|
Article Excerpt
Abstract
In an influential article, [Romer, Christina "The Great Crash and the Onset of the Great Depression," Quarterly Journal of Economics, 105, 1990, pp. 597-624.] estimates the magnitudes of the uncertainty and wealth effects. She reports that before and after the Great Depression, the uncertainty effect has a large and statistically significant influence on durable good production, while the wealth effect is negative but negligible. When the authors of this paper change the specification of the model with respect to the amount of time necessary for stock returns to translate into changes in consumption, they reach the exact opposite conclusions that Romer does. Specifically, when the authors allow consumers 12 or more months to alter consumption behavior, rather than Romer's three, stock price uncertainty did not significantly affect the durable goods production before, during, or after the Great Depression. The authors also find that stock market returns from the previous year have a positive and statistically significant impact on the durable goods production, indicating the importance of the wealth effect. (JEL D10)
Introduction
Two different theories attempt to explain the relationship between changes in the stock market and real activity (production, consumption, or real investment). The first theory, the wealth effect, claims that there is a positive correlation between current stock returns and future real activity. Fama [1990] suggests a reason why he expects to find this relationship. Stock prices are a leading indicator for the overall economy--so when stock prices rise, investors think that the economy will grow quickly. Friedman's permanent income hypothesis suggests that people smooth out consumption over their lifetime. Therefore, higher expected future wealth will lead to more consumption in the current period, while lower expected future wealth will decrease current consumption. Several studies have found empirical support for the wealth effect, most notably Fama [1981, 1990], Geske and Roll [1983], Huang and Kracaw [1984], Kaul [1987], Barro [1989, 1990], and Schwert [1990].
The other theory, the uncertainty effect [Bernanke 1983a, b], claims stock volatility, not stock returns, affects real activity. The work of Romer [1990], Pindyck [1991], and Bittlingmayer [1998] present statistical evidence to support the uncertainty effect. For instance, Romer [1990] argues that stock price fluctuations cause uncertainty about future income, depressing consumer spending on durable goods. In her model, she included a measure for the wealth effect and one for the uncertainty effect, assuming that it took three months for changes in the stock market to influence the durable goods market. With a 3-month lag, she found that the uncertainty effect does influence durable goods production, while the wealth effect has either no influence or a perverse one. She went on to conclude that the uncertainty effect that she estimated with data through 1928 explained the out-of-sample downturn in 1930.
However, a literature on the wealth effect provides evidence that there is more than a 3-month lag between increases in stock market returns and changes in consumption. Therefore, this study modifies Romer's model by extending the time lag to 12 months, and then re-estimates the model. With this specification, the findings are the exact opposite of what Romer reported. This study finds that the wealth effect influences consumer durables before the Great Crash and after World War II. But there is no evidence of the uncertainty effect during these periods.
Romer's Model and Data
Romer estimates the following model over two samples: 1) 1891-1913 and 1921-1928 (a total of 31 observations); and 2) 1949 to 1986 (a total of 38 observations):
[Y.sub.it] = f ([Y.sub.i(t - 1)], [Y.sub.(t - 1)], [V.sub.t], [W.sub.t]),
where:
1) [Y.sub.it] is the first difference of natural logarithms of commodity output group i from year t - 1 to year t. There is only one commodity group considered: consumer durables. The years t and t - 1 are calendar years.
2) [Y.sub.i(t - 1)] is the one-year lagged value of the dependent variable.
3) [Y.sub.(t-1)] is the first difference of natural logarithms of total commodity output from year t - 2 to year t - 1. The years are calendar years.
4) [V.sub.t] is a measure of the variability of real stock prices. In both sample periods, the difference of the deflated natural logarithms of the monthly stock market index are squared and then averaged over 12 months from October of year t - 1 to September of year t. Romer uses the Cowles Commissions Series P Stock Price Index in the early sample period and the monthly Standard and Poor's 500 (S & P 500) Index in the second sample period.
5) [W.sub.t] is the one-year holding period return of the deflated stock market index calculated from September of year t - 1 to September of year t. Romer uses the Cowles Commissions Series P Stock Price Index in the early sample period and the monthly Standard and Poor's 500 (S & P 500) Index in the second sample period.
This study uses the same data sources that Romer used in her study. In the earlier sample, the output indexes used to create [Y.sub.it] and [Y.sub.(t-1)] come from Shaw [1947], Tables 1-3, pp. 70-77, and are all annual data. In order to discover the proper lag time between the stock market variables and the dependent variable, various specifications of [W.sub.t] and [V.sub.t] are also considered.
In the later sample, the output indexes come from the Federal Reserve Board. This monthly data is averaged into an annual series. In order to create [Y.sub.(t-1)] in the later sample, an index is created by combining total final products and intermediate construction supplies. Weights are used to combine these indexes. The detailed information about constructing this series, as well as the other series, can be found in Appendix.
The monthly S & P 500 Composite Index is from the Security Price Index Record of Standard & Poor's [2002]. The monthly Cowles Stock Index is from Common Stock Indexes [1939]. In the earlier sample, the stock price index is deflated with Warren and Pearson's Wholesale Price Index [1933], pp. 10-13. In the later sample, the stock price series is deflated by the Producer Price Index for All Commodities compiled by Bureau of Labor Statistics, U.S. Department of Labor.
Hypotheses
Romer uses this model to test her uncertainty hypothesis. To paraphrase, the uncertainty hypothesis states that stock market volatility makes people uncertain about their level of future income and therefore, causes them to postpone the purchase of durable goods. As Romer writes: If the consumer buys the durable, then he obviously gets the utility from the durable. However, the consumer is then locked into the durable before the level of future income is learned. Hence, he may choose a quality level that is either too luxurious or too modest relative to his future income, and thus he may be very far from the optimal level of consumption for the life of the durable. On the other hand, if he waits, the consumer is very far from the optimal level of consumption while he is waiting, but he is then able to choose the appropriate durable good once the uncertainty about future income is resolved. Given this trade-off, it is clear that a temporary rise in uncertainty will tend to increase the value of waiting. In such circumstances, consumers may find it advantageous to put off purchasing durables until they are more certain about the course of future income. (p. 602) The model above can test the uncertainty hypothesis. Her theory predicts that the coefficient of [V.sub.t] (stock price variability) will be negative. That is, an increase in stock...
|
