Monday, September 27, 2004

Does sentiment matter?

Does sentiment matter? By Anchada Charoenrook

Super Short version: Yes!

Slightly longer version:

Sentiment, as measured by the University of Michigan Consumer Sentiment Index, does affect stock prices. Charoenrook finds that “changes in consumer sentiment reliably predict excess stock market returns at one-month and one year horizons.

Long version:

This paper tries to settle the debate that exists in finance as to whether sentiment plays a role in asset pricing. This is interesting question for, as the paper states, in a purely rational market, sentiment would play no role.

Alternatively, sentiment plays a role in many behavioral finance markets: “Delong, Shleifer, Summers, and Waldman (1990) propose a model of asset pricing based on the idea that irrational investors guided by sentiment misprice stocks, and the unpredictability of investor sentiment impounds resale risk on assets that they trade. In other behavior-based asset-pricing models, investor sentiment or belief distorted by psychological attributes drives stock prices away from their fundamental valuations.”

Past empirical evidence does not provide a clear answer as to whether sentiment matters or even how to most efficiently measure sentiment. For instance, “in the closed-end fund literature, some researchers argue that small investor sentiment can be measured by change in the discount on closed-end fund equity returns.” Consequently, many finance papers have used closed end fund discounts as a proxy for market sentiment.

This proxy has occasionally led to conflicting conclusions. “Lee, Shleifer, and Thaler (1991) report empirical evidence that the discount on closed-end fund return is a factor in the stock return-generating process.” While on the other hand “Elton, Gruber, and Busse (1998) find that the discount on closed-end fund return is not priced and hence is unimportant in the return-generating process.”
In this current paper, Charenrook around the improper proxy problem relating changes in the widely reported University of Michigan Consumer Sentiment Index to changes in stock market returns.

She finds “that change in the consumer sentiment index is negatively related to future value-weighted and equal-weighted excess aggregate stock market returns at one-month and one-year horizons.” That is if investors are happy (higher sentiment) the returns one month and one year out, tend to be lower.

This relationship “remains a strong and consistent predictor of returns after controlling for other established predictors. [Such as] dividend yield, the book-to-market ratio of the Dow Jones Industrial Average (DJIA), the slope of the term structure, the yield spread between Baa and Aaa bonds, the short rate yield, lagged excess market returns, and the consumption-wealth ratio.”

The author disputes suggestions that such a relationship is due to ties to the business cycle: “Empirical test results…show that the predictability of change in consumer sentiment is unrelated to economic cycles measured by real gross domestic product growth or consumption growth. Moreover, change in consumer sentiment has incremental predictive power for aggregate stock return after controlling for lagged consumption-wealth ratio, which is a strong predictor of business cycles (Lettau and Ludvigson, 2001).”

It is important to note that this relationship is not just statistically significant but economically significant as well: “in the one-year returns sample, a one-standard deviation improvement in consumer sentiment predicts a 6 percentage points a year lower excess return relative to the unconditional mean. Moreover, change in consumer sentiment index performs better than the benchmark ARI model in out-of-sample forecasting.”

In conclusion the author identifies the paper’s main contributions: “First it uses a direct survey of sentiment instead of proxies such as closed-end fund discounts….Second, this study contributes to the debate on whether sentiment can cause systematic mispricing in the aggregate stock market….The results suggest that it is premature to reject a behavioral explanation.”

Very interesting and well done.

BTW the discussion of how the Sentiment Index is calculated in well worth your time!

1 comment:

Anonymous said...

The link to the paper does not work. Can you fix this?