Expectations and Long-Term Interest Rates: Comment on Bovenberg

This paper examines the macroeconomic effects of IMF-supported adjustment programs. The evidence is reviewed for such effects, and new estimates are provided of these effects for 69 developing countries with programs during 1973–88. The empirical analysis indicates that in the short term, programs have led to an improvement in the current account and the balance of payments, a lowering of inflation, and a decline in growth. In the longer term, the positive effects of programs on the external balance and inflation are strengthened, and the adverse growth effects reduced.

Abstract

This paper examines the macroeconomic effects of IMF-supported adjustment programs. The evidence is reviewed for such effects, and new estimates are provided of these effects for 69 developing countries with programs during 1973–88. The empirical analysis indicates that in the short term, programs have led to an improvement in the current account and the balance of payments, a lowering of inflation, and a decline in growth. In the longer term, the positive effects of programs on the external balance and inflation are strengthened, and the adverse growth effects reduced.

It is widely believed that increases in budget deficits cause higher interest rates. With the exception of the now-familiar Ricardian equivalence view, standard macroeconomic models suggest that interest rates (nominal or real) are an increasing function of budget deficits. Despite the predictions of these models, the empirical results have not provided consistent support for this hypothesis. For example, Hoelscher (1983), Evans (1985, 1987a, 1987b), and Kolluri and Giannaros (1987) do not find a positive and significant effect of deficits on interest rates.

A number of recent studies have focused on the effects of expected deficits on interest rates. This approach is motivated by the belief, as Kim and Lombra (1989, p. 242) argue, “that current changes in interest rates are essentially a function of changes in future expected deficits rather than of changes in current and past deficits.” This point is also clearly made by Bovenberg (1988, pp. 383–84). In each of these studies, the results indicate a significant and positive effect of expected deficits on long-term interest rates. In particular, Bovenberg shows that the interest rate on ten-year and three-year U.S. Treasury bonds is an increasing function of the expected deficit variables.

The purpose of this comment is to present and to test a number of extensions of Bovenberg (1988). Two alternative specifications of the model are estimated in order to determine the robustness of his results. First, as suggested by Barth, Iden, and Russek (1985), I include government expenditures in the interest rate equation. Second, a money supply variable is introduced. It is assumed that individuals use the same process to form expectations of government expenditures and money growth as they do to form expectations of the deficit variables. The objective of this note is not to argue that deficits (expected or actual) have no effect on interest rates, but rather to illustrate that the positive effect of expected deficits on long-term interest rates appears to be sensitive to the specification of the model.

Barth, Iden, and Russek (1985) emphasize the importance of the specification of the interest rate equation. In particular, they note that “… the estimated coefficients on the tax and purchase variables in this [interest rate] equation cannot be equal in magnitude. … If one includes the deficit variable rather than Tx [tax revenues] and G [government expenditures] separately, then one should also include G (or Tx) to capture differences in coefficients” (pp. 557–58). This suggests that it is important to include government expenditures as a right-hand side variable and to determine whether the effects of expected deficits on interest rates are sensitive to government expenditures. I have also introduced the ratio of Ml to the natural level of gross national product (GNP) to capture the effects of monetary policy on long-term interest rates.

The interest rate equations are estimated using quarterly data for the period 1960:1 to 1983:4.1 Because of the presence of autocorrelation in several equations, I have estimated the equations using the maximum-likelihood procedure. The dependent variables are the nominal rates on the ten-year and three-year U.S. Treasury bonds. The expected inflation series is taken from the Carlson-Livingston data on 12-month inflation forecasts; a quarterly series is obtained by interpolation. For a more detailed discussion of the Carlson-Livingston data, see Carlson (1977).

The deficit variable is defined as the ratio of the cyclically adjusted deficit to the natural level of output, CADEFY. The data on the cyclically adjusted deficit and natural level of GNP were obtained from the St. Louis Federal Reserve Bank and Gordon (1987), respectively. Three approaches were used to generate a series on expected deficits. First, the expected deficit, DEF 21, is assumed to be a simple average of CADEFY for five years ahead (including the current quarter). A second measure, DEF 13, is given by a three-year-ahead average. Finally, I have also used a four-year centered moving average, DEF 17, as a measure of expected deficits.2 While Bovenberg uses the recent U.S. Congressional Budget Office (CBO) forecasts as a measure of expected deficits, I use the actual cyclically adjusted deficit for the entire period to construct DEF 21, DEF 13, and DEF 17. I do this for two reasons. First, the CBO forecasts are not cyclically adjusted. And second, the use of the CBO forecasts with the cyclically adjusted deficit data introduces a “break” in the series.3

The results for equations (1.1), (2.1), and (2.5) in Tables 1 and 2 represent replications of Bovenberg’s results. The results provide additional support for his findings; increases in expected deficits have significant and positive effects on the dependent variable.4 Increases in expected inflation, P, are also shown to have positive effects on nominal interest rates. Bovenberg’s results, therefore, are unaffected by the use of: (1) quarterly data; (2) the above measures of expected deficits; and (3) the Carlson-Livingston series on expected inflation.5

Table 1.

Maximum Likelihood Estimates: Five-Year-Ahead Forecasts

(1960:1 to 1983:4)

article image
Note: The dependent variable in equations (1.1)–(1.4) is the nominal rate on ten-year U.S. Treasury bonds. The values in parentheses are the absolute values of the t-statistics; P denotes the expected inflation rate; DEF 21 denotes the expected deficit; G 21 denotes expected government expenditure; M 21 denotes expected money growth; R2 denotes the adjusted coefficient of determination; DW denotes the Durbin-Watson statistic; F denotes the F-statistic; (**) denotes significance at the 1 percent level; and (*) denotes significance at the 5 percent level.
Table 2.

Maximum Likelihood Estimates: Three-Year-Ahead Forecasts

(1960:1 to 1983:4)

article image
Note: The dependent variable in equations (2.1)–(2.4) is the nominal rate on ten-year U.S. Treasury bonds; the dependent variable in equations (2.5)–(2.8) is the nominal rate on three-year U.S. Treasury bonds. The values in parentheses are the absolute values of the t-statistics; P denotes the expected inflation rate; DEF 13 denotes the expected deficit; G 13 denotes expected government expenditures; M 13 denotes expected money growth; R2 denotes the adjusted coefficient of determination; DW denotes the Durbin-Watson statistic; F denotes the F-statistic; (**) denotes significance at the 1 percent level; and (*) denotes significance at the 5 percent level.

I will now focus on the sensitivity of the results to the introduction of the expected government expenditures variable (G 21 and G 13), which is given by the ratio of cyclically adjusted expenditures to natural output. As previously suggested by Barth, Iden, and Russek (1985, p. 558), one should include a government expenditures variable in order “to capture differences in coefficients.” The results for equations (1.2), (2.2), and (2.6) indicate that the previous empirical results are affected by the inclusion of an expenditures variable. For all three equations, expected deficits no longer have a significant effect on long-term interest rates.

I have also reported the results for the interest rate equations, which include expected money growth (with and without government expenditures). The coefficients on M 21 and M 13 all have the hypothesized negative sign (two of the coefficients are significant at the 5 percent level, with the remaining four coefficients marginally significant). Only one of the coefficients on the expected deficit variable is significant when M 21 and M 13 are included (see equation (1.3)). Once again, expected deficits appear to have no effect on three-year or ten-year treasury bond rates when expected government expenditures are also included (see equations (1.4), (2.4), and (2.8)).

Empirical support for a positive effect of current or lagged deficits on interest rates remains, as Spiro (1987, p. 403) notes, “as elusive as ever.” However, the approach used by Bovenberg is both intuitively appealing and clearly motivated. Specifically, it is the expectations of deficits (and of other variables) that influence nominal interest rates.6 Bovenberg found that increases in expected deficits cause long-term interest rates to increase; consequently, increases in expected deficits may result in the crowding-out of private expenditures.

The results that I have obtained provide mixed support for the hypothesis that long-term interest rates are an increasing function of expected deficits. The positive relationship between expected deficits and the interest rate variables does not appear to be sensitive either to the use of quarterly data or my measures of expected deficits and expected inflation. However, alternative specifications of the model do not provide additional support for the above hypothesis. If, as Barth, Iden, and Russek (1985) suggest, government expenditures must be included in the interest rate equations, expected deficits no longer have a positive effect on either interest rate. Furthermore, the introduction of expected money growth reduces the significance of the interest rate effects of expected deficits.

It is not my intention to use these results to argue that deficits (actual or expected) have no effect on interest rates. To the contrary, 4 out of the 12 estimated coefficients on the expected deficit variables are either significant or marginally significant. I would, however, like to emphasize that the above results do suggest that additional research, like that presented by Bovenberg, is needed on the formation of expectations and on the appropriate specification of interest rate equations.

REFERENCES

  • Barth, James R., George R. Iden, and Frank S. Russek,Federal Borrowing and Short-Term Interest Rates: Comment,Southern Economic Journal, Vol. 52 (October 1985), pp. 55459.

    • Search Google Scholar
    • Export Citation
  • Bovenberg, A. Lans,Long-Term Interest Rates in the United States: An Empirical Analysis,Staff Papers, International Monetary Fund (Washington), Vol. 35 (June 1988), pp. 38290.

    • Search Google Scholar
    • Export Citation
  • Carlson, John A.,A Study of Price Forecasts,Annals of Economic and Social Measurement, Vol. 6 (Winter 1977), pp. 2756.

  • Evans, Paul,Do Large Deficits Produce High Interest Rates?American Economic Review, Vol. 75 (March 1985), pp. 6888.

  • Evans, Paul, (1987a), “Do Budget Deficits Raise Nominal Interest Rates?Journal of Monetary Economics, Vol. 20 (September), pp. 281300.

    • Search Google Scholar
    • Export Citation
  • Evans, Paul, (1987b), “Interest Rates and Expected Future Budget Deficits in the United States,Journal of Political Economy, Vol. 95 (February), pp. 3458.

    • Search Google Scholar
    • Export Citation
  • Gordon, Robert J., Macroeconomics (Boston: Little, Brown and Company, 4th ed., 1987).

  • Hoelscher, Gregory,Federal Borrowing and Short Term Interest Rates,Southern Economic Journal, Vol. 50 (October 1983), pp. 31933.

    • Search Google Scholar
    • Export Citation
  • Kim, Sun-Young, and Raymond E. Lombra,Why the Empirical Relationship Between Deficits and Interest Rates Appears So Fragile,Journal of Economics and Business, Vol. 41 (August 1989), pp. 24151.

    • Search Google Scholar
    • Export Citation
  • Kolluri, Bharat R., and Demetrios S. Giannaros,Budget Deficits and Short-Term Real Interest Rate Forecasting,Journal of Macroeconomics, Vol. 9 (Winter 1987), pp. 10925.

    • Search Google Scholar
    • Export Citation
  • Muller, Patrice, and Robert Price, “Public Sector Indebtedness and Long-Term Interest Rates,” paper presented at the World Bank-Brookings Workshop (unpublished; Washington, September 1984).

    • Search Google Scholar
    • Export Citation
  • Spiro, Peter S.,The Elusive Effect of Fiscal Deficits on Interest Rates: Comment on Tanzi,Staff Papers, International Monetary Fund (Washington), Vol. 34 (June 1987), pp. 40003.

    • Search Google Scholar
    • Export Citation
  • Wachtel, Paul, and John Young,Deficit Announcements and Interest Rates,American Economic Review, Vol. 77 (December 1987), pp. 100712.

    • Search Google Scholar
    • Export Citation
*

David W. Findlay is an Assistant Professor in the Economics Department at Colby College, Waterville, Maine.

1

Bovenberg uses both semi-annual and annual data.

2

None of the results obtained with this measure of expected deficits yielded positive and significant coefficients. This is consistent with the results of Muller and Price (1984) who show that a centered moving average of over seven to nine years is necessary to yield significant and positive results. For space considerations, I do not report the results here.

3

It is not clear whether individuals adopted the CBO forecasts (when they became available) in favor of, for example, the five-year ahead moving average of actual deficits (that is, DEF 21).

4

The coefficient on DEF 13 in equation (2.1) is marginally significant at the 5 percent level; given the degrees of freedom, the critical value of the t-statistic is 2.00.

5

I did not estimate the effects of DEF 21, G 21, and M 21 on the three-year treasury bond, because these series include expectations of, for example, deficits that occur after the security matures. This is similar to a point that Wachtel and Young (1987) make.

6

Current and lagged values of deficits will influence interest rates to the extent that they have an effect on the expected deficit.