«Exporting Hyperinflation: The Long Arm of Chiang Kai-shek Richard C. K. Burdekin and Hsin-hui I. H. Whited* Claremont McKenna College and University ...»
With the official exchange rate between the gold yuan and the taipi held unchanged in spite of the growing loss of confidence in the gold yuan, the attraction of exchanging gold yuan for taipi was obvious. The accelerating net capital inflows into Taiwan after the August 1948 reform are depicted in Figure 2. Perhaps reflecting the fact that the “failure of the monetary reform seemed to be anticipated by the public” (Li and Wu, 1997, p. 7), the net capital inflow between August and October 1948 was equivalent to 95,981 million taipi. This capital inflow was the major force causing the money supply to double over this same period, accounting for 86.8% of the M1 money supply increase in September 1948 and 97.9% of the October increase (Lin and Wu, 1989, pp. 935-36). The Bank of Taiwan stopped accepting deposits in gold yuan notes from the mainland on October 23, 1948 and, on November 1, 1948, the Nationalist government belatedly adjusted the exchange rate between the taipi and the gold yuan. The official rate was re-set at 1:1000 -- that is, one taipi was equal to 1000 gold yuan notes. Also, the exchange rate was now allowed to change more often in response to market forces. The exchange rate adjustment was accompanied by a capital outflow that grew 8.8-fold between October and November 1948 and created the first net capital outflow since the August 1948 reform was enacted.
Taiwan’s respite at the end of 1948 was short lived, however. Capital flight to Taiwan reaccelerated in the face of the Nationalists’ crumbling position on the mainland and the Nationalist government itself began to transfer its remaining resources to Taiwan. In the face of the mounting deficits on the mainland, the 2 billion yuan maximum for the gold yuan notes was formally set aside on November 11, 1948.6 Meanwhile the price controls in Shanghai unraveled during the middle of October 1948 in the midst of panic buying when retailers “withheld their goods from sale rather than sell them at official prices, and even the restaurants refused to do business” (Chang, 1958, p. 80). Prices were formally freed on November 1, 1948 under a Financial Emergency Executive Order. While nearly 3.4 billion gold yuan (equivalent to more than 10,182 thousand billion fapi) were in circulation on November 30, 1948, by the end of April 1949, total note issue had risen to more than 5 thousand billion gold yuan (equivalent to a staggering 15,483,720 thousand billion fapi--see Table 1).7 Shanghai wholesale prices rose 59,374 times between September 1948 and April 1949 before rising a further 84 times between April and May 1949 on the eve of the fall of that city to the Communists on May 27, 1949.8 Taiwan’s November 1948 net capital outflow of -23,277 million taipi reversed to a capital inflow of 214,495 million taipi in December 1948. December’s capital inflow amounted to 73.4% of the total increase in currency in that month (Li and Wu, 1997, p. 8). The Taiwanese money supply grew 1.75-fold between November and December 1948. Pressures on the money supply increased when many divisions of the Nationalist army began moving to Taiwan in 1949.
The Bank of Taiwan now not only had to support military expenses in mainland China but also needed to make massive loans to support the newly-arriving Nationalist troops. The growth rate of the money supply increased to 41.6% in January 1949, 26.1% in February, 18.8% in March, 42.8% in April, and 118.7% in May. Inflation peaked during this period, reaching 45.3% in January, 56.8% in February, 35.7% in March, 46.5% in April, and 104.8% in May. This still paled in comparison to the depreciation of the gold yuan, however. By late April and early May 1949, the gold yuan's initial official exchange ratio of 4 gold yuan to the US dollar had fallen to the point that open market quotations ranged from 5 million to 10 million gold yuan per US dollar (US Department of State, 1949, p. 401).
2. EMPIRICAL EVIDENCEThe underlying data on the Taiwanese money supply and the Taipei wholesale price index are drawn from Liu (1970). Given that, during this period, currency was apparently “the only means of payment in retail business in Taiwan” (Liu, 1970, p. 24n), we attach more significance to the net currency measure of the money supply than to the broader M1 measure.9 The data on capital inflows into Taiwan from mainland China are given by Lin and Wu (1989, p. 934). The mainland Chinese money supply and inflation series are based on the currency and Shanghai wholesale price data provided by Quddus, Liu and Butler (1989). None of the series on inflation, money growth and capital inflows is subject to unit roots. The presence of unit roots is rejected at the 99% confidence level when we allow for up to three autoregressive factors in an application of the Augmented Dickey-Fuller test.
Taiwanese inflation and money growth are regressed on their own past values, past values of mainland Chinese money growth and inflation, and the level of capital inflows. This allows us to assess the relative importance of domestic and external monetary conditions to the hyperinflation process in Taiwan. We also examine the determination of the capital inflows themselves and regress the inflows on the difference between mainland Chinese and Taiwanese money growth and inflation rates. Increases in these inflation and money growth differentials could be expected to fuel capital outflows from mainland China to Taiwan through increasing the degree of overvaluation of the gold yuan exchange rate. According to the literature on exchange market pressure (see, for example, Girton and Roper, 1977; Burdekin and Burkett, 1990), excess money growth and excess inflation in mainland China vis-à-vis Taiwan should be reflected in some combination of currency depreciation and international reserve losses. If the exchange rate is held constant, loss of international reserves, or capital outflow from mainland China, becomes the only outlet for the exchange market pressure. A number of Latin American countries experienced this phenomenon in the run-up to the debt crisis of the 1980s, where “the general tendency toward overvaluation fueled massive capital flight... In country after country, the public speculated against the central bank by acquiring foreign exchange and moving it abroad” (Edwards, 1995, p. 23).
The exact lag order of the variables in the inflation, money growth and capital inflow regressions is selected – from a maximum of four lags – according to the Akaike information criterion (AIC). We also allow for the presence of contemporaneous values of the right-handside variables. All estimation is performed over monthly data from 1947 through April 1949 (the last month prior to the fall of Shanghai). We test for shifts in the relationships at the time of the August 19, 1948 reform in mainland China and, given evidence that such a shift occurred, reestimate the equations with the sample truncated at July 1948.
The results for the Taiwanese inflation and money growth are presented in Table 2.
Lagged Chinese money growth, and contemporaneous and lagged values of the capital inflows from mainland China are significant in each equation.10 Lagged Chinese inflation is also significant in the inflation equation. Lagged Taiwanese money growth is selected by the AIC criterion for the money growth equation but is not significant at the 10% level or better. Both equations were initially estimated using two-stage least squares (2SLS) – incorporating an autoregressive term – and then subjected to a battery of specification tests.11 In each case, Qtests and LM tests for serial correlation, the White test for heteroskedasticity, and Ramsey’s RESET test indicated that the null hypothesis of no specification error could not be rejected. The Jarque-Bera normality test suggested, however, that the residuals from the 2SLS estimates of the inflation equation were significantly non-normal. Accordingly, the inflation equation was reestimated using the generalized method of moments (GMM) – which does not require normality – to produce the results reported in the table.12 The full sample results are followed by results for the 1947:6-1948:7 sub-sample that excludes the effects of mainland China’s August 1948 monetary reform. The Chow breakpoint test for the money growth equation supports the existence of a structural break at August 1948.
No such Chow test can be applied under the GMM used for the inflation equation but this equation, too, was re-estimated over the sub-sample and the findings compared to the full-sample results.13 The inflation equation reveals the expected positive effects of higher Chinese money growth and rising capital inflows (the positive coefficient on lagged inflows being ten times larger than the negative coefficient on the contemporaneous value). There is, however, a counterintuitive negative sign on lagged Chinese inflation. While lagged Chinese inflation has the expected positive, and significant effect, in the pre-August 1948 sub-sample, these latter results must be qualified by the very short sample period employed. Nevertheless, the inflation equation results do seem to imply a dominant role for the mainland China variables and the inflows from mainland China. Indeed, neither Taiwanese money growth nor lagged Taiwanese inflation was even selected by the AIC criterion when entered alongside these other variables.
The above findings certainly do not mean that Taiwanese money growth was irrelevant to Taiwan’s inflation over this period. Rather, these estimation results imply that Taiwan’s money supply process may have been endogenously driven by the mainland Chinese variables and capital inflows. The full-sample results for the money growth equation are not inconsistent with this view. Lagged Chinese money growth and capital inflows are the only significant variables in the equation. The effects of Chinese money growth and capital inflows are positive overall although, in the case of Chinese money growth, a significant negative effect on the first lag partially offsets the larger significant coefficient on the second lag.14 Unlike the inflation equation, lagged Chinese inflation is not selected by the AIC criterion. Re-estimation over the pre-August 1948 sub-sample shows a stronger positive effect of capital inflows but no significant role for lagged Chinese money growth. For the sub-sample, not only the contemporaneous capital inflow but also the two lags are now statistically significant. Although the first lag (only) has a negative sign, a Wald test confirms an overall positive effect that is significant at better than the 1% level. Finally, the first lag of Taiwanese money growth becomes significant at least at the 11% level over the pre-August 1948 period.