The stock market moves a lot on a day-to-day basis. Today the S&P 500 index may be up 1 per cent, and tomorrow it may be down 1 per cent. And although news commentators may be forced to relate it to some news about the economy, it is rare to come across news that may add or destroy the value of the largest U.S. public companies by as much as 1 per cent.

The asset pricing literature recognises that besides news about the fundamentals, non-informational factors such as investor sentiments can also generate short-term movements in the stock market. But testing this idea is a challenge because of the difficulty of identifying and separating trading behaviour of the informed and noise traders in the aggregate stock market.

My idea is to use a natural experiment in the late 1800s, in which exogenous weather shocks help separate trading behaviours of these two types of traders, to understand the extent to which short-term aggregate stock market fluctuations are driven by uninformed noise trades.

My empirical tests focus on daily stock market performance from 1888 to 1903, the period after the modernisation of the stock market but before the introduction of air-conditioning and communications technologies that enabled investors to minimise the adverse effects of weather shocks on trading. By this period, the New York Stock Exchange (NYSE) had already adopted the modern continuous trading mechanism. However, due to the lack of advanced transportation and communication technologies such as subways and telephones, investors needed to visit their brokerage houses in person to participate in the daily stock market. This meant that those who did not possess great information to trade on were likely to stay out of the market on hotter days, while traders with new information would not risk their informational advantage by delaying their trades. Given this and the fact that the majority of NYSE trades at the time originated from around New York, I use adverse weather conditions, heat waves in the summer in particular, as proxies for the reduction in non-informational trades from the NYSE.

What do I find? First I find that extreme weather shocks such as hotter days in summer, heavy rainfall, and snow tend to reduce the aggregate level of non-informational trades in the market. Various other market signals such as the bid-ask spread further indicate that less informed, sentiment-driven investors are likely to be behind this reduction in the trade volume.

But my main object of interest is the short-term volatility of the aggregate stock market. Indeed, I find that on days with adverse weather shocks, the stock market tends to make more robust movements that do not reverse over subsequent days. In contrast, stock market movements on days when noise traders are more likely to be active tend to generate subsequent reversals as investors realize that the prior movement was due to the buying or selling pressure of noise traders rather than news about the fundamentals.

What can we learn from the paper? Most importantly, the paper’s findings provide evidence for the idea that day-to-day movements in the stock market are largely driven by non-informational reasons such as transitory sentiments that are correlated across investors. This means that investors should not attach too much meaning to daily changes in the stock market. This also means that investors who want to buy or sell a large amount of stocks may want to spread their trades over multiple days so that they do not unluckily buy stocks when there is a strong non-informational buying pressure or sell stocks when there is a strong selling pressure.

The paper also tells us that stock trading in the past was significantly affected by weather conditions, and this adds to existing evidence that adverse weather conditions can affect economic activities. Perhaps a similar effect still exists in parts of the world where temperature control is not an everyday feature.

In this way, a “natural experiment” that happened in Manhattan about 100 years ago offers key insights into today’s market performance, where similar tests cannot be easily performed.



Thummim Cho is an assistant professor of finance at LSE. He has a PhD in economics from Harvard University and a B.A. in Mathematics from Cornell. His main research interests are asset pricing and macroeconomics.