Estimation Results

Our application of the McCallum rule to China utilizes quarterly data on money supply series, nominal GDP, foreign exchange reserves, and the real exchange rate that are drawn from the Great China Database and the International Monetary Fund’s International Financial Statistics.

Figure 4.2. Actual vs. Simulated Values of Nominal Output, 1990-2006.

In our analysis, the unobservable target nominal GDP growth is modeled based on the trend properties of the raw data, which is not only required by the absence of a consistent series of official target values but also in keeping with the long-standing Friedmanite rationale for linking money growth to trend growth in output.

In reapplying the McCallum rule to an extended 1990-2006 dataset, the estimates reported in this chapter not only allow the response to the nominal GDP output gap to be determined by the data but also simultaneously allow for money supply reactions to the rate ofchange of foreign exchange reserves,

Table 4.3. Money supply reactions based on extended McCallum Rule Estimates, 1990-2006

Notes: Standard errors are reported beneath the coefficients; ***, **, and * indicate significance at the 99%, 95%, and 90% confidence levels, respectively.

^a Denotes significance at the 89.5% confidence level.

^b Denotes the exact significance level for the J-test statistic - with the insignificant values implying that the over-identifying restrictions required for the GMM estimation could not be rejected (see text under equation (2) for the list of instruments).

the rate of change of the real exchange rate, and a post-Asian financial crisis dummy set equal to one after 1997. The estimation period is from the second quarter of 1991 through the second quarter of2006. Table 4.3 reports a series of regressions of the form:

and all other variables are as previously defined.

Ordinary least squares (OLS) estimates are reported first. These estimates may not accurately reflect the People’s Bank of China’s true reaction to the

right-hand-side variables in equation (2), however, as OLS estimation co­mingles the central bank’s policy response with more general economy-wide reactions. This is due to the endogeneity, or non-predetermined nature, of the right-hand-side variables. Hence, the final two column of Table 4.3 show the results of reestimation using the Generalized Method of Moments (GMM) procedure to correct for this endogeneity issue. The constant, post­Asian financial crisis dummy, two lags of each of the variables in the equa­tion, and two lags of export growth were used as instruments to apply the GMM procedure.^[68]

The 0.5 coefficient value for β ₁ hypothesized by McCallum (1988) is not supported in the Chinese case, and the GDP gap is generally statistically insignificant - in the one case where it is marginally significant, the negative sign actually implies an accommodative, procyclical response to nominal output movements.^[69] The post-Asian financial crisis dummy is negative and highly significant for both M2 and the monetary base across each of the two estimation procedures.

Insulating the overall level of the monetary base from the effects of reserve inflows requires that the domestic component of the base be reduced, on a one-to-one basis, when the foreign component is forced up by new reserve accumulation. Complete, or near-complete sterilization, as discerned by both He et al. (2005) and Ouyang, Rajan, and Willett (2007), implies the overall monetary base remaining invariant to increased reserve inflows,

Figure 4.3. Actual vs. Fitted M2 Growth Rates Based on GMM Estimation.

therefore, consistent with the present findings.^[70] On the other hand, failure to fully sterilize reserve inflows should be associated with a positive mone­tary base reaction to higher rates of reserve accumulation. Finally, there is evidence of a negative reaction to the real exchange rate in three cases out of four, indicating a countercyclical (stabilizing) response to real exchange rate appreciation. Real exchange rate appreciation, as experienced during much of the 1990s (Chapter 1), can arise when Chinese prices are rising faster than prices abroad. Tighter monetary policy that reduced Chinese inflation rates would tend to offset such real exchange rate appreciation.

Our final exercise is to compare the predicted values from the regressions reported in Table 4.3 to actual money growth rates observed over the 1991— 2006 period.

Figure 4.4. Actual vs. Fitted Monetary Base Growth Rates Based on GMM Estimation.

loose during the 1993-1995 inflation spike - but otherwise shows actual M2 growth according quite well with the forecast values implied by the extended McCallum policy rule. Meanwhile, Figure 4.4 suggests a more moderate, but also more extended, tendency for overly high growth rates for the monetary base during 1993-1998. After that, monetary base growth looks to have become too tight in 1998-1999, at the time that the Chinese economy entered deflation, but subsequently follows a similar trend to the forecast values. Overall, the strong suggestion of overly loose policy in the early to mid-1990s followed by indications (from the performance of monetary base growth) of overly tight policy at the end of the 1990s seems quite consistent with actual experience. It is also interesting to see that sustained deviations from forecast values were generally avoided after 2000 in spite of the pressures arising from record rates of foreign reserve accumulation and the claimed undervaluation of the renminbi.