign capital inflows into four different broad categories: direct investment, portfolio investment, derivatives and other investments.

1

by the NTN-F (Notas do Tesouro Nacional – série F) and the LTN (Letras do Tesouro Nacional)2. According to the data available from Banco Central do Brasil (BCB), the shares of NTN-F and LTN in the hands of non-resident clients were, respectively, 21 and 14 percent in November 2009, which rose to 44 and 18 percent in July 2012. Therefore, the prices of these securities might be affected at some extent by the presence of foreign investors in their market. Vale (2012) has just concluded a study that shows the influence of foreign participation upon the domestic interest rates of long-term fixed-rate Brazilian bonds (NTN-F bonds) in line with other empirical results, for example, by Peiris (2012). This paper aims at giving special emphasis to the behavior of the prices of assets, which perform as collateral in credit transactions, in response to fluctuations in capital inflows to Brazil. The Brazilian market of DFPD bonds is reasonably developed and these securities have a special roll of being used as collateral in repurchase operations. On the other hand, local credit lines that use households as collateral (known as home equity) have so far limited room in the biggest Brazilian banks. These banks estimate that home equity credit lines represent at most one percent of total credit operations to the housing sector. On its turn, data on the price of land, which is another asset that might be used as collateral in credit transactions, is available only on a six-month basis. Therefore, in view of the reduced significance of credit operations that use households as collateral and the scarce data on the price of land, to analyze the effect of foreign capital inflows on the price of DFPD bonds is more attractive. Some recent papers focus on the effect of foreign capital inflows on the yields of long-term public debt bonds. Warnock and Warnock (2005) investigate the impact of international capital inflows on the yields of ten-year U.S. Treasury bond. In particular, they consider two types of foreign investors (institutional and private ones) and they argue that bond yields are forward-looking prices and consequently their model includes expected inflation and expected growth as explanatory variables, besides the budget deficit, the interest-rate risk premium and variables that capture the effects of monetary and fiscal policies. Pradhan et al. (2011) followed the methodology of Warnock and Warnock (2005) and focused on the impact of nonresident participation on the yields of local-currency bonds, for a group of selected emerging market countries from Asia, as well as non-Asian economies, including Brazil. We aim at estimating a similar model exclusively for Brazil. Our estimation uses net portfolio capital inflows as a substitute for the nonresident holdings of public bonds, which is available only from 2007 on.

2. Recent evolution of foreign participation in assets of the Brazilian economy As shown in Figure 1, in response to the financial opening to capital from abroad, Brazil experienced during the second half of the 1990’s a rise in foreign direct investment and significant inflows of portfolio investments, specifically in 1994, 1996 and 1998. During the next decade, these international flows reduced during two episodes. Portfolio investments fell from 8.7 billion in 2000 to minus 4.8 billion two years later as a consequence from the high risk aversion that hit the international financial markets during this time and, in particular, from the presidential elections of 2002 in Brazil. Foreign direct investment also showed a downturn trend during this period, but reached again the peak of 33 billion dollars of 2000, as of 2007 on. Portfolio investments also portray a sharp rise in its levels after 2006, reaching 68 billion dollars in 2010. However, they experienced a sudden fall in 2008, following the collapse of the Lehman Brothers on September of this year, and also, in 2011, as a result from the

2

The first NTN-F was issued as of December 2003, with maturities in January 2008 - a 4-year period. These bonds have biannual interest coupons yielding 10% per year and allow for the stripping of the coupons. This characteristic allows investors to decompose these bonds into several fixed-rate “zero-coupon” bonds with different maturities, similar to the existing National Treasury Bills (LTN). Copied from Secretaria do Tesouro Nacional (2005).

2

Brazilian government high tax rate on foreign portfolio investment introduced in October of the previous year. Recently, the Brazilian government has taxed foreign investment in fixed-income and in the National Treasury bonds by 1.5% since March 2008 (Decree 6391, from March 13, 2008), which became zero in October of the same year (Decree 6613) as a result from the international financial crisis. In 2009 some emerging market economies have applied measures to curb the high and volatile short-term foreign capital inflows, which directed to their economies as a consequence of the excess liquidity in the international capital markets that followed the monetary easing policy adopted by major developed economies. Brazil, for instance, re-established the IOF tax at the rate of 2 percent, in October 2009, on non-resident portfolio investment. One year later, it was raised to 4 percent, excluding investment on individual stocks, and in a few days later it increased to 6 percent and extended to the margin requirements in the derivatives market. These measures, combined with other macro prudential policies, have contributed to reduce the net inflows of speculative foreign capital. Figure 1: Annual Foreign Capital Inflows in Brazil during 1980-2011 80000

60000

USD millions

40000

20000

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

0

-20000

-40000

Foreign direct investment (net)

Foreign portfolio investment (net)

other foreign investments (net)

According to Table 1, the portfolio composition of foreign investors in Brazil in the month of December from 2000 to 2011 shows that investment in equities represents a major part  more than 50 percent  of their total assets in each year. The share of private debt bonds and of banking deposits is insignificant, but the portions of their total assets allocated in public debt bonds and in investment funds are also meaningful. Between 2006 and 2011, the share of investment in public debt bond has increased almost three times from 11 percent to 28 percent, in part as a result from an exemption from income tax on the yield of foreign investments in domestic federal public debt or in mutual funds that are comprised of 98 percent of federal internal public debt bonds. Medida Provisória 281, from February 16, 2006, reduced the income tax rate from 15 percent to zero non-resident investments3.

3

Investors from countries that do not tax income or do it at a maximum rate of 20 percent, the so-called fiscal paradises, are not included in this measure. MP 281 was converted into Lei 11.312 in June 27, 2006. See Costa (2008, p. 8).

3

Table 1: Composition of foreign investors' portfolio - nominal BRL millions and share (italics) Types of assets Equities Private debt bonds Banking deposits Public debt bonds Investment funds (public bonds) Total of assets

DecDecDecDecDec00 01 02 03 04 Dec-05 Dec-06 Dec-07 Dec-08 Dec-09 Dec-10 Dec-11 33306 31812 27450 50452 69515 113812 177440 293519 166745 357222 423538 411598 79.5% 73.3% 63.5% 67.8% 69.8% 77.1% 73.2% 71.9% 54.0% 65.2% 62.0% 59.4% 360 0.9%

698 1.6%

498 1.2%

564 0.8%

393 0.4%

2201 1.5%

2151 0.9%

10367 2.5%

13155 4.3%

19913 3.6%

5748 0.8%

9271 1.3%

29 0.1%

29 0.1%

29 0.1%

46 0.1%

62 0.1%

100 0.1%

173 0.1%

303 0.1%

230 0.1%

423 0.1%

506 0.1%

732 0.1%

609 1.5%

1772 4.1%

2768 6.4%

4494 6.0%

5235 5.3%

6973 4.7%

26814 11.1%

48987 12.0%

72492 107215 175292 194650 23.5% 19.6% 25.7% 28.1%

7592

9104

12495 18855 24451

24536

35819

55214

56113

63509

77515

76655

18.1% 21.0% 28.9% 25.3% 24.5% 16.6%

14.8%

13.5%

18.2%

11.6%

11.4%

11.1%

41896 43415 43239 74410 99656 147623 242397 408390 308734 548283 682600 692905 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%

Source: CEMEC - Centro de Estudos de Mercado de Capitais, Contas Financeiras CEMEC: Retrospecto 2000 a 2011,Tabelas 8A e 8D, Mar. 2012.

Compared to other types of non-financial investors, foreign investors have increased their share in the total assets of the Brazilian economy between 2000 and 2011. According to Table 2, they have participated significantly in different types of assets in the local capital market. Their share in the equities segment has risen from 8.2 percent in December 2000 to 20.3 percent in December 2011. Moreover, its participation in the total assets represented by public debt bonds has had an impressive change from 0.9 percent in Dec. 2000 to 59.4 percent in Dec. 2011, shifting almost completely the position previously occupied by the group of other non-financial investors. When the banking sector is included, the share of foreign investors in the total stock of domestic federal public debt (DFPD) is not as big. According to CEMEC data, foreign investors’ investment in funds constituted of public debt bonds is added to their investment in public debt bonds and the sum of these two alternative types of investment represented 1.8 percent in December 2000 and increased almost linearly to 15.4 percent in December 2011 (see Table 3).

4

Table 2: Portfolio composition by type of investidor - nominal BRL millions and percent share (italics) Dec-00 Types of assets Equities Private debt bonds Private debt bonds launched by banks Banking deposits Public debt bonds Investment funds (public bonds) Others Total of Assets

Types of assets Equities

Institutional investors (1)

Foreign investors

Other nonfinancial investors

32431 8.0 2679 17.5 4163 1.8 17591 26.9 94078 29.0 10964 100 161905 15.3

33306 8.2 360 2.4 29 0.0 609 0.9 7592 2.3

339184 83.8 12243 80.1 6668 100 228536 98.2 47173 72.2 222714 68.7

0 41896 4.0

0 856518 80.8

Institutional investors (1)

Dec-11 Other nonForeign financial investors investors 408900 1524695 20.3 75.8 9490 118130 6.4 79.4 265902 100 732 1095207 0.1 98.7 194650 15334 59.4 4.7 76655 1195503 4.0 62.6

76922 3.8 Private debt bonds 21201 14.2 Private debt bonds launched by banks Banking deposits 13476 1.2 Public debt bonds 117666 35.9 Investment funds (public bonds) 636553 33.3 Others 32252 100 0 0 Total of Assets 898070 690427 4214771 15.5 11.9 72.6 (1) Closed and open pension funds, insurance and special savings companies. Source: CEMEC - Centro de Estudos de Mercado de Capitais, Contas Financeiras CEMEC: Tabelas Mensais - Tabela 14, fev. 2012.

Total 404921 100 15282 100 6668 100 232727 100 65372 100 324384 100 10964 100 1060319 100

Total 2010516 100 148820 100 265902 100 1109416 100 327650 100 1908712 100.0 32252 100 5803269 100

5

Table 3: Percentage share of foreign investors in total DFPD (including investment funds) Dates Dec 2000 Dec 2001 Dec 2002 Dec 2003 Dec 2004 Dec 2005 Dec 2006 Dec 2007 Dec 2008 Dec 2009 Dec 2010 Dec 2011 4 Source: CEMEC

% 1.8 1.9 2.6 3.4 3.9 3.3 5.9 8.8 10.4 12.5 16.0 15.4

CEMEC data does not match exactly the data from the Brazilian National Treasury, which publishes the monthly share of foreign investment in the stock of DFPD since January 2007. For December 2011, the foreign investors’ participation in the DFPD is 11.3 percent according to the Treasury, while it is 15.4 according to CEMEC data. Table 6 presents the share of foreign investors’ share in the total DFPD, according to the Brazilian NationalTreasury data. Figure 2: Percentage share of foreign investors in total DFPD (including investment funds) 14.0

12.0

percent

10.0

8.0

6.0

4.0

2.0

0.0

Source: STN (Secretaria do Tesouro Nacional), Anexo Relatório Mensal da Dívida Pública Federal, mar. 2012.

5

The public debt bonds that have had a major part in the portfolio of nonresident investors in Brazilian DFPD are the fixed-rate bonds, represented by the NTN-F and the LTN bonds. Figure 3 shows that, according to the data available from Banco Central do Brasil (BCB), the shares of NTN-F and LTN in the 4

Tabela 25D – Títulos Públicos por Detentor (DPMFi): inclusive setor bancário - % do total, CEMEC: Centro de Estudos de Mercado de Capitais: Contas Financeiras CEMEC – Tabelas Mensais, Março 2012. 5 http://www.tesouro.fazenda.gov.br/hp/relatorios_divida_publica.asp

6

hands of non-resident clients were, respectively, 21 and 14 percent in November 2009, which rose to 44 and 18 percent in July 20126. Therefore, the prices of these securities might be affected at some extent by the presence of foreign investors in their market. The NTN-I bonds and the NTN-B are the third and fourth most preferred bonds held by nonresident investors7. As can be seen in Figure 3, their share in these bonds is below 10 percent during the period Nov. 2009 to Jul. 2012. Figure 3: Percentage share of federal securities held by the public in the hands of non-resident clients 50.0 45.0 40.0 35.0

percent

30.0 25.0 20.0 15.0 10.0 5.0

LTN

LFT

NTN-B

NTN-I

Jul-12

Jun-12

Apr-12

May-12

Jan-12

Feb-12

Mar-12

Dec-11

Oct-11

Sep-11

Nov-11

Jul-11

Jun-11

Aug-11

Apr-11

NTN-C

May-11

Jan-11

Feb-11

Mar-11

Oct-10

Dec-10

Nov-10

Jul-10

Sep-10

Aug-10

Jun-10

Apr-10

May-10

Jan-10

Feb-10

Mar-10

Dec-09

Nov-09

0.0

NTN-F

Source: Banco Central do Brasil, Notas Econômico-Financeiras para a Imprensa: Mercado Aberto.

Figure 4 shows the behavior of the yields of local currency domestic bond obtained from Bloomberg database, which is referred to as “generic local currency domestic government bonds” for different maturities, but correspond exactly to NTN-F. This data is available from January 2007 on, but there are some months that no bonds of a certain maturity are negotiated and we plot them as zero yield8. A negative relation between the yields and the participation of foreign investors in the total DFPB is slightly indicated by Figure 4 in the months since the end of 2009. A stronger negative relation is more apparent after Aug. 2009.

6

These shares do not include nonresident participation in investment funds, which can be easily done. The NTN-I are employed for fundraising the payment of interest rate equalization in export financing of domestic goods and services supported by the Programa de Financiamento às Exportações – PROEX. Its nominal value is corrected by the dollar exchange rate. The NTN-B are price-indexed bonds that use the IPCA price index (See DECRETO Nº 3.859, July 4, 2001). 8 The historical data correspond to the “last price” of the last day of the month that the bond was negotiated and the indexes are GEBR03Y, GEBR04Y, GEBR05Y, GEBR07Y and GEBR010Y. 7

7

Figure 4: NTN-F nominal yields and the share of foreign investors in DPFD 20 18 16 14 12 10 8 6 4 2

3 years

5 years

7 years

10 years

7/1/2012

5/1/2012

3/1/2012

1/1/2012

9/1/2011

11/1/2011

7/1/2011

5/1/2011

3/1/2011

1/1/2011

9/1/2010

11/1/2010

7/1/2010

5/1/2010

3/1/2010

1/1/2010

9/1/2009

7/1/2009

11/1/2009

5/1/2009

3/1/2009

1/1/2009

9/1/2008

4 years

11/1/2008

7/1/2008

5/1/2008

3/1/2008

1/1/2008

9/1/2007

7/1/2007

11/1/2007

5/1/2007

3/1/2007

1/1/2007

0

Share in Total DFPD

Source: Bloomberg and Secretaria do Tesouro Nacional, Anexo 2.7 RMD.

3. Estimation Foreign investment in capital markets, as pointed out by Peiris (2010), is a source of demand that can lower yields of domestic bonds and also increase the liquidity of these markets. To test the impact of foreign participation on the long-term local currency government bond yields of a group of emerging market economies, Peiris (2010) estimated a panel data of 10 countries during the period 2000q12009q1. The model specification is based on Baldacci and Kumar (2010) that also apply, as Peiris (2010), panel data approach for a group of 31 emerging and developed countries using annual data for period 1980-2008. Baldacci and Kumar (2010) are also concerned about the determinants of local-currency public bond yields, but they focus on the effects of fiscal deficits and public debt, instead of foreign participation. A recent paper by Pradhan et al (2011) is explicitly interested in the impact of foreign holdings of bonds in the local bond market. Their model follows the one by Warnock and Warnock (2005), which considers bond yields as forward-looking asset prices and, as such, its explanatory variables include forward-looking expectations usually from surveys. Warnock and Warnock (2005) estimate the effect of international inflows on the bond yields of just one country. They concentrate on the 10-year U.S. Treasury bond. Our model specification is based on Warnock and Warnock (2005) and Pradhan et al. (2011). We aim at evaluating the impact of foreign participation on the yields of long-term fixed-rate local-currency public bonds prices (the NTN-F bonds) in Brazil. Vale (2012) has just concluded a study that tested the influence of foreign participation upon the domestic interest rates of NTN-F bonds, as well. His paper develops a specific model for the Brazilian case, based on a simple identity established between the variation of DFPD in two consecutive periods and public sector borrowing requirements and estimated by vector autoregressive estimation. The results indicated that a one-percent increase in foreign participation decreases the monthly average long-term interest rates of NTN-F bonds by 6.58 basis points. His result is in line with one found by Peiris (2010, p. 13): a one percentage point increase in the share of foreign investors decreases bond yields by 6 bps on average. In this paper, the econometric analysis is based on the following reduced-form equation:

rt LT  c  1rt ST  2 te  3 xte  4 t  5 ft  6 yte  7 Bt   t

(1)

8

where, rt LT denotes nominal yields of long-term fixed-rate NTN-F bonds. To capture the effect of local monetary policy on the long-term rates, we follow Warnock and Warnock (2005) and use the target short-term Selic interest rate, rt ST . As the nominal interest rate is a function of expected inflation,  te , the 12-month ahead expected inflation rate is also included in the model specification,  te . The expected exchange rate variation in the following 12 months, xte , is considered a determinant of the long-term bond yields in Pradhan et al. (2011), but not in Warnock and Warnock (2005). Foreign investors expect higher returns on their external investments whenever they expect an appreciation of the currency from abroad, because in this way, their return denominated in their currency also rises. For parsimony, our model does not include this variable, because the local market for future dollars is well developed and since foreign investors represent a significant share of it, probably for hedging. In order to account for the interest-rate risk premium WW includes the variable  t to convey the idea that declining interest rate risk premium contribute to lower long-term rates. Fiscal policy is represented by the fiscal balance in percent of GDP, f t , which is the public sector borrowing requirements (PSBR) as percent of GDP. Baldacci and Kumar (2010, p.9, footnote 17) also use the primary fiscal balance, the change in debt and the level of gross general government debt  all in percent of GDP. The last two variables included in the model are the expected economic growth in the following year, yte , and the nonresidents’ share in DFPD, Bt . The expected signs for each of the variables are shown below: target Selic interest rate, rt ST

positive

12-month ahead expected inflation rate,  te

positive

expected exchange rate variation in the following 12 months, xte

negative

interest-rate risk premium,  t

positive

fiscal balance in percent of GDP, f t

negative

expected economic growth in the following year, yte

positive

foreign participation in percent of DFPD, Bt

negative

3.1 Data For rt LT , Pradhan et al. (2010) use the nominal yield on 10-year sovereign bonds denominated in dollars from Bloomberg, specified as GEBU10Y (see Pradhan et al. (2010, Figure II.1, from Annex II)). Warnock and Warnock (2005) use the nominal ten-year U.S. Treasury yield and Baldacci and Kumar (2010), the nominal yields on 10-year government bonds, in bps. Peiris (2010, p. 12) data for long-term nominal local-currency bond yields for Brazil comes from Bloomberg “generic 5-year local-currency domestic government bond”. We use the same data, which is the GEBR5Y Index, from Bloomberg, which refers to the 5-year NTN-F bonds9.

9

In view of the lack of data for GEBR5Y Index during the months of September 2009 to June 2011, we use data from GEBR04y Index.

9

The short-term interest rate, rt ST , is described by the Brazilian monetary policy target rate (target Selic rate) in annual terms. The 12-month-ahead expectations for inflation and exchange rate are obtained from market expectations time series produced in Banco Central do Brasil. The risk premium series is obtained in a similar way to Warnock and Warnock (2005). It is the volatility of the NTN-F interest rates calculated as the 36-month standard deviation of changes in long rates. Including this variable in the model, calculated in this manner, reduces our sample in half (from 67 observations to 32 observations). Contrary to Pradhan et al. (2010), we do not include the Brazilian Credit Default Swaps (CDS) spreads as proxy for risk premium, because they are generally employed to measure default risk, not interest-rate risk. Alternatively, Pradhan et al. used the VIX (Chicago Board Options Exchange Market Volatility Index), which is a measure of volatility of index prices of options. The fiscal balance is represented by the monthly public sector borrowing requirements (PSBR) as percent of GDP. The expected economic growth in the following year is represented by the expected growth of the industrial production from market expectations compiled and published by the Banco Central do Brasil. The share of non-resident investors as holders of the DFPD is available only since January 2007 from the Tesouro Nacional. As an alternative to this short time series is the net foreign portfolio investment in long-term fixed-income bonds negotiated in the country. Figure 5 shows the net foreign portfolio investment in long-term fixedincome securities in real terms (USD millions of July 2005). Figure 5: Net foreign portfolio investment in long-term fixed-income securities in real terms 4,000

3,000

2,000

1,000

0

-1,000

-2,000 2004 2005

2006

2007

2008

2009

2010

2011

2012

A dummy variable that portrays the IOF tax rate on foreign fixed income investment is included. It is specified as a step function, where each level, in the interval [0,1] is proportional to four different values that the tax rate has taken from 2007 to 2012. Also, to control for a liquidity effect, the effective rate of reserve requirements is used as explanatory variable. Whenever this rate is reduced by the government, the amount of public bonds supplied to the market rises because less public bonds are necessary to fulfill the monetary authority rule and, as such, their yields increase. Alternatively, Peiris (2010) considers the broad money growth for this purpose. Finally, another dummy variable, specifically for October 2010, takes part in the model specification. In this month, the nominal NTN-F yield jumps and its inclusion improved the problem of non-normality and heteroscedasticity of the residuals.

3.2 Nonstationarity and endogeneity Estimations require, at first, only ordinary least squares regression. Two issues are considered in the estimations, as highlighted by Warnock and Warnock (as from here, denoted by WW): nonstationarity and endogeneity. WW (2005, p. 12-13) point out that U.S interest rates and expected inflation appear to be nonstationary. Therefore, they impose an assumption that the long-term Treasury yields are

10

cointegrated with expected inflation and the Federal funds rate and that the sum of the coefficients of these two explanatory variables sum to one. A recent paper by Gomes da Silva and Leme (2011) shows that inflation expectations in Brazil are stationary with mean reversion. They use the 12-month-ahead inflation expectations from July 2001 to December 2010 and achieve this result when the possibility of structural breaks is taken into account (breakpoints in 2003:3 and 2008:1), otherwise conventional unit root tests (ADF, KPSS and PP) showed contradictory results10 11. Figure 6 shows the nominal yield of NTN-F bonds and the market expected inflation rate for IPCA. In our estimations, the nominal Selic rate is also considered stationary, based on the evaluation of Medeiros et al. (2011, p.10-11). They applied conventional unit root tests for the annualized monthly nominal Selic interest rate for the period Dec. 2001 to Dec. 2010, used other unit root tests that consider breakpoint of unknown date and found that the Selic rate is stationary. Figure 6: Nominal yields and expected inflation in Brazil 14 12 10 8 20

6

18

4

16

2

14 12 10 8 01

02

03

04

05

06

07

08

09

10

11

12

5-year nominal yield on NTN-F bonds Expected 12 month month infaltion rate (IPCA)

The Johnasen cointegration test (with no deterministic trend) is applied to the three variables, as done by WW, and indicates that there are no cointegrating relations at the 5% level by the trace test and the maximum eigenvalue test. Even among the two non-stationary variables the test does not indicate any cointegration relation12. The model is estimated with and without this restriction. We also applied three conventional unit root tests and show the results in Table A1 in the Appendix for all the variables used in the estimations. For the period Jan. 2007 to Jul. 2012, the test statistics show that the target Selic rate is stationary according to the ADF and KPSS tests, but not the PP test. The rate of reserve requirements and the participation of foreign investors in the DPFD are both nonstationary according to the unit root tests13. However, by construction, these variables are limited in the interval [0, 100] percent and, therefore, their values cannot sustain an upward or downward trend indefinitely. Besides, the rate of reserve requirements seems to have a structural break during the period at two times after 2008, which could help accept a false unit root. The expected industrial production growth over the next year and the fiscal balance (including the primary fiscal balance) in percent of GDP appear 10

I am grateful to Osmani Guillen, from the Research Department of Banco Central do Brasil for pointing out this issue. ADF, PP and KPSS stands for Augmented Dickey-Fuller, Phillips-Perron and Kwiatkowski-Phillips-Schmidt-Shin univariate unit root tests. They are applied with EViews software. 12 This result probably stems from the small sample size. 13 The KPSS does not reject the null hypothesis of stationarity of the foreign participation in the DFPD. 11

11

to be stationary according to the three unit root tests. But, the level of gross general government debt in percent of GDP is considered nonstationary by the ADF and the PP tests, but not the KPSS. However, this result might be influenced by structural breaks in the series as shown in Figure 7. The idea of including the stock of public debt in percent of GDP in the model comes from the work of Baldacci and Kumar (2010). They estimate the impact of fiscal deficits and public debt on the sovereign bond yields for a panel of 31 advanced and emerging market economies. Figure 7: Some of the explanatory variables TARGET INTEREST RATE

EXPECTED IND. PROD. GROWTH OVER NEXT YEAR

14

6

13 5

12 11

4 10 9

3

8 7

2 01

02

03

04

05

06

07

08

09

10

11

12

01

FISCAL BALANCE IN % GDP SEAS. ADJ.

02

03

04

05

06

07

08

09

10

11

12

PRIMARY FISCAL BALANCE IN % GDP SEAS. ADJ.

12

4

8

0

4

-4

0

-8

-4

-12 01

02

03

04

05

06

07

08

09

10

11

12

LEVEL OF GROSS GENERAL GOVT. DEBT IN % GDP 64

01

02

03

04

05

06

07

08

09

10

11

12

EFFECTIVE RATE OF RESERVE REQUIREMENTS (%) 40

62

35

60 30 58 25 56 20

54 52

15 01

02

03

04

05

06

07

08

09

10

11

12

01

02

03

04

05

06

07

08

09

10

11

12

In our estimations endogeneity may arise because one of the explanatory variables is determined simultaneously with the dependent variable. The right-hand-side variables of equation (1) that could answer contemporaneously to a shock in the long-term nominal yields of NTN-F bonds are the interest rate risk premium and the foreign capital inflows. The interest-rate risk premium could be simultaneously determined by construction. To avoid endogeneity probably Pradhan et al. (2011) chose

12

the VIX index, which is highly exogenous to all nominal yields in emerging markets14. To estimate equation (1) including the measure of interest-rate risk premium, as defined by WW, diminishes significantly the sample and, consequently, the inference we would like to do with the results. Therefore, we should try to use the VIX, as well. As a preliminary result, one of our equations is estimated by OLS, not with 2SLS even though the endogenous risk premium is included. Furthermore, the initial estimations use the share of nonresident investors on the total DFPD, not the inflows of foreign investments. A shock to the nominal yields of long-term domestic bonds might produce a contemporaneously rise in fixed-income investments from abroad, but it does not mean that the share of foreign investors on DFPD would rise as well. Therefore, Bt is exogenous, while foreign capital inflows are not. In a second round of estimations, we estimate equation (1) with foreign capital inflows as one of the explanatory variables. In this case, 2SLS has to be applied. 3.3 Results Table 4 presents the results of five preliminary specifications of equation (1). Specification [1] shows that a one percentage point increase of foreign participation in DFPD reduces 5-year nominal yields of NTN-F bonds by 33 basis points with a one-percent significance level. Pradhan et al. (2011) also found that this coefficient is significant and with a negative effect of 5 bps, while Peiris (2010), 6 bps15. However, this specification has a very low adjusted-R2, meanwhile the target interest rate and the fiscal deficit in percent of GDP are not significant. In fact, for the specifications presented in Table 4, the explanatory variable that is highly significant in all the five cases is the one-year ahead expected inflation. For specifications [1], [2] and [3] that do not impose a cointegration restriction on the Treasury yields and the inflation expectation, a one-percentage point rise in inflation expectation increases the 5year nominal yields of NTN-F bonds by about 150 basis points. The sign of this coefficient is the same as the one suggested by previous works (see Table A2, in the Appendix). However, the 150 bps effect is almost two times the one found by WW, BK, Pradhan et al. and Peiris. Looking at Table 4, the adjusted R2 increased when the effective rate of reserve requirements is included as an additional explanatory variable16. Also, it improved the significance of the target interest rate, the expected inflation rate, the fiscal balance as percent of GDP, which all became significant at the one-percent level. Besides, for a one-percentage point increase in the target inflation rate, the nominal NTN-F yields would rise by 42 bps, which is in line with the result of WW (30 bps)17. The coefficient of fiscal balance as percent of GDP (PSBN relative to GDP) is negative. As the public sector deficit (-PSBN) relative to GDP increases, then the federal government should issue new debt, reducing the price of new bonds and, consequently, the yields rise. Baldacci and Kumar (2010, p.13) also find a negative sign for the effect of fiscal balance on the long-term bond yields, while and WW (2005) and Peiris (2011) find a positive sign for the size of the fiscal deficit-to-GDP ratio on long-term yields. For the case of an increase of a one-percentage point in the fiscal-balance-to-GDP ratio pulls down bond yields by 19 bps, which is in accordance with the 13 bps rise of WW estimations that occur after a onepercentage point increase in the deficit-to-GDP-ratio.

14

I am grateful to an anonymous referee for this comment. WW does not include foreign participation as explanatory variable, but instead foreign inflows under three alternative definitions. According to their estimations all three foreign flow variables exhibit a significant negative impact on long rates. 16 I am grateful to Waldyr Areosa, from the Research Department, of Banco Central do Brasil, for this suggestion. 17 BK (70 bps), Pradhan et al. (65 bps) and Peiris (12 bps). 15

13

Table 4: Impact of the Foreign Participation in DFPD on NTN-F Interest Rates Dependent Variable: 5-year nominal yields measured in percentage points Sample Jan. 2007 to Jul. 2012 p-values in italics [1] [2] [3]

[4]*

[5]*

Target interest rate

0.12673 0.41788 0.47049 0.57435 0.39913

One-year ahead inflation expectation

1.53215 1.49929 1.42311 0.42566 0.60088

0.4212

0.0019

0.01114 0.00154

0.0261 0.0093

0.0188 0.00000

0.6458 0.57597

Expected industrial production growth over the next year

0.0023 Fiscal balance in percent of GDP seasonally adjusted

0.0188 0.00000

0.003

-0.085 -0.1905 -0.1106 -0.1077 0.28027

0.0033

0.0723

0.0683

Primary fiscal balance in percent of GDP seasonally adjusted (%) Level of gross general government debt in percent of GDP Level of gross general government debt in percent of GDP squared Foreign Participation in DFPD (%)

-0.332 -0.1593 -0.1488 0.04101 -0.7791 0.01651

0.1427

0.2668

0.67 0.00000

-0.1762 -0.1698 -0.1575

Effective rate of reserve requirements (%)

0.000

0.0000

0.0001

1.26611

Interest rate risk premium

0.0009

3.10104 3.67444

Dummy for 2008.10

0.0000 Constant Number of observations Adjusted R2 F-statistic

0.0000

6.5814 7.16564 3.70782 5.67241 11.6396 0.00014

0.00

0.0533

0.0006

0.0000

67

67

67

67

32

0.250

0.553

0.679

0.647

0.820

6.5

17.3

21.0

21.1

48.1

*Imposition of cointegration relation among the 5-year nominal yields on NTN-F bonds, the target Selic rate and the one-year ahead expected inflation rate. The dummy variable tor October 2008 improves the results of the residuals tests.

14

WW (2005) estimate that for a one-percentage point increase in expected GDP growth tends to increase nominal long-term yields by 35 bps. Pradhan et al. (2011) also find a positive effect of economic growth on long-term yield of public bonds: for each percent increase in quarterly growth, long-term yields rise by 5 bps. Baldacci and Kumar (2010, p 12) show that economic growth (not expected growth) affects the nominal yields on 10-year government bonds unless primary balance is used as fiscal indicator: a onepercentage point increase in growth rates diminishes long-term yields by 70 basis points. They argue that higher growth leads to higher tax revenues that reduce fiscal vulnerability, decreasing risk premia. Further estimations should include net foreign portfolio investment in long-term fixed-income as explanatory variable, instead of the foreign participation in PFPD. Furthermore, the expected depreciation of the local currency is another explanatory variable to be tested.

5. Conclusion To evaluate the effect of foreign participation on the Brazilian domestic public federal debt has been done in this paper for the yields of the long-term fixed-rate bonds (NTN-F bonds) for the period January 2007 to July 2012). Our estimations are based on a linear regression analysis, following the methodoly developed in the papers of Warnock and Warnock, Baldacci and Kumar, Pradhan et al. and Peiris. As the preliminary results show, the effect of foreign participation is in general not significant. One particular specification, however, shows a reasonable fit of the model and a highly significant coefficient for the foreign participation. This is one line of research to be investigated further  the inclusion of the interest-rate risk premium, measured as the volatility of the nominal bond yields. Unfortunately, the volatility as calculated for this specification reduces sharply the sample and also brings the problem of endogeneity, which has not been taken care of so far.

6. References ANBIMA (2011), IMA – Indice de Mercado (http://www.andima.com.br/ima/arqs/ima_cartilha.pdf).

ANBIMA

–

Metodologia

Baldacci, Emanuele; Kumar, Manmohan (2010). Fiscal Deficits, Public Debt, and Sovereign Bond Yields. International Monetary Fund IMF Working Paper, WP/10/184. Costa, Daniel B. R. Estudo da Diferença entre as Curvas de Título Público Prefixadas em Reais Interna e Externa Brasileiras, Dissertação (Mestrado profissional) – Escola de Economia de São Paulo, 2008. Gomes da Silva, Cleomar; Leme, Maria C. S. (2011). An Analysis of the Degrees of Persistence of Inflation, Inflation Expectations and Real Interest Rate in Brazil, Revista Brasileira de Economia (RBE), v.65, n.3, p. 289-302, Jul-Set. 2011. Medeiros, Gabriela B.; Freitas Filho, Paulo R. S.; Aragón, Edilean K. S. B. Uma Avaliação Empírica dos Efeitos da Taxa de Juros sobre a Taxa de Câmbio em um Ambiente de Déficit Nominal Zero, http://www.ufrgs.br/PPGE/pcientifica/2011_12.pdf. Acessado em 27/09/2012. Peiris, Shanaka (2010). Foreign Participation in Emerging Markets’ Local Currency Bond Markets, International Monetary Fund: IMF Working Paper, WP/10/88.

15

Pradhan, Mahmood; Balakrishnan, Ravi; Baqir, Reza; Heenan, Geoffrey; Vowak, Sylvia; Oner, Ceyda; Panth, Sanjaya. Policy Responses to Capital Flows in Emerging Markets, IMF Staff Discussion Note, April 21, 2011. Secretaria do Tesouro Nacional (STN), Debt Report: The Treasury increases maturity and reduces cost with the issuance of fixed-rate five-year bonds, February, 2005. Vale, Elton M. do (2012). Melhorias para a Dívida Pública e para a Sociedade Decorrentes da Participação Estrangeira na Dívida Pública Mobiliária Federal Interna, Tesouro Nacional: textos para Discussão, n.6, 2012. Van Der Laan, Cesar R. A Liberalização da Conta de Capitais no Brasil Recente (1990-2005), Rio de Janeiro: Revista do BNDES, v.14, n. 28, p. 425-458, dez. 2007. http://www.bndes.gov.br/SiteBNDES/export/sites/default/bndes_pt/Galerias/Arquivos/conhecimento/r evista/rev2814.pdf. Acessado em 10/09/2012. Warnock, Francis; Warnock, Veronica (2005). International Capital Flows and U.S. Interest Rates, Board of Governors of the Federal Reserve System: International Finance Discussion Paper 840.

16

7. Appendix Table A1: Conventional Unit Root Tests Series

ADF H0: has unit root PP H0: has unit root KPSS H0: is stationary Test Critical values Test Band Critical values Test Band Critical values Lag statistics 5% 10% statistics with 5% 10% statistics with 5% 10%

Target interest rate -4.057985* One-year ahead inflation expectation -0.818566 Expected industrial production growth over the next year -2.580523 Fiscal balance

3 -3.48

-3.17

-1.096503

6

-1.95

-1.61

0.082346

6

0.146

0.119

1 -1.94

-1.62

-0.600312

3

-1.94

-1.62 0.198357*

8

0.146

0.119

4 -2.91

-2.59

-4.146398*

0

-2.91

-2.59

0.091368

5

0.463

0.347

-9.201013*

0 -3.45

-3.15

-9.160909*

4

-3.45

-3.15

0.098739

0

0.146

0.119

Primary fiscal balance in percent of GDP seas. adjusted -9.393828*

0 -3.45

-3.15

-9.484019*

4

-3.45

-3.15

0.081747

5

0.146

0.119

4

-1.95

-1.61

0.117323

6

0.146

0.119

0

-3.48

-3.17

0.115675

6

0.146

0.119

3

-1.94

-1.61

0.185213

9

0.146

0.119

2

-1.95

-1.61

0.085122

4

0.146

0.119

in percent of GDP seasonally adjusted

Level of gross general government debt in percent of GDP 0.131695 0 -1.95 -1.61 -1.613633 Foreign Participation in DFPD (%) -2.270202 0 -3.48 -3.17 -2.270202 Effective rate of reserve requirements (%) -0.681165 0 -1.94 -1.61 -0.660599 Interest rate risk premium -0.814183 1 -1.95 -1.61 -1.033322 * and ** Rejection of the null at 5% and 10% level respectively. The sample sizes are not the same for all series.

17

Table A2: Results from other estimations Authors

Warnock and Warnock (2005)

Country study Dependent Variable

EUA

Sample

Nominal ten-year Jan 1984 U.S. Treasury yield May 2005

Explanatory Variables

positive

A one percentage-point increase in long-term inflation expectations tends to increase nominal yields by 70 basis point.

expected GDP growth in the subsequent year

positive

A one percentage-point increase in expected GDP growth tends to increase nominal by 35 basis point.

risk premium

positive

target federal funds rate

positive

A one percentage point of Fed tightening results in a 30 basis point increase.

positive

A one percentage point increase in the deficit-to-GDP ratio increases long rates by 13 basis points.

negative

For 2 percent of GDP overall bond inflows, long rates would be 105 basis point higher (p.16)

positive

A one percentage point increase in inflation expectationd increases long-term bond yields by 70 basis points (p. 14).

negative

A one percent increase of the overall fiscal deficit relative to GDP pushes bond yields by 17 basis points.

negative

Higher growth leads to lower yields. A one-percentage point increase in growth rates diminshes long-term yields by 70 basis points.

nonresident participation on local bond matket

negative

Each percentage point increase in nonresident participation reduces long-term bond yields by about 5 bps on average.

policy rate

positive

economic growth

positive

fiscal balance

positive

A one percent increase of the overall fiscal balance relative to GDP pushes bond yields by 6 basis points.

positive

For the monthly model, a one percentage point increase in expected inflation increases bond yields by 50 bps.

31 advanced Annual Nominal yields on 10Baldacci and and emerging frequency inflation year government Kumar (2010) market from 1980- expectations bonds economies 2008 overall fiscal balance

economic growth

Pradhan et al. (2011)

Eight emerging markets

Empirical results

long-term inflation expectations

budget deficit as percent of lagged GDP 12-month flows in U.S bonds scaled by GDP

Two samples: quarterly 2000 q1 to Ten-year bond yields 2010 q4; and monthly 2006 m1 to 2010 m12

Suggested Signs

expected inflation

A one percentage point increase in the policy rate increases long-term bond yield by 65 bps. For each percent increase in quarterly growth increase long term yields by 5 bps.

18

Explanatory Variables

Authors (Continued) Pradhan et al. (2011)

Peiris (2010)

10 emerging market economies

Long-term local Quarterly currency government data from bond yield 2000-2009

Suggested Signs

Empirical results

expected depreciation

negative

A one percentage point depreciation lower long term rates by 5 to 23 bps.

risk premium

positive

In the monthly model, each percentage point increase in sovereign CDS spreads add 4 bps to bond yields.

negative

A one-percentage point increase in the share of foreign investors in the government bond market tends to lower yields by 6 bps on average.

positive

The annualized impact of a one percent rise in the fiscal deficit to GDP on EMs’ long-term yields is about 20bps (see Table 3, p. 14).

foreign participation

fiscal deficit

19