Testing For The Asymmetry Of Monetary Policy In China

This paper tests for the presence of asymmetric effects of monetary policy on aggregate activity using Chinese quarterly data from 1991 to 2006. The types of asymmetry effects of monetary we will test for include: (1) whether negative and positive monetary policy shocks have different effects on aggregate output;(2) whether big or small shocks have different effects; and/or (3) whether low-variance, negative shocks have real effects on aggregate output. We will introduce the three types of the monetary policy and explain under which condition these asymmetries might take place.To date, the empirical literature has focused on a particular asymmetry that we call"the traditional Keynesian asymmetry", which states that positive monetary policy shocks have smaller real effects than negative monetary policy shocks, or, in a more extreme form, that only the negative monetary shocks have real effects. This asymmetry can be derived under the assumption of either downward (upward) sticky (flexible) nominal wages or sticky prices together with rationing of demand.We also test for the asymmetric effects that are implied by models with menu costs (Ball and Romer, 1990, and Ball and Mankiw, 1994) in this paper. In static (deterministic) settings, standard menu-cost models imply that"big"monetary policy shocks are neutral because firms would find it optimal to adjust nominal prices, while"small"monetary policy shocks would have real effects because whether to adjust nominal prices for a firm is related to menu-cost. In other words, the firms have to decide whether to index their prices (at the cost of paying the menu cost) or not before the monetary policy shock is observed. Firms will choose indexation (which implies neutrality) only if the variance of monetary policy shocks is high. We extend the analysis by assuming that the monetary policy process can change between having a"high"variance and a"low"variance.Finally, we consider the case in which only small negative shocks to nominal demand affect real aggregate activity. Consider a dynamic menu-cost model in which there is positive steady-state inflation; firms can change prices costlessly every second period, but, if firms want to change prices in between the two periods. They must pay the menu cost. The conclusion draw from the above model is that only the little negative monetary shocks effect aggregate output. We call the asymmetry type as"hybrid asymmetry", because it has similarities both to the traditional Keynesian asymmetry and to the menu-cost asymmetry.To test for the asymmetric effects described above, we use a procedure that consists of estimating a monetary policy process that allows for changes in regime using the regime-switching model of Hamilton (1988) appropriately modified to our setting. We assume that the money supply is a regime-switching process that allows for changes in the mean and in the variance of the monetary supply shocks. This implies that we can distinguish between four different shocks to monetary policy: big positive shocks, big negative shocks, small positive shocks, and small negative shocks. The distinction between"big"and"small"here refers to the variance of the shocks in two states.The monetary policy equation we estimate in this paper is: GM 1t -μ( st )=Φ( L )( GM1t -1 -μ( st -1 )) +Θxt -1+σ( st )ηt Where GM 1is the rate of money supply M 1 for the same quarter,Φ( L) is a lag polynomial,Θis a vector of parameters, xt ' -1 is a vector of de-meaned predetermined variables(we includes as regresses the rate of GDP for the same quarter, defined as GY , the difference of interest rate, defined as Dr , and the rate of consume price index for the same quarter, defined as GCPI ), defined as x -μx.μ( st)is a state-dependent mean, st is the discrete-valued state variable, andηt is an i.i.d N (0,1) error term that is independent of st .We find that the changes in both the mean and the variance of the process are significant from the result of the estimation of the above equation. The estimates suggest that there are a low-mean, low-variance regime where the mean is around 0.235, and variance is around 0.014 and a high-mean, high-variance regime where the mean is around 0.338, and variance is around 0.056. We note also both regimes are quite persistent, since the transition probabilitiesπ11 andπ22 are both in excess of 0.97. Through estimating the money supply equation, we can gain four residual series, defined as big positive shocks, big negative shocks, small positive shocks, and small negative shocks.This technique allows us to test for the existence of the three cases of asymmetric effects discussed above after we distinguish the monetary supply shocks into four different types. We estimate an output equation, which includes the (change in the) current unanticipated shocks from the"monetary policy"relationship. We then test for asymmetries by introducing various parameter restrictions on the four different types of unanticipated monetary policy shocks in the output equation and by applying wald-test or likelihood ratio tests.The output equation we apply in this paper is:We find the evidence that small money shock have more effects on output than big money shock. But we can not find the evidence of traditional Keynes asymmetry and hybrid asymmetry.