The asymmetric effects of demand shocks: international evidence on determinants and implications

Abstract

The analysis focuses on the asymmetric effects of demand shocks. The evidence across a sample of 19 industrial countries differentiates the effects of expansionary and contractionary aggregate demand shocks on real output growth and nominal wage and price inflation. The difference appears consistent with a kinked supply curve that is dependent on the asymmetric flexibility of wages and/or prices across countries. Furthermore, the evidence does not support the endogeneity of asymmetric nominal flexibility with respect to demand variability or trend price inflation across countries. On average, across countries, demand variability increases nominal wage and price inflation relative to deflation, while exacerbating output contraction relative to expansion. The apparent trade-off between changes in real and nominal trends provides further support to the supply side explanation of asymmetry.

Notes

¹ Cover (1992) and Kandil (1996) illustrate the evidence using data for the United States. Kandil (1995) demonstrates the asymmetric effects of demand shocks using quarterly data for real output, the nominal wage and the price level across industrial countries. Other research on asymmetry includes De Long and Summers (1988) and Romer and Romer (1989). Karras (1996a, b) demonstrates the asymmetric effects of monetary shocks using annual data for output and price. Kandil (1998) demonstrates the nonneutrality of aggregate demand shocks using aggregate data for a sample of developing and developed countries. Kandil (1999) demonstrates the asymmetric stabilizing effects of price flexibility using pre- and post-war data for a sample of industrial countries. Weise (1999) tests the asymmetric effects of monetary policy using a VAR-based model. Kandil (2001) and (2002a) demonstrates asymmetry in the effects of government spending shocks using quarterly data for the United States. Kandil and Mirzaie (2002) investigate possible asymmetry in the effects of exchange rate fluctuations on sectoral output and price in the United States.

² For more details on this theoretical possibility, see Cover and Van Hoose (1993) and Kandil (2002b). Insider–outsider models attempt an alternative explanation for asymmetric nominal wage adjustments (see e.g. Lindbeck and Snower, 1986; Blanchard and Summers, 1988).

³ For more details on this theoretical possibility, see Caballero and Engle (1991), Caplin and Leahy (1991), Tsiddon (1991) and Ball and Mankiw (1994).

⁴ Examples are models of credit rationing policies (e.g. Bernanke, 1983).

⁵ For related illustrations, see Karras (1996b).

⁶ The mean growth rate is determined by productive resources, e.g. technical progress and population growth.

⁷ The relationship in (1) is implied by the reduced-form solutions in standard business-cycle models that incorporate the rational expectation hypothesis. These include equilibrium business-cycle models pioneered by Lucas (1973) and new Keynesian models advocating nominal wage rigidity, for example, Fischer (1977), or price rigidity, for example, Ball et al. (1988).

⁸ This correlation is the basic premise of theoretical models that emphasize wage rigidity in the labour market as an explanation of business cycles.

⁹ The analysis focuses on asymmetry in the effects of the shocks, i.e. α_pj and α_nj . For a theoretical illustration of variation in the degree of asymmetry with the size of the shocks, pos_nt and neg_nt , see Evans (1985).

¹⁰ During boom periods, cost of living adjustments may be specified to guarantee workers upward adjustment of wages to keep up with inflation. In contrast, firms may be reluctant to take aggressive measures towards adjusting wages in the downward direction. This is because the search and training cost of hiring new workers to accommodate a future rise in demand may actually exceed the perceived loss of retaining workers at wages that exceed the marginal physical product of labour during recessionary periods.

¹¹ Jackman and Sutton (1982) set a similar argument by focusing on the effect of interest rate changes on spending. They report that as interest rates rise (in response, e.g. a tight monetary policy) consumption spending falls the full amount as a result of the increase in debt payments. In contrast, a decrease in interest rates (in response, e.g. an expansionary monetary policy) induce higher levels of spending, but by an amount less than the change in liabilities. Similarly, Bernanke and Gertler (1989) analyse the relation between changes in the interest rate and investment demand. They find that large drops in investment are more likely to occur than large increases.

¹² Quarterly data are not available for all variables over this span.

¹³ Theoretical explanations have focused on asymmetry in the propagation mechanism of unanticipated shocks. In contrast, agents adjust fully to anticipated shifts, which renders the explanation of asymmetry, in general, unapplicable.

¹⁴ Detailed results are available upon request.

¹⁵ Nominal Gross National Product (GNP) or GDP approximates the total value of aggregate spending on goods and services in the economy. It is likely, however, that this measure is affected by major supply-side factors. To trace the shifts of the AD curve for a given constant AS curve, it is important, therefore, to account for major sources of supply-side shifts, e.g. the energy price.

¹⁶ Many standard business-cycle models share the neutrality hypothesis. Accordingly, anticipated demand shifts are absorbed in nominal variables without affecting real output. By construction, anticipated demand shifts are a function of lagged changes in wage, price and output. This eliminates the need to account for the lagged dependent variable in Equations 7 and 8.

¹⁷ Negative parameters would indicate counter-cyclical response to demand shocks. For example, real output growth may be increasing despite recessionary demand conditions.

¹⁸ Negative parameters would signal nominal rigidity. For example, price inflation may be increasing despite recessionary demand conditions.

¹⁹ t-statistic is the ratio of the difference in coefficients to the square root of its variance.

²⁰ By construction, α_py + α_pp = α_ny + α_np = 1. This constraint holds statistically for various countries.

²¹ The cross-section analysis employs estimates from the time-series regressions. Point estimates approximate the average response of variables to aggregate demand shocks over time. The cross-country regressions follow the generalized least squares estimation method suggested by Saxonhouse (1977). To account for the variability of time-series estimates, each parameter is weighted by the inverse of the SE in the time-series regression. That is, estimates with high SEs (statistically insignificant) are weighted less heavily in the cross-country analysis.

²² Correlation measures the closeness of a linear relationship between variables. A correlation of 0 between two variables means that each variable has no linear predictive ability for the other. The sample correlation approximates the Pearson product-moment correlation. It is computed:

where, and are the sample means of x and y. The significance probability approximates the significance level for the null hypothesis of a zero correlation.

²³ Note that the coefficients measuring price flexibility are different across various countries. The correlation coefficients measure how this variation, in the upward direction, correlates with variation in the downward direction, across countries.

²⁴ By construction, aggregate demand shocks are distributed between real output growth and price inflation.

²⁵ By construction, demand shocks are exogenous to the economic system. Data for the cross-section analysis are provided in Table A2.

²⁶ The highest trend inflation is evident for Spain, 11% on average over time, which is characterized by a higher downward price flexibility relative to upward flexibility in response to aggregate demand shocks (Table 1).

²⁷ This evidence is consistent with earlier discussion on experiences with stopping inflation, see e.g. Yeager (1981).

²⁸ Details are provided in Table A2 of Appendix 2.

²⁹ The evidence in Table 4 indicates, however, that higher trend inflation counters the effect of demand variability as countries attempt to slow down inflation.