130 Years of Turbulence, and No Good Explanation: Schwert on Why Volatility Changes

By the late 1980s, the profession had got very good at describing how volatility behaves. ARCH and GARCH could tell you that volatility clusters, that it is persistent, that a turbulent week predicts a turbulent month.

What nobody could tell you was why. Why is the market violently unstable in some decades and placid in others? Volatility is not a mood, it is presumably driven by something in the real world. What?

G. William Schwert asked that question with unusual seriousness and unusual data. He assembled monthly US stock market data going back to 1857, well over a century of it, and went hunting for the economic forces that make markets turbulent.

The problem: the theory says volatility should have a cause

A stock is a claim on a company's future cash flows. Its price should move when news arrives about those cash flows or about the rate at which they are discounted. So volatility, the amount that prices move, ought to reflect how much genuine uncertainty there is about the economy.

This gives you an obvious list of suspects, and Schwert investigated all of them:

• Macroeconomic volatility. When inflation, industrial production and money growth are themselves volatile and unpredictable, the future is murkier, so stock prices should move around more.
• Recessions. Economic downturns are periods of genuine uncertainty. Volatility should be higher when the economy is contracting.
• Financial leverage. This one has real theoretical bite. When a company's stock price falls, its debt stays the same, so its debt-to-equity ratio rises, and the equity becomes a more leveraged, riskier claim. Mechanically, falling prices should make the stock more volatile. This is the original meaning of the "leverage effect."
• Trading activity. More volume, more information flowing, more price movement.

Each of these is plausible. Schwert measured them all against 130 years of data.

The key idea via analogy: the detective who eliminates every suspect

The paper reads like a detective story where the detective interviews every suspect and finds that none of them could have done it. Not quite none, but the confessions are all disappointingly thin.

Macroeconomic volatility: there is a relationship, and it runs the way theory predicts. But it is weak. The wild swings in stock market volatility are far larger than the swings in the volatility of the underlying economic fundamentals. The stock market is much more turbulent than the economy it is supposed to be a claim on.

Recessions: here Schwert finds something real. Volatility is genuinely higher during economic contractions. This is one of the more durable findings in the paper and it holds up in later work. But it is a partial answer, and it does not explain the magnitude of the swings.

Leverage: the mechanism is real and it is statistically detectable. But when Schwert measures how much of the movement in volatility it can actually account for, the answer is a relatively small part. The leverage story is a genuine channel and a badly insufficient one.

Trading activity: correlated with volatility, certainly. But this is uncomfortably close to circular. Volume and volatility move together, but which causes which, and is either really an explanation? Both are probably responding to the same underlying flow of information.

And then there is the Great Depression, which sits in the data like a bomb crater. Stock market volatility from 1929 to 1939 was extraordinarily high, far higher than anything the ordinary economic variables can begin to justify. Schwert is blunt about this: the amplitude of these fluctuations is very hard to explain with simple models of stock valuation.

What it means

The conclusion the paper drives you toward is uncomfortable, and it is why the paper still matters.

Stock market volatility is much larger, and much more variable, than any straightforward account of economic fundamentals can explain. The market is not simply a mirror held up to the economy. Something else is going on: shifting risk appetite, waves of fear, feedback between prices and beliefs, the sheer social dynamics of markets. Whatever it is, it is not in the national accounts.

This connects directly to Robert Shiller's argument from the same era that stock prices move far more than the dividends they are claims on. Schwert's finding is the volatility-side version of the same unsettling observation.

Why it mattered

• It is the definitive empirical catalogue. If you want to know what volatility did, and what it correlated with, over more than a century, this is the source. The data assembly alone was a substantial contribution.
• It established the business-cycle link. Volatility rises in recessions. That fact is now standard and it shows up in modern models that link volatility to the macroeconomy.
• It deflated the leverage explanation. The leverage effect had been the leading structural story for why volatility rises when prices fall. Schwert showed it is nowhere near big enough to carry the load. Later work confirmed that most of the asymmetry must come from somewhere else, probably from changing risk premia and investor fear rather than from balance sheet mechanics.
• It framed the puzzle that motivates everything after. Every subsequent attempt to explain volatility, behavioural, structural, macro-finance, is trying to fill the hole this paper dug.

The honest limitations

• Monthly data over 130 years is coarse. It cannot see intraday dynamics, and the early data is of uncertain quality. Building a consistent stock index back to 1857 involves a lot of judgement.
• Correlation is not mechanism. The paper documents relationships. Establishing what causes what, when everything is jointly determined and endogenous, is far beyond what regressions on macro data can settle.
• The economy changed enormously over the sample. The market of 1860 and the market of 1987 differ in almost every respect: what is listed, who trades, how information travels, what regulation exists. Treating them as one homogeneous sample is a stretch, even if a necessary one.
• The failure to explain is not itself an explanation. The paper is much better at ruling suspects out than at naming the culprit. It leaves you knowing that fundamentals cannot do the job, without a satisfying alternative.
• Macro variables are measured badly. Historical inflation and industrial production series are estimates, and their own volatility is measured with substantial error, which biases the tests toward finding weak relationships.

The one-line takeaway

Schwert took 130 years of stock market data and asked what makes markets turbulent, and the enduring result is a negative one: recessions matter, leverage matters a little, and none of the obvious economic explanations come remotely close to accounting for how violently the stock market actually moves.