I Why the New York Stock Exchange Crashed in 1929 and 1987 and Why It Could Crash Again

Size: px

Start display at page:

Download "I Why the New York Stock Exchange Crashed in 1929 and 1987 and Why It Could Crash Again"

Felicity Potter
6 years ago
Views:

Why It Could Crash Again Lester G Telser, University of

1 From the SelectedWorks of Lester G Telser November, 2010 I Why the New York Stock Exchange Crashed in 1929 and 1987 and Why It Could Crash Again Lester G Telser, University of Chicago Available at:

2 1 veloc.nb I. Why the New York Stock Exchange Crashed in 1929 and 1987 and Why It Could Crash Again Lester G. Telser Abstract. Between January 2, 1929 and December 31, 1988 the Dow-Jones Index of 30 Industrials moved up or down more than 2 percent on 958 trading sessions out of the 16,084 total. Twice in October, 1929 and October, 1987, the drop was more than 23 percent. An unmistakable striking similarity in the pattern of trade before both Crashes explains why they occurred. Information and Market Efficiency, Event Studies. JEL G14. ü Prologue So far the New York Stock Exchange has crashed twice, first at the end of October, 1929 and second in the third week of October, The Dow-Jones Index of 30 Industrials fell more than 23 percent in both Crashes. The volume of trade relative to the number of shares outstanding rose to unprecedented heights on both. No news could account for these Crashes. Nothing like these Crashes occurred at the outbreak of the Second World War, the attack on Pearl Harbor, the assassination of President Kennedy, or the heart attack of President Eisenhower. Suddenly, without warning, stocks traded on the New York Stock Exchange plunged. The explanation emerges from a detailed examination of the daily volume of trading and the size of the open interest. There is an unmistakable and striking similarity in the pattern of trade before and after the two Crashes. Analysis of this pattern casts bright light on what lies behind the Crashes and the aftermath. My analysis begins with an examination of the velocity of trade. Velocity is the ratio of the volume of trade to the number of shares outstanding. These data are available for each trading session from January 2, 1929 to December 31, I divide these data into two samples. The first ends on September 26, The second begins on September 29, 1952 and goes to the end of The first sample has 6967 trading sessions. The second has 9117 trading sessions. In this Part It suffices to have in mind a very simple model. There is a population of potential shareholders described by their own valuations of shares. Each is willing to buy if the current price is below his valuation and is willing to sell if the current price is above his valuation. Parts II and III delve more deeply into what lies behind these valuations. The main result of Part II is a derivation of the properties of a stable distribution of share holders' valuations. Part III describes some determinants of individuals' valuations. The estimated holding period of a stock is the reciprocal of its velocity. A numerical example makes this clear. Say that the daily volume of trade in a stock were of the number of shares outstanding. It would follow that the holding period would be 1000 days. Thinking in terms of holding periods instead of velocity turns our attention to the reasons for obtaining or keeping a position in a stock.

3 2 veloc.nb instead of velocity turns our attention to the reasons for obtaining or keeping a position in a stock. Among these presumably is the hope of profit and fear of loss. Confining attention to the velocity of trade is ill advised owing to the different character of the two time series it uses, volume of trade and number of shares outstanding. Trading volume changes considerably from one session to the next. The number of shares outstanding does not. It has short stretches without any change interrupted by jumps when changes occur. The number of shares outstanding is a step function unlike trading volume that changes daily. Therefore, I calculate the log of each series separately and fit them separately to linear trends for each of the two main samples and to suitable sub samples of each of the main samples. Velocity trends are derived from coefficients of these separate trend regressions. Hence I focus is on trends of volume of trade and the number of shares outstanding, not on levels of these variables. Corporations decide how many shares they issue. The number changes due to stock splits, reverse stock splits, issues from new corporations, disappearances, acquisitions by merger, buy backs, and the like. Although neither stock splits nor reverse stock splits affect the market value of a corporation in theory, corporations carefully decide these with an eye on potential owners they wish to attract. For example, the value of a round lot of Google stock is about 10 times the average value of a round lot of NYSE stock. Google's reasons are of no concern to my analysis. Splits or reverse splits can change the value of a round lot to an amount more appealing to individual investors because a round lot (100 shares) is less costly to trade than an odd lot (< 100 shares). Hence splits and reverse splits can be expected to raise the volume of trade relative to the number of shares outstanding, that is, velocity. The excess demand for stocks determines the volume of trade. Individuals, financial institutions, brokers, floor traders, trust funds, foreigners and others account for the daily volume of stocks. The number of shares outstanding is the supply. ü Findings The first clue toward solving the problem comes from a regression that relates the velocity of trade to the absolute percentage change of the Dow-Jones Index for those 958 trading sessions that were events. An event is a session in which the Index moves up or down by at least two percent. The regression coefficient of velocity in this regression is positive with a t-ratio nearly 12. This means the higher the velocity, the greater is the pressure on the Index. One would see this in a market where the short run excess demand is inelastic. The stronger the pressure on one side of the market as measured by velocity, the bigger is the percentage change in the Index. It must be so in order to attract enough trade on the other side of the market that can accommodate the pressure. This empirical result is encouraging because it shows that standard economic theory can explain these data.

4 3 veloc.nb ü Table : Summary of Regression Statistics for Log of Volume and Log of Open Interest on Trend Sample Variable NOBS a Trend Coefficient b t - ratio R 2 Trend Velocity c all sample 1 log volume * *10-4 1ê 2ê 29-9ê 26@52 log open interest * post Crash 1 log volume * * ê 20 ê ê 31 ê 88 log open interest * test sample 1 log volume * *10-4 1ê 2ê ê27 ê 29 log open interest * all sample 2 log volume * *10-4 9ê 29 ê52-12 ê 31ê 88 log open interest * pre Crash 2 log volume * *10-4 9ê 29 ê52-10 ê 16ê 87 log open interest * test sample 2 log volume * * ê 13 ê ê 16 ê 87 log open interest * Post Crash 2 log volume * * ê 20 ê ê 31 ê 88 log open interest a. NOBS = Number of Observations = sample size b. The dependent variable for the trend coefficient is in the first column. Thus the coefficient of the trend for Log Volume in Sample 1 = *10-4 with t-ratio = Because the dependent variables are in logs, their trend coefficients do not depend on the units in which volume and open interest are measured. c. Trend coefficient for Velocity = Trend coefficient for Log Volume minus Trend coefficient for Log Open Interest. The trend coefficient for the Holding Period would be (-1) X trend coefficient of velocity. Thus Velocity trends down in Sample 1 and Holding Period trends up. The Table has the whole story. Velocity falls during the first sample and rises during the second. This means holding periods lengthened during the first sample and shortened during the second. Now the expected holding period equals the reciprocal of the real rate of return. Therefore the real rate of return was falling during the first sample and rising during the second. Since the first sample covers the Great Depression while the second was very prosperous on the whole, these findings seem plausible but they do not explain what causes the Crashes. Correcting for the effects of the huge increase in velocity on the days of the Crashes, we must look at the trend of velocity after the first Crash and before the second Crash. After the first Crash velocity rises only slightly so the holding period falls only slightly. Hence there is a change in the patterns of velocity before and after the first Crash. However, before the second Crash velocity still shows an upward trend. Correcting for the outlier effects of velocity on the days of the Crashes, the results after

5 4 veloc.nb upward trend. Correcting for the outlier effects of velocity on the days of the Crashes, the results after the First Crash and before the Second Crash agree in sign. Next we turn to the results for the 10 months preceding the Crashes. Now it is better to examine trends of volume and open interest separately. Test 1 has the results for the first Crash. We see that the trend of volume is very slightly down with a t-ratio of only The trend of open interest is decidedly positive with a t-ratio of This shows a rising trend for the supply of stocks and almost no change in excess demand as shown by the absence of a trend in volume. Test 2 has the results for trends in volume and open interest in the 10 months before the second Crash. In this case while it is true that volume trends up with a t-ratio of 3.16, open interest trends up at double the rate of the volume trend. The t-ratio for open interest trend coefficient is The supply of stocks rises twice as rapidly as the demand before the second Crash. Both Crashes have the same explanation. Rapidly rising supply outpaced the demand. Standard economics says that under these circumstances, prices collapse. ü Conclusions The New York Stock Exchange has not crashed since October, However, the market has been very turbulent since summer, 2008 judging by the large number of events so far. It would be desirable to obtain daily figures on volume and shares outstanding from January 1989 to the present. With such fresh data one could see whether the mechanisms I have described here are the same now as then.

Financial Markets I The Stock, Bond, and Money Markets Every economy must solve the basic problems of production and distribution of goods and

Financial Markets I The Stock, Bond, and Money Markets Every economy must solve the basic problems of production and distribution of goods and services. Financial markets perform an important function