Daily Speculations

The Web Site of Victor Niederhoffer & Laurel Kenner

Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter;  a forum for us to use our meager abilities to make the world of specinvestments a better place.



Write to us at: (address is not clickable)

Using Merrill Lynch stock and the stock of a Brokerage house Index to predict the S&P 500 futures, by Owen Wilson

In his book, ‘The Education of a Speculator’, Victor tests an idea by Gerald Loeb, that the performance of brokerage houses can be used to predict future market performance. Loeb was a stock broker on Wall Street for 40 years, a celebrated financial writer for Barrons and author of the classic book ‘The Battle for Investment Survival’. Loeb’s recommendation based on his experiences was to sell stocks when brokerage profits are at a high and buy them when brokerage houses are in the red. In EdSpec Victor recounts one of Loeb’s experiences shortly before the 1929 crash.

‘During these times, Mike Meehan, the famous speculator and floor specialist in the high flying RCA stock of 1929, opened the first office on a steamer – the North German Lloyd luxury express liner Bremen. I sailed on her for Europe in early October 1929. I think it was the maiden voyage of the seagoing board room. At least, so far no firm has opened an office on a plane. I was not immune to the optimism of this period. I laid the groundwork for a stock market country club … fine brokerage offices were other straws in the wind. We built a showplace in Palm Beach. The interior wood was all weather-beaten and genuine, collected along the Atlantic Coast. We had a patio, a fountain, palm trees, of course, a real fireplace and two or three cars to lend just in case a client needed some transportation …’

Shortly afterwards, Loeb formed and quickly sold out an offering for a brokerage house in a golf country club.

Victor tested the idea using data on Merrill Lynch, the largest U.S. brokerage house, from the time it was listed on the NYSE in 1972 through to year end 1995. On a monthly basis Victor showed that unfortunately, at least as far as Merrill Lynch is representative of brokerage stocks, there was no support for Loeb’s inverse hypothesis. He did however find that the excess return of Merrill Lynch and the S & P change the next year were correlated at –0.3.


‘There were five years when Merrill Lynch’s excess return was 30 percentage points or more. In the year immediately following three of these years, the S & P declined. The odds in favor of a decline in those years were 3 to 2, compared to 2 to 15 in the other years. The odds ratio of 11 to 1 suggests that, for yearly returns, brokerage house fortunes are a negative harbinger.’

I am an International Relations student from St Andrews, Scotland, working for Victor through the summer, and my search for studies to run brought me round to updating this one. To substantiate the study, as well as using Merrill Lynch data I formed a brokerage index of 12 of the top 15 brokerage houses, evaluated by market cap. I looked at the percentage returns / excess returns, (over the dependent variable), of these two indicators for time periods t-1 to t-5 to try and predict the S&P 500 futures percentage return for time period t. I used weekly, monthly, quarterly and yearly data.

I found that for the last twenty year period and ten year period, from beginning 1985 and beginning 1995 consecutively there was nothing to indicate an inverse relationship of brokerage profits and the futures market in weekly or monthly data. Both time periods returned correlations of less than 0.01 for all t-1 stats, whilst showing the highest significant monthly correlation for Merrill Lynch % returns and excess returns at (t-5) at approx 0.1.

Looking at quarterly data over twenty years returned interesting, but sometimes insignificant, results

Using data from the beginning of 2000, however, showed much more correlated results.

Looking at weekly data, only for times when the respective independent variable was negative, there was a -0.147 correlation between Merrill Lynch % returns (t-1) and the S&P 500 futures, with a t stat of 2.61.

The following results for monthly data did not all show up as significant, mainly due to a short count of 67 but the Merrill Lynch data and the brokerage data, whilst slightly overlapping, often support each other’s case.

The following tables are divided into three rows; the first row looks at all t-1 data, the second row looks only at data where the independent variable was positive and the third looks only at data where the independent variable was negative.

The correlation then inverts for time period t-2:

For monthly data, when the independent variable was negative, Merrill Lynch % returns and Merrill Lynch excess returns show consecutively;

t-3 : -0.263 , -0.346
t-4 :  0.204 , 0.252