The Ricardian equivalence hypothesis: evidence from Bangladesh

Abstract

This article examines the Ricardian equivalence hypothesis (REH) and its sources of failure in the case of Bangladesh using various theoretical specifications, annual data from 1974–2001 and linear and nonlinear time series techniques. The general findings tend to invalidate the REH: a finite time horizon and the presence of liquidity-constrained individuals are the sources of deviation from the REH. Empirical results reveal that real per-capita private consumption (C) under various specifications is cointegrated generally at the 5% level with real per-capita income (Y), government expenditure before and after interest rate repayments (G and G2), taxes (T) and the interest rate (r). Results reveal that an increase in G, G2, T and r reduces C and that an increase in budget deficits raises trade deficits. These results highlight the importance of fiscal policies in boosting private consumption and controlling trade deficits, which are the prime goals of stabilization policies being followed by Bangladesh.

Acknowledgements

I am particularly grateful to Paul Auerbach for his helpful comments on the previous drafts of this article. I am also grateful to an anonymous referee for his useful comments to improve this article and to Vince Daly for his useful comments on Section II on the methodology. The usual disclaimer applies.

Notes

¹The average of fiscal deficits during our sample periods is about 6% of GDP, ranging from 3–9% of GDP and the average of trade deficits is more than 7% of GDP, ranging from 3.54–12.51% of GDP.

²These policies are prescribed by international institutions such as the World Bank and the International Monetary Fund, and are based on demand management policies which suggest reducing the budget deficits in order to reduce trade deficits and to increase private investment, thereby increasing the income and consumption.

³For example, Ghatak and Ghatak (1996); Gupta (1992); Haque (1988); Khalid (1996).

⁴The relationship, according to the twin deficits hypothesis, between government budget deficits and trade deficits can be summarized as follows (see Khalid and Guan (1999) for a good review). Firstly, in a Mundell–Flemming framework, an increase in government deficits are thought to exert an upward pressure on real interest rates, which boosts capital inflows and hence, causes an appreciation in real exchange rates and a reduction in competitiveness, causing trade (or current account) deficits (Rosenweig and Tallman, 1993, p. 580; Khalid and Guan, 1999, p. 390). This mechanism is effective under both fixed and flexible exchange rate regimes. Under a fixed exchange rate regime, trade deficits deteriorate due to a positive income effect caused by the government's excess expenditures and due to an appreciation in real exchange rates. Secondly, the Keynesian absorption theory predicts that a rise in budget deficits increases domestic absorption and hence an expansion in imports, causing current account deficits (see from Khalid and Guan, 1999, p. 390). A strong correlation between saving and investment (Feldstein and Horioka, 1980) also causes budget deficits and the current accounts of the balance of payments to move together, supporting the twin deficits hypothesis.

⁵Note, however, that LDCs in general, and Bangladesh in particular, are characterized by imperfect capital markets (Ghatak, 1995; Siddiki, 2000, 2002; Auerbach and Siddiki, 2004).

⁶The violation or failure of the REH implies that a government can affect trade deficits or private consumption by changing the timing of taxes.

⁷I am thankful to Vince Daly for his through and useful comments on this section. I am thankful to the referee for suggesting me to apply the three-step Engle and Yoo (1991) method in our analyses.

⁸ A positive value of θ gives a negative coefficient for G2 and thus implies that an increase in G2 reduces C, i.e. government spending is a substitute for private spending. On the other hand, a negative value of θ gives a positive coefficient of G2, implying that an increase in G2 raises C, i.e. government spending is a complement to private spending.

⁹A zero value of μ supports the assumption of infinite horizon that the individual's subjective probability of survival is unity while a positive value of μ, i.e. a fraction of population (μ) dies each period, indicates a finite horizon or survival rate.

¹⁰I am thankful to the referee for pointing the weaknesses of the Engle and Granger (Engle and Granger, 1987) method and suggesting me to apply the three-step Engle and Yoo (1991) method in our analyses.

¹¹I am thankful to the referee for suggesting me to incorporate both types of analyses.

¹²The restriction a ₁ < 1 implies that marginal propensity to consume is less than one; a ₁ = |a ₂| implies that consumption losses due to an imposition of taxes are equal to consumption gains resulting from a same amount of increase in income or vice versa; a ₂ = a ₃ asserts that deficits incurred by the government today will be completely offset by rising taxes in the next period.

¹³As explained in the footnote of Table 4, Eviews gives somewhat unstable and implausible results, which are mainly caused by the misspecification of the model, since only past values of G2 and d may not be enough to estimate E_{t −1} G2 _{t −1}. In addition, selecting the number of lags to be used for annual data is arbitrary and difficult and such problems become very acute with the short time series as is the case for Bangladesh.

¹⁴As explained above, the finite horizon (μ > 0) and the presence of liquidity-constrained individuals are considered as the main sources of deviation from the REH. Estimated results from the linear and nonlinear version of Equations 3.11c, 3.11d and 3.11e are used to explore the sources of the departures of the REH (Tables 5 and 6). The presence of infinite horizon, i.e. μ = 0, suggests that consumption in all the three approaches follows a random walk, only lagged values of consumption rather than any other variables explains current consumption (Hall, 1978). A linear model test the hypothesis that all coefficients other than the coefficient of lagged consumption are insignificant, i.e. consumption follows a random walk model, implying that the coefficient of lagged consumption is one (Hall, 1978). On the other hand the nonlinear models examine whether μ = 0 and α = 1.