«Exporting Hyperinflation: The Long Arm of Chiang Kai-shek Richard C. K. Burdekin and Hsin-hui I. H. Whited* Claremont McKenna College and University ...»
The Table 2 results clearly support an important role for capital inflows in the Taiwanese inflation process. An important motivation behind these capital flows was undoubtedly flight from the political and military risks on the mainland to the relative safety of Taiwan. But these capital inflows were also likely responsive to exchange rate overvaluation associated with any worsening of inflationary pressures in mainland China over and above the levels prevailing in Taiwan. Accordingly, we next examine whether the volume of inflows can be explained by the differential between mainland Chinese and Taiwanese money growth and inflation rates. We would expect any such relationship to be strongest over the post-July 1948 period – during which the rate of exchange between the taipi and the new gold yuan was fixed notwithstanding a tremendous surge of inflationary pressure on the mainland. In Table 3, we examine the effects on capital inflows of money growth and inflation differentials between mainland China and Taiwan over the full sample and over the pre-August 1948 sub-sample. The contemporaneous inflation differential plus two lagged values are selected by the AIC criterion together with the third lag of the money growth differential. The effect of the inflation differential is actually ambiguous given the mixed significant negative and positive coefficients and a Wald test reveals only an overall 47% significance level.
An important qualification to all the above results is the fact that even our “full sample” comprises little more than two year’s worth of data. Nevertheless, the findings do most consistently point to the importance of mainland Chinese monetary variables and capital inflows as determinants of Taiwanese money growth and inflation during the “twilight period” of Nationalist rule on the mainland. The suggested endogeneity of the Taiwanese variables to the mainland Chinese “forcing variables” receives further confirmation if, following Lin and Wu (1989), we apply Granger-causality tests to our set of five variables (allowing for two lags on each).15 The causality test results are shown in Table 4 and are in line with the inferences drawn from Table 2 above. First, capital inflows Granger-cause both Taiwanese inflation and Taiwanese money growth. Second, Chinese money growth and Chinese inflation also Grangercause Taiwanese inflation as suggested in Table 2. Third, Taiwanese money growth does not Granger-cause Taiwanese inflation. The only differences from the implications derived from Table 2 are that there is no confirmed causal effect of Chinese money growth on Taiwanese money growth in the Granger tests and that there is a suggested causal effect of Taiwanese inflation on Taiwanese money growth that was not indicated in the prior regression results.
The implication that Taiwanese inflation Granger-causes Taiwanese money growth but that Taiwanese money growth does not Granger-cause Taiwanese inflation matches the conclusions drawn by Makinen and Woodward (1989) and Quddus, Liu and Butler (1989) based on their bivariate causality testing with these two variables.16 This is also consistent with the potential feedback from inflation to future rates of money growth in high inflation episodes emphasized by Sargent and Wallace (1973). Surprisingly Lin and Wu (1989) reach opposite conclusions even though their approach is seemingly more similar to ours than to Makinen and Woodward (1989) and Quddus, Liu and Butler (1989) -- neither of whom control for the role played by the mainland Chinese variables in their causality testing. Lin and Wu’s other finding that Chinese inflation Granger-causes Taiwanese inflation remains in line with our results.
Estimation of impulse response functions confirms that the suggested causal effects of the Chinese variables on Taiwanese inflation and the effects of capital inflows are economically significant in terms of their magnitudes. Figures 3 and 4 show that one standard deviation shocks to Chinese money growth and Chinese inflation produce an up to 10% shift in Taiwanese inflation across the different lag orders. Figures 5 and 6 suggest similar-size shifts in Taiwanese money growth and Taiwanese inflation following a one standard deviation shock to the level of capital inflows. Each figure shows the impulse response with one-standard error bands that are calculated using the Monte Carlo method with 100 repetitions. These results are for an ordering with capital inflows first followed by Chinese money growth and Chinese inflation and then Taiwanese money growth and Taiwanese inflation. But a similar pattern emerges if we experiment with other orderings of the variables. The effects of Taiwanese inflation on Taiwanese money growth were less clear in the impulse response analysis, however. While a significant impact was evident with some variable orderings the magnitude was always smaller, as well as less consistent, than the impulse responses to Chinese money growth, Chinese inflation and capital inflows.
3. CONCLUDING COMMENTSThe importance of mainland China’s inflationary process to Taiwan’s own hyperinflation has received surprisingly little recognition in the academic literature. Yet this episode seems to be a textbook case of the exporting of inflation from the fading Nationalist regime on the mainland to their new power base on Taiwan. This process was facilitated by the fixed, and overvalued, rate of exchange for the new gold yuan against the Taiwanese taipi that was adopted in August 1948. The inflationary spiral in Taiwan continued into 1950, when it was finally stopped with the help of a deposit program designed to draw currency out of circulation and with the arrival of US aid (see, for example, the discussion and references in Burdekin and Whited, 2001). Meanwhile, the last Nationalist forces outside Taiwan were evacuated from the Choushan Islands on May 17, 1950.
3. The inflationary consequences were, however, mitigated by central banks’ ability to sterilize the effects of the international reserve flows on the money stock (Laney and
4. Meanwhile, Lin and Wu (1989), Makinen and Woodward (1989), and Quddus, Liu and Butler (1989) all test causality between money growth and inflation in Taiwan but Lin and Wu’s conclusions are opposite to the other two studies.
3. Regarding the importance of this reform to the inflation process on the mainland, Quddus, Liu and Butler (1989) find that truncating the sample period at August 1948 eliminates the bi-directional feedback between inflation and money growth in mainland China that applies if the sample is extended into 1949 (see also Tang and Hu, 1983).
Meanwhile, our own prior univariate analysis of inflation and the growth rates of nominal and real money balances in Taiwan over the 1945-1953 period (Burdekin and Whited,
2001) suggests that the August 1948 reform coincides with the beginnings of an extended period of instability in Taiwanese currency and inflation processes that continues into
4. The area of circulation of the fapi also was reduced as local Communist currency was issued in its stead. Hsia (1953, pp. 73-74) details ten different currencies circulating in the Communist-controlled areas in January 1948.
5. The total capital flight from China to Hong Kong between 1947 and 1949 has been estimated at around HK$500 million (Chang, 1958, p. 320).
6. Chang (1958, p. 80) incorrectly puts this maximum at 200 million. Note that the 374 thousand billion fapi in circulation in July 1948 were already equivalent to 125 million gold yuan. US Department of State (1949, p. 200) confirms the 2 billion limit and refers to the 200 million figure as being the value of the gold, silver and foreign exchange that was available to back the currency.
7. On July 3, 1949, the director of the People's Bank of China at Shanghai stated that the total issue of gold yuan notes reached 60,000 billion on May 25, 1949 (equivalent to 180,000 quadrillion fapi, or 11.6 times larger than Table 1's April 1949 figure given by the Nationalist central bank).
8. There was a subsequent ten-fold increase in Shanghai wholesale prices after the Communist takeover before stabilization was finally achieved in March 1950 (see Burdekin and Wang, 1999). While Nationalist forces remained on the field in mainland China until 1950, Chou (1963, p. 27) states that after the fall of Shanghai “the gold yuan notes were rejected by the public.” Attempts to replace the gold yuan notes with a new “silver yuan” standard on July 1, 1949 were an utter failure. Reportedly, the budget deficit in the second half of 1949 totaled 88% of expenditures--a deficit that "had to be met with gold, silver, and foreign exchange since bank notes no longer enjoyed the people's acceptance" (Chang, 1958, p. 169).
9. As with Makinen and Woodward (1989), Quddus, Liu and Butler (1989) and Burdekin and Whited (2001), our currency series does not include the emergency money that was issued in 1948 and 1949. As the reported emergency money is only 2% of other currency in June 1949 and less than 1% in December 1948 (Chen, 1955, p. 87), it is unlikely that excluding this emergency money has any major effect on the results.
10. Besides capital inflows, the addition of contemporaneous values of the other variables was always rejected on the basis of an F-test.
11. Additional lagged values of the variables included in the equations were used to form the instruments in the 2SLS estimation. In the inflation equation the instrument were lags two through four of Chinese inflation, Chinese money growth and capital inflows. In the money growth equation the instruments were lags three through five of Taiwanese money growth, Chinese money growth and capital inflows.
12. While it is far from ideal to apply GMM to such a small sample as this, the results for the 1947:6-1949:4 period actually seem to be quite robust to the estimation procedure employed. The application of 2SLS, for example, has no effect on the signs and significance levels of the coefficients reported in Table 2. The results for the 1947:6sub-sample are not as consistent, however, and the right-hand-side variables there are generally not statistically significant under 2SLS estimation – the only exception being that the lagged capital inflow remains significant at approximately the 7% level.
13. Simply adding a dummy variable defined as one for the post-July 1948 period and zero otherwise did not capture the source of the instability and, when such a dummy was added to either the inflation, money growth, or capital inflow equation, it was always insignificant. This suggests that any instability is reflected more in the slope coefficients but, at the same time, there were insufficient degrees of freedom to add further dummies.
We also could not sensibly estimate over the post-1948 period alone and, as it is, the preAugust 1948 estimates cannot be considered anything more than suggestive given that they use barely more than a year’s worth of data.
14. Application of a Wald test showed the overall positive effect to be significant only at approximately the 13% level. There are also partially offsetting effects for the capital inflows variable but neither of the negative coefficients on the lagged terms is significant at even the 20% level. Re-estimating the equation over the full-sample with the statistically insignificant lagged Taiwanese money growth and lagged capital inflow
deleted actually improves the overall goodness of fit and, for comparison, the results are:
Taiwanese money growth = 13.812*** - 0.096** Chinese Money Growth (-1)
Adjusted R2 = 0.88; F-Statistic = 49.71***; Chow Breakpoint Test = 2.4*