Using Stock Returns to Predict GDP Growth

Summary

• A new NBER working paper looks at variations in firm-level stock returns to predict next-quarter real GDP growth.
• It finds that while greater macro uncertainty predicts lower growth, more volatility in firm-level stock prices leads to higher next-quarter GDP growth.
• Their measure of idiosyncratic stock return variation is better at predicting short-term growth than other traditional measures.

Introduction

At the heart of Economic Growth Theory is technological progress. A country whose productivity is advancing faster should gain a higher growth trajectory, all else equal. In DSGE models, often employed by monetary and fiscal authorities, this phenomenon is captured within a firm’s production function. In other words, new technologies are reflected in the capabilities of firms which, once aggregated, lead to changes in economic growth.

A new NBER working paper exploits the forward-looking nature of financial markets to determine whether differences in the stock returns of firms can predict next-quarter economic growth. They can.

And the results are very intuitive. Over time, differences in firm values should reflect differences in their relative technological progress or productivity. Larger variations in productivity within a quarter, just like greater creative destruction, should lead to higher subsequent growth, which the authors find.

Idiosyncratic Stock Return Variation

Understanding the intuition of this paper is the key to understanding the results. So I focus on that first.

Conceptually, movements in daily stock prices reflect either (i) market-wide information, (ii) firm-specific fundamentals, or (iii) noise. Market-wide factors, such as geopolitical risks, typically explain close to 10% of daily equity market price action. These macro factors often increase during recessions and, as is currently the case, during globally significant events.

However, outside of market meltdowns, firm fundamentals are king. In a recent paper we reviewed, the authors found that firm-specific information accounts for more than half of daily stock market movements. Furthermore, their contribution tends to be a pro-cyclical, i.e., idiosyncratic variations in firm fundamentals explain relatively more of daily stock price movements during expansions.

Against this backdrop then, netting-out market-wide factors and noise allows us to determine what proportion of a firm’s daily return can be explained by its fundamentals. Why may this vary from firm to firm or over time? According to the authors, if we return to Economic Growth Theory, a key reason should be productivity. Firms deploying new technologies should become more valuable relative to their peers.

And so, because markets are forward-looking and stock prices capitalise new information about individual firms’ altered prospects, productivity improvements are likely to show up first in stock prices. Therefore, greater variations in the stock prices of firms within a quarter will likely reflect greater differences in productivity. These productivity ‘shocks’ reflect what the authors call ‘elevated idiosyncrasy in stock returns’.

What is the paper’s hypothesis? Well, if greater productivity leads to higher growth, then greater variation in idiosyncratic stock returns this quarter should lead to higher growth in the subsequent quarter. It does, hence the paper’s title: ‘Idiosyncrasy as a Leading Indicator’.

Data and Methodology

The authors collect stock price data for all NYSE, Nasdaq, and Amex-listed common stocks from 1926 to 2021. To calculate the idiosyncratic variation in stock returns, they regress daily firm returns on market returns. Anything unexplained by the market will be left in the error term, or residual. So, the residual from these regressions is firm-specific information. They sum this information for each firm for all days in the quarter and square it to get a variance. They then aggregate across all firms in the quarter using a value-weighting method. What is left is their aggregate idiosyncratic variation (AIV) measure.

They then do the same for the market-wide information. But rather than use the residuals from the regression, they use the coefficient on market returns. In other words, they use the information in daily firm returns that is explained by market-wide returns. Again, they aggregate across all firms in the quarter and call this the aggregate systematic variation (ASV).

Armed with their AIV and ASV variables, they then want to see if these measures can predict GDP growth in the following quarter. To ensure no other variables are driving that prediction, they need to control for other common predictors of GDP growth. These are:

• Credit spreads.
• Term spreads (10y minus 3m).
• Bond yields (change in 10y).
• Dividend yield.
• Excess market return.
• Inflation.
• Stock market illiquidity.

The authors then regress AIV, ASV and the above measures on GDP growth to see which measures are the best predictors.

Does AIV Work?

AIV captures the variation in stock returns driven by differences in firm-specific information. This, the authors argue, captures the market pricing in ‘disequilibrating shocks’, which most often reflect technological progress and creative destruction. Indeed, the authors find that two of the three prominent spikes in their AIV measure occur during unusually large and sustained economic booms driven by tech – 1920s and 1990s (Chart 1).

Next, the ratio of AIV to ASV, which measures the relative idiosyncratic variation, i.e., the proportion stock returns are moving due to firm-specific changes versus market-wide changes, rises during expansions and drops preceding recessions (Chart 2). This reaffirms that macro factors become more important around recessions. It also shows that their measures are correlated to next-quarter macroeconomic growth.

Results

The authors find that a one-standard deviation increase in current quarter AIV predicts a 0.18pp faster real GDP growth the next quarter. The same holds true for ASV, but with the opposite sign. In other words, an increase in idiosyncratic stock return variation predicts an increase in GDP growth. Meanwhile, when macro conditions become a more important driver of stock market volatility, i.e., there is greater macro-level uncertainty, real GDP growth is predicted to be weaker in the subsequent quarter.

Furthermore, it turns out that these measures are better at predicting next-quarter growth than all the other commonly used predictors of economic growth, except changes in credit spreads. For credit spreads, the tighter they are, the healthier real GDP growth is next quarter.

Why Does Idiosyncrasy Lead to Higher Growth?

First, greater volatility in daily stock returns, which is ultimately what this measure is identifying, could reflect greater uncertainty. This may induce greater government spending that in turn increases growth. However, the authors find no evidence of this.

Second, firm-level volatility is intricately linked to household-level risk and wealth. That is, increased dispersion in households’ human and portfolio capital leads to increased aggregate consumption because of greater firm-level volatility. Indeed, the authors find that higher idiosyncratic stock return variance forecasts higher next-quarter aggregate consumption.

Third, the authors argued that this disequilibrating shock is akin to a technological shock. If true, some firms will benefit from new technologies, while others will not. Indeed, the authors find that higher idiosyncratic stock return variation forecasts higher next-quarter growth in patent applications, patent citations, total factor productivity, and labour productivity.

Bottom Line

The paper discovers that movements in stock market prices can predict future economic growth. The key to this finding comes from decomposing the variation in returns into macro-driven volatility and firm-level volatility. While macro uncertainty hurts future growth, greater variation in firm-level prices supports next-quarter growth. The key reason is this variation manifests itself as an improvement in aggregate productivity.

Is this increase in investment in innovation driven by firms having to become more efficient in more volatile markets? Or does it reflect investors pricing in new information about individual firms’ altered prospects because of planned innovation ahead of time? I am unsure. Regardless, the paper’s AIV measure predicts next-quarter GDP growth and productivity well.