Between the mid-1990s and 2003, China had an average annual inflation rate of only about 1 percent, with periods of mild deflation, even though output growth was very strong, averaging more than 8 percent a year (Figure 3.1). Moreover, price increases were minimal despite the accommodative stances of fiscal and monetary policies. This section analyzes the key factors that have led to low inflation or deflation in this high-growth environment. The econometric results suggest that declining commodity prices and tariff cuts exerted significant downward pressures on inflation in China, while demand factors played a smaller role.
Real GDP Growth and CPI Inflation
(Annual percent change)
Source: CEIC database.Real GDP Growth and CPI Inflation
(Annual percent change)
Source: CEIC database.Real GDP Growth and CPI Inflation
(Annual percent change)
Source: CEIC database.Stylized Facts
After a period of sharp price increases, peaking at 27 percent in late 1994, inflation declined rapidly and, by early 1998, China entered into its first episode of deflation (Figure 3.2). Deflation lasted for around two years, with price declines averaging around 1 percent and the largest decline at 2.2 percent in April-May 1999.1 In early 2000, mild inflation emerged, but it was short lived. The largest price increase registered during this period was 1.7 percent (May 2001). Deflation reemerged in late 2001 and lasted for more than a year. Since the beginning of 2003, China has been experiencing low inflation, which reached 3.2 percent (year-on-year) by end-2003.
Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.In examining price changes at a disaggregated level, one pattern stands out: prices of tradable consumption goods declined persistently between 1997 and 2002 (Figure 3.3). In particular, prices of clothing and housing appliances (which have an estimated weight of 14 percent in the aggregate consumer price index (CPI)) declined by a cumulative 8 percent during 1998–2002, and food prices (which have an estimated weight of 35 percent) by a cumulative 10 percent during 1997–2000.2 Starting in late 2002, however, food and energy prices started to increase, and through 2003, tradable goods’ prices have been increasing at a moderate pace.
Components of CPI Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.Components of CPI Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.Components of CPI Inflation
(12-month percent change)
Sources: CEIC database; and IMF staff calculations.Prices of nontradables, on the other hand, increased significantly until 2001. In particular, service prices increased, on average, by 12 percent during 1998–2000. Starting in mid-2001, however, prices of most services began to decline (e.g., medical care) or ceased to increase (e.g., housing). This slowdown seems to have come to an end by late 2002, and prices of most services, including housing, have started to pick up in early 2003. These price developments indicate that, while the first deflationary episode was the result of strong price declines in the tradable goods sectors, weaker service prices were important in the emergence of the second deflationary episode.
While most prices in China are now determined by market forces, the remaining price controls still play a role in aggregate price dynamics. Most price controls were abolished by 1993, but pharmaceutical prices and prices for health care and education are still administratively determined. Most controls are in the form of “guidelines” issued by the State Development and Reform Commission, and are ceilings for prices and fees. The weights of these components in the aggregate CPI are estimated to be around 10–15 percent. There are indications that price increases in these sectors were kept at moderate levels because of concerns about social stability (rising unemployment associated with state-owned enterprise (SOE) reforms and reforms in the agriculture sector), particularly in recent years, when these concerns have intensified.
Potential measurement problems, especially in service sector prices, suggest that changes in the aggregate CPI should be interpreted with caution. While the administered prices in health care and education have not increased significantly, there are indications that prices of similar services provided by nonstate entities have been increasing markedly. Also, the housing price index could be underestimating the actual services received from housing, because the index is linked to government-controlled housing units, which do not reflect actual rental price increases in the market. In addition, the imputed rent is linked to the interest rate and falls as interest rates decline, as was the case in 2002. Nevertheless, the total weight of these components is not large enough to play a significant role in the changes in the aggregate index.
Other measures of aggregate prices show similar movements as in the CPI (Figure 3.4). For example, the retail sales price index, which covers most of the prices that are included in the CPI, but excludes services and includes small retail items that might also be used for investment purposes, moves in parallel to the CPI index. During the sample period, changes in this index remained consistently below changes in the CPI index, reflecting the fact that inflation in services was higher than inflation in other prices. Inflation based on the producer price index also does not deviate from CPI inflation for long periods, although it is more volatile. Moreover, it leads CPI inflation in the sense that it switches direction several months before CPI inflation does.
Alternative Measures of Inflation
(12-month percent change)
Source: CEIC database.Alternative Measures of Inflation
(12-month percent change)
Source: CEIC database.Alternative Measures of Inflation
(12-month percent change)
Source: CEIC database.Key Sources of Price Changes
Both demand and supply factors could potentially account for changes in prices. On the supply side, labor productivity in China has been increasing rapidly as a result of strong investment rates, which averaged 40 percent of GDP in the last 10 years. High investment rates have been supported by a high domestic saving rate and strong foreign direct investment (FDI), which averaged more than $35 billion annually during this period. Competition in domestic markets has intensified, as China has been lowering trade barriers and opening up its markets to foreign competition for many years. This process has been spurred further by China’s WTO accession in December 2001 (see Lardy, 2002; and Section II). In addition, monopolies are in the process of being broken up, and SOE reform is allowing the private sector to increase its share in the economy.
There have also been changes in external and domestic demand. External demand was weak during the Asian crisis, which intensified the downward pressures on prices, not only because of a slowdown in overall economic activity, but also through diversion of some export goods to local markets. On the domestic side, the relatively slow growth of income in rural areas may also have contributed to lower overall prices. On the other hand, macroeconomic policies were accommodative during most of this period, helping to stimulate demand and prop up prices. In particular, the fiscal deficit widened from less than 1 percent of GDP (official definition) in 1996 to 3 percent of GDP in 2002. Monetary policy was also accommodative, especially in the last few years.
Other more transitory factors may also have had a significant impact on inflation in China. World nonfuel commodity prices declined significantly during both deflationary episodes.3 In particular, the world nonfuel commodity index declined by 23 percent (year-on-year) by August 1998, started to recover in early 1999, peaked in late 2000, and then started to decline again. Import tariffs have also been reduced significantly in recent years, with the average tariff rate (unweighted) declining from 23.6 percent in 1996 to 11 percent in 2003 (see Section II). Such a decline would have a direct impact on consumer prices, as well as an indirect effect through lower costs of imports of intermediate goods and other inputs. Moreover, exchange rate changes may have had some impact. Although the renminbi has been closely tied to the U.S. dollar since 1994, China’s nominal effective exchange rate has fluctuated significantly, registering a cumulative appreciation of 35 percent between 1994 and early 2002. Between then and end-2003, the nominal effective exchange rate depreciated by 15 percent.
Regression Results
To capture the potential impact of these factors on price dynamics in China, the following regression was estimated in linear form using quarterly data for the period 1996–2003:4
where inf is CPI inflation; prod is productivity growth in the whole economy; ygap is the output gap, calculated as the deviations from trend output obtained using the Hodrick-Prescott filter; g is the quarterly fiscal balance as a ratio to output; M2 is the rate of growth of broad money; comm is the change in the world nonfuel commodity price index; tariff is the change in the average tariff rate; and neer is the change in the nominal effective exchange rate (Figure 3.5).5
Variables Used in the Regression
Sources: CEIC database; and IMF staff calculations.Variables Used in the Regression
Sources: CEIC database; and IMF staff calculations.Variables Used in the Regression
Sources: CEIC database; and IMF staff calculations.The model assumes that broad money growth affects price changes with a lag; therefore, several lag values of this variable are used in the regression.6 It is also assumed that fiscal balances are not affected contemporaneously by inflation, because taxes are collected with a lag. Regarding the nonfuel commodity prices, China’s increasing trade and demand for investment goods have boosted some specific commodity prices; however, given the relative size of the Chinese economy in the global economy (3–6 percent during the sample period), it is assumed that China does not have a significant impact on the overall commodity price index.
Inflation expectations are not modeled explicitly; however, the estimated coefficients could be considered as the reduced-form coefficients from a richer model that incorporates expectations. In that sense, the coefficients capture direct effects as well as indirect effects through expectation formation. A single equation approach was preferred to a structural vector autoregression (VAR) model commonly used in estimating the impact of demand and supply shocks on output and inflation, because a VAR model that is as rich and contains as many variables requires a longer time series than is available.
The model is estimated using ordinary least squares. First, a general model was estimated that included all the variables that could potentially affect inflation. Next, the variables and lags that were not statistically significant were eliminated sequentially, with the variable that has the lowest probability of having a significant coefficient eliminated first. The final result was not sensitive to the elimination sequence, since different elimination sequences produced similar results. A set of diagnostic tests on the final regression results did not indicate the presence of serial correlation in the error terms or significant break points.7
The results for the final specification are presented in Table 3.1. The results suggest strong persistence in CPI inflation, since the estimated coefficient on the lagged dependent variable is large (0.71), even though it is significantly smaller than one.8 This also implies that the total impact of a shock to inflation is spread across two years, with two-thirds of the total impact observed within the first year.
OLS Estimates of a Reduced-Form Inflation Equation1
(Dependent variable: four-quarter CPI inflation)
Sample period is Q1:1996-Q4:2003. The figures in parentheses are absolute t-statistics, based on standard errors calculated using Newey-West heteroscedasticity and autocorrelation consistent covariances (lag truncation = 3).
OLS Estimates of a Reduced-Form Inflation Equation1
(Dependent variable: four-quarter CPI inflation)
| Variable | Coefficient (t-statistics) |
|---|---|
| Constant | 0.50 |
| (0.41) | |
| Lagged CPI inflation | 0.71 |
| (22.15) | |
| Productivity growth | -0.34 |
| (2.65) | |
| Changes in average tariff rates, lagged twice | 0.08 |
| (2.60) | |
| Output gap as a percent of GDP | 0.51 |
| (3.07) | |
| Fiscal balance in percent of GDP | -0.09 |
| (1.87) | |
| Broad money growth, lagged once | 0.13 |
| (4.68) | |
| Changes in the world commodity price index | 0.05 |
| (6.75) | |
| Number of observations | 32 |
| R-squared | 0.98 |
| Prob (F-statistic) | 0.00 |
Sample period is Q1:1996-Q4:2003. The figures in parentheses are absolute t-statistics, based on standard errors calculated using Newey-West heteroscedasticity and autocorrelation consistent covariances (lag truncation = 3).
OLS Estimates of a Reduced-Form Inflation Equation1
(Dependent variable: four-quarter CPI inflation)
| Variable | Coefficient (t-statistics) |
|---|---|
| Constant | 0.50 |
| (0.41) | |
| Lagged CPI inflation | 0.71 |
| (22.15) | |
| Productivity growth | -0.34 |
| (2.65) | |
| Changes in average tariff rates, lagged twice | 0.08 |
| (2.60) | |
| Output gap as a percent of GDP | 0.51 |
| (3.07) | |
| Fiscal balance in percent of GDP | -0.09 |
| (1.87) | |
| Broad money growth, lagged once | 0.13 |
| (4.68) | |
| Changes in the world commodity price index | 0.05 |
| (6.75) | |
| Number of observations | 32 |
| R-squared | 0.98 |
| Prob (F-statistic) | 0.00 |
Sample period is Q1:1996-Q4:2003. The figures in parentheses are absolute t-statistics, based on standard errors calculated using Newey-West heteroscedasticity and autocorrelation consistent covariances (lag truncation = 3).
Structural factors, in particular strong productivity growth and lower tariffs, appear to have been important determinants of inflation. The coefficient estimates suggest that a 1 percentage point increase in productivity growth would have lowered CPI inflation (or increased deflation) by more than 0.3 percentage point during the same quarter and, owing to the persistence in the inflation variable, inflation would have continued to decline in subsequent periods, but at a diminishing rate. Similarly, a 1 percentage point cut in the average tariff rate would have lowered inflationary pressures by close to 0.1 percentage point with a half-year lag.
Demand-side factors also seem to have played a role in price formation during the sample period. The coefficient on the output gap variable is significant, and implies that a 1 percentage point increase in GDP growth above the trend would have increased inflation by 0.5 percentage point. Similarly, a 1 percentage point increase in growth of broad money would have resulted in a more than 0.1 percentage point increase in inflation. The coefficient on the fiscal balance is significant but small, and suggests that a 1 percent of GDP increase in the fiscal deficit would have increased inflation by less than 0.1 percentage point.
World nonfuel commodity prices also appear to affect inflation in China. The coefficient estimate implies that a 10 percent increase in the world nonfuel commodity price index would have increased the inflation rate by 0.5 percentage point. The nominal effective exchange rate, on the other hand, was not a significant determinant of inflation during the sample period. Its coefficient was consistently insignificant in various regression specifications (sometimes with the wrong sign), and this variable was dropped from the final equation.
Observed movements of the explanatory variables, coupled with the coefficient estimates, suggest that structural factors and world commodity prices were key factors that led to both deflationary episodes (Figure 3.6). Prior to and at the beginning of the first deflationary episode, there were large reductions in tariff rates, exerting significant downward pressures on prices during this period. There were also further cuts in tariffs before and during the second deflationary episode. In addition, during these periods productivity growth remained high—in particular during the second deflationary episode—and commodity prices declined significantly, further increasing deflationary pressures. Demand factors also contributed, as the output gap remained negative, albeit small, during both deflationary periods.
First-Round Impact of the Regression Variables on Inflation1
Source: IMF staff calculations.1 The figures show the initial impact of the variables on inflation; the second-round impact of these variables on inflation through lagged inflation is not included. The shaded areas indicate deflationary periods.2 Excludes the impact of the lagged inflation variable.First-Round Impact of the Regression Variables on Inflation1
Source: IMF staff calculations.1 The figures show the initial impact of the variables on inflation; the second-round impact of these variables on inflation through lagged inflation is not included. The shaded areas indicate deflationary periods.2 Excludes the impact of the lagged inflation variable.First-Round Impact of the Regression Variables on Inflation1
Source: IMF staff calculations.1 The figures show the initial impact of the variables on inflation; the second-round impact of these variables on inflation through lagged inflation is not included. The shaded areas indicate deflationary periods.2 Excludes the impact of the lagged inflation variable.Several of these factors were reversed in 2002–03, ending the second deflationary episode. Commodity prices in particular changed sharply, increasing around 20 percent during this period. In addition, demand factors picked up, especially in 2003 (except during the second quarter that reflected the SARS epidemic): output growth was significantly above trend, the fiscal deficit remained close to 3 percent of GDP, and money growth had picked up. Combined, these inflationary pressures were stronger than the downward price pressures from continued productivity growth, and led to a sharp increase of inflation by the end of 2003.
Appendix: Data Sources and Calculations
All data are year-on-year growth rates of quarterly data from the CEIC database, unless otherwise noted.
| inf: | the CPI inflation rate. |
| prod: | aggregated differences of the output growth rates and the employment growth rates in the primary, secondary, and tertiary sectors. Quarterly employment figures are interpolated from annual sectoral employment data provided by the National Bureau of Statistics of China. |
| ygap: | the deviation of quarterly output growth rates from the trend, which is obtained using the Hodrick-Prescott filter with smoothing parameter equal to 1600. |
| g: | the ratio of the quarterly fiscal balance of the general government to quarterly GDP; seasonally adjusted using X-12. |
| M2: | growth of broad money. |
| Comm: | growth rate of the world commodity price index (source: IMF, Research Department). |
| Tariff: | unweighted average tariff rates (source: Rumbaugh and Blancher, 2004). |
| Neer: | trade-weighted index of the nominal exchange rates; bilateral exchange rates are provided by the IMF. |
| inf: | the CPI inflation rate. |
| prod: | aggregated differences of the output growth rates and the employment growth rates in the primary, secondary, and tertiary sectors. Quarterly employment figures are interpolated from annual sectoral employment data provided by the National Bureau of Statistics of China. |
| ygap: | the deviation of quarterly output growth rates from the trend, which is obtained using the Hodrick-Prescott filter with smoothing parameter equal to 1600. |
| g: | the ratio of the quarterly fiscal balance of the general government to quarterly GDP; seasonally adjusted using X-12. |
| M2: | growth of broad money. |
| Comm: | growth rate of the world commodity price index (source: IMF, Research Department). |
| Tariff: | unweighted average tariff rates (source: Rumbaugh and Blancher, 2004). |
| Neer: | trade-weighted index of the nominal exchange rates; bilateral exchange rates are provided by the IMF. |
All growth rates are year-on-year changes. Seasonally adjusted monthly changes could indicate earlier turning points; however, data limitations—in particular, breaks in the series—do not allow proper seasonal adjustment.
The weights are the estimated coefficients from a least squares regression, where the dependent variable is CPI growth and the explanatory variables are the growth rates of its components. Official data on the weights of the components of the CPI basket are not publicly available.
The index used in this analysis is the overall index aggregated using global weights.
The sample is truncated at 1996 because prior data cover China’s high-inflation period, for which only partial data are available on a quarterly basis. Moreover, only partially appending this high-inflation period to the low-inflation period could potentially distort the estimates for the latter period, which is of more interest.
A11 growth rates are year-on-year. See the appendix for a detailed description of the data sources and calculations.
The use of domestic credit growth as an alternative variable to capture the impact of monetary developments on inflation did not fundamentally change the results. The correlation between M2 growth and credit growth over the period is about 50 percent.
The tests included recursive coefficient estimates and CUSUM of squares test.
Augmented Dickey-Fuller and Phillips-Perron tests rejected the null hypothesis of a unit root in the inflation rate. The large coefficient on the lagged dependent variable probably partly explains the high R-squared statistic.
