Effects of fiscal policy in the goods market, evidence of asymmetric adjustment in USA and Canada: a cointegration analysis.

AuthorWane, Abdoul

    Recent years have been characterized by a lively debate on the economic role that governments should play. Many authors have investigated the effects of government fiscal policy on output and several views have sprung. We do not have the pretensions of presenting all those views but rather we will discuss just a few of them.

    Kandil (2001) using quarterly data for the United States studied the asymmetry in the effects of the government spending shocks. She showed that while interest rates increase in the face of expansionary government spending shocks, there is no evidence of a reduction in the face of contractionary shocks. Consequently, the increased government spending crowds out private investment. Moreover, Kandil (2001) showed that there is evidence of a reduction in private consumption as agents anticipate a future increase in taxes to finance the increased government spending. As a result, output growth and price inflation are decreasing despite expansionary government spending shocks, on average, over time. In view of this evidence, Kandil (2001) suggested that public finance considerations ought to dominate attempts to stimulate demand using government near full-equilibrium capacity utilization in the economy. In contrast, contractionary government spending shocks are not offset by an increase in private spending. Hence, demand contraction is pronounced, slowing output growth and price inflation in the face of a reduction in government spending. For Kandil (2001), the implication is that concerns over the pronounced contractionary effects of a reduction in government spending ought to dominate public finance considerations near full-equilibrium.

    Sorensen and Yosha (2001) examined the business cycle behavior of state fiscal policy to determine whether policy is asymmetric and, if so, to identify the causes. They concluded that state revenue and expenditure display significant asymmetry over the business cycle, with nearly offsetting effects on the budget surplus. As a result, state fiscal policy tends to mute economic booms to roughly the same degree it mitigates slowdowns. The asymmetries in revenue and expenditure appeared to be associated with balanced budget rules, although their fundamental causes cannot be clearly identified. Sorensen and Yosha (2001) tried to find out the potential causes of asymmetric fiscal policy. They pointed out that asymmetry in fiscal policy may arise for a variety of reasons, including market constraints, balanced budgets rules, lack of budget discipline in upturns, and incumbent political parties trying to influence voting patterns or force the hand of future governments.

    Sorensen, Wu and Yosha (2001) found evidence of political business cycle asymmetry in the fiscal policy of U.S. state and local governments during the period 1978-1994.

    The objective of this paper is to investigate the asymmetric effects of government spending shocks (i.e. fiscal policy) on variables in the goods market (such as the real output and the price level), and the asymmetric adjustment of fiscal policy on the two major components of output will be also examined. Specifically, we will examine whether expansionary and contractionary government spending shocks have the same effects on the variables mentioned above. Employing newly developed techniques, the analysis will provide evidence on asymmetry in the effects policy shocks.


    We will modify and improve Kandil's (2001) methodology which is very similar to that of Cover (1992). We will use new and more appropriate estimating techniques such as unit-root and cointegration tests with possibility of error correction. Since Kandil (2001) used data containing over 30 years of observations, it is possible, perhaps likely, that the data are non-stationary. Therefore, estimation of equations based on such data would subject the results to "spurious" concerns raised by Granger and Newbold (1974). Another concern pertaining to time-series considerations as pointed out by Hendry (1986), Granger (1998) and Hakkio and Rusch (1989), is that estimating economic relationships such as productivity, unemployment and wage to the exclusion of the long-run association between the variables may create biases. This is particularly the case if the variables are cointegrated. Therefore, a better way of proceeding is to investigate whether there exists a long-run relationship between the variables. Building upon Kandil's (2001) empirical models, several models will be used to test the asymmetric effects of government spending shocks.

    To test the asymmetric effects of government spending shocks in the goods market, the following two models will be estimated:

    (1) [y.sub.t] = [[gamma].sub.1][g.sub.t] + [[gamma].sub.2] [o.sub.t] + [[gamma].sub.3] [t.sub.t] + [u.sub.yt]. (1)

    (2) [p.sub.t] = [[beta].sub.0] + [[beta].sub.1][g.sub.t] + [[beta].sub.2][o.sub.1] + [u.sub.pt] (2)

    Where y delineates the log value of real output; t is the log value of tax revenues; p is the log value of the price level; and [u.sub.yt] and [u.sub.pt] are the disturbance terms.

    In terms of Equation (1), if the output, the government spending, the energy price and the tax revenues are all I (1) and the linear combination [y.sub.t] = [[gamma].sub.0] - [[gamma].sub.1][g.sub.t] - [[gamma].sub.1][o.sub.t] - [[gamma].sub.3][t.sub.t] = [u.sub.yt] is stationary, and then the

    variables are cointegrated of order (1.1). The vector [x.sub.t] is ([y.sub.t], 1, [g.sub.t], [o.sub.t], [t.sub.t]) and the cointegrating vector [gamma] is (1 - [[gamma].sub.0] - [[gamma].sub.1] - [[gamma].sub.2] - [[gamma].sub.3]). According to Enders (1995), the system is in long-run equilibrium when y[x.sub.t] = 0. The deviation from long-run equilibrium-called the equilibrium error- is [u.sub.yt] so that [u.sub.yt] = y[x.sub.t]. In terms of equations (2) the same procedure also applies.

    The equations mentioned above will be estimated for long-run relationship and for cointegration allowing for TAR and MTAR adjustment following Engle and Granger's (1987) methodology. The testing procedure is described in the following section.

    Threshold and Momentum Models of Cointegration

    The Engle and Granger (1987) methodology as applied to the fiscal policy model begins by positing a long-run equilibrium relationship of the forms given for example in equation (1). The next step in the Engle and Granger procedure focuses on the OLS estimate of ? in the following regression equation:

    [DELTA][[mu].sub.t] = [DELTA][[mu].sub.t-1] + [[epsilon].sub.t] (3)

    Where the estimated regression residuals from (1) and (2) are used to estimate (3).

    Rejecting the null hypothesis of no cointegration (i.e., accepting the alternative hypothesis -2

    The implicit assumption of symmetric adjustment is problematic if money-supply shock adjustment is asymmetric. A formal way to introduce asymmetric adjustment is to let the deviations from the long-run equilibrium in equation (3) behave as a Threshold Autoregressive (TAR) process. Thus, it is possible to replace (4) with:

    [DELTA][[mu].sub.t] = [I.sub.t][[rho].sub.1][[mu].sub.t-1] + (1 - [I.sub.t])[[rho].sub.2][[mu].sub.t-1] + [[epsilon].sub.t] (4)

    Where [l.sub.t] is the Heaviside indicator such that:


    Asymmetric adjustment is implied by different values of [[rho].sub.1] and [[rho].sub.2]; when [[mu].sub.t] is positive, the adjustment is [[rho].sub.1] [[mu].sub.t], and if [[mu].sub.t] is negative, the adjustment is [[rho].sub.2] [[mu].sub.t]. A sufficient condition for stationarity of {[[mu].sub.t]} is for: -2

    Since the exact nature of the non-linearity may not be known, it is also possible to allow the adjustment to depend on the change in [[mu].sub.t] (i.e. [DELTA}[[mu].sub.t]) instead of the level of [[mu].sub.t]. In this case, the Heaviside indicator of (5) becomes:


    Even though Hansen (1997) shows that setting the Heaviside indicator using [DELTA] [[mu].sub.t] can perform better than the specification using pure TAR adjustment, Enders and Granger (1998) and Enders and Siklos (2001) show that the series exhibits more "momentum" in one direction than the other. They call this model Momentum-Threshold Autoregressive (M-TAR) model. Respectively, the F-statistics for the null hypothesis [[rho].sub.1] = [[rho].sub.2] = 0 using the TAR specification of (5) and the M-TAR specification of (6) are called [[PHI].sub.[mu]] and [[PHI]*.sub.[mu]]. As there is generally no presumption as to whether to use (5) or (6), the recommendation is to select the adjustment mechanism by a model selection criterion such as the AIC.

    If the errors in equation (4) are serially correlated, it is possible to use an augmented threshold model for the residuals. In this circumstance, equation (5) is replaced by:


    The distributions of [[PHI].sub[mu]] and [[PHI]*.sub.[mu]] depend on the number of observations, the number of lags in equation (4) and the number of...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT