The Taylor Rule - an Empirical Review of China’s Monetary Policy

...

Real GDP, potential real GDP and the output gap

We obtained the real GDP data from the National Bureau of Statics of China (NBS), which is the accumulated GDP per quarter in 100 million current Yuan. This data was converted into quarterly GDP by subtracting the accumulated GDP in quarter 2 from the GDP quarter 1 etc. We estimated the potential real GDP by using the linear trend estimate, as has been used by Ping and Xiong (2002). Since the output figures we have demonstrate potential quarter-related varied characteristics, we implemented three dummy variables.

[pic 3][pic 4]

The potential output can be based on the following equation

[pic 5][pic 6][pic 7][pic 8] (1)

(t) (547) (89.2) (-16.0) (-12.0) (-11.0)

Sample period: 2001Q1-2015Q1

Oberservations: 61[pic 9][pic 10]: 0.99

F-statistic 2082.84

In accordance with equation (1), the potential GDP can be estimated using

[pic 11][pic 12](2)

Using this information we can estimate the output gap with the following expression:

[pic 13][pic 14] (3)

Inflation rate and inflation target

The inflation rate used is measured by using the CPI. This CPI data can be found in the database of the Federal Reserve Bank of St. Louis. Information regarding the inflation rate target was collected by reading “The Report on Plan Implementation for National Economic and Social Development of the year and the Draft plan for National Economic and Social Development of the following year”. This report is composed annuallyby the State Development and Reform Commission of China. However, in this paper quarterly data is used, while the inflation targets annual, therefore, we set the quarterly target equal to the annual target (table 1).

Interest rate and long run real interest rate

The official interest rate used is the average of the 1-year lending rate, this rate is set by the PBC and can be found on their website. The long run equilibrium real interest rate posed a dilemma, the literature does not provide a clear cut answer to which rate should be used. In the literature both the median and the average real interest rates are used as the long run equilibrium real interest rate (Ping, Xiong 2002). We chose to go with 4.7%, since this is the most recent data we could find in the literature concerning the equilibrium long real run interest rate (He, Wang 2012).

-

stationarity tests

The Augmented Dickey-Fuller (ADF) test was used to test the variables used in the Non-Linear Least Squares (NLS) regression. The outcomes of those tests can be found in table 2.[pic 15][pic 16]

- Establishment of model

The original Taylor rule states that the interest rate determined by the central bank can be described by Taylor (1993):

[pic 17][pic 18] (4)

Where[pic 19][pic 20] is the interest rate set by the central bank, LRRI is the long run equilibrium real interest rate, [pic 21][pic 22] is the realized inflation as measured by the Consumer Price Index (CPI) on a quarterly basis,[pic 23][pic 24] is the information available to the central bank at time t, [pic 25][pic 26] is the inflation target set by the State Development and Reform Commission of China,[pic 27][pic 28] can be described and will be referred to in this paper as the CPI-gap, [pic 29][pic 30] is the output gap. We expect the parameters [pic 31][pic 32] and [pic 33][pic 34], since this is in line with the Taylor (1993). What’s more, [pic 35][pic 36] would represent an active interest rate policy towards the CPI gap, whereas [pic 37][pic 38] would indicate a more passive policy. Furthermore, a higher value for [pic 39][pic 40] would indicate a more responsive interest policy towards the output gap. The inflation rate is highly persistent, therefore [pic 41][pic 42] can be applied (Fan et al. 2011).

However, in reality central banks apply a degree of smoothness to the rate. In the literature a weighted average method is used which comprises of the target interest rate, [pic 43][pic 44], the existing interest rate [pic 45][pic 46], and an error term, [pic 47][pic 48]. This can be summarized in the following expression:

[pic 49][pic 50] (5)

The [pic 51][pic 52] conforms the original Taylor rule, furthermore, [pic 53][pic 54]. When (4) and (5) are combined we obtain the following definition for the Taylor rule for the benchmark interest rate, [pic 55][pic 56]:

[pic 57]

Results

This section presents our findings based on the Taylor rule (6) from section 4. We faced substantial heteroskedasticity in the error term. Interestingly this seemed to be due to the observations from 2008:3 to 2009:1. This period is especially interesting, since on 9 November 2008 the State Council of the People’s Republic of China announced a $586 billion (current dollars) stimulus package, before this period the PBC already cut the interest rate three times in an effort to bolster economic growth[1]. Furthermore, the PBC cut interest rates for the fourth time that year with 108 basis point, the biggest drop in 11 years, on 26 November 2008[2]. In the literature we could not found similar mentions of the heteroskedasticity issue, which causes the test statics to be incorrect and overstate precision.

At first, we tried to get rid of the heteroskedasticity by applying White heteroskedasticity-corrected standard errors. However, this did not remove the heteroskedasticity issue. What’s more, when these observations are removed we found that the [pic 58][pic 59] increases from 0.81 to 0.92, that the output-gap([pic 60][pic 61]y) becomes more significant, and that the long run equilibrium real interest rate and the CPI-gap become less significant after the removal of these observations (see (7) and (8) and table 3).

[pic 62][pic 63] (7)

(t) (2.21) (3.57) (2.11) (18.03)

Sample period: 2000:1 – 2008:2 2009:2-2015:1

...