GMO’s Predictions: A Useful Guide For Investors?

Edward Tower

Chulalongkorn University-Economics Department; Duke University – Department of Economics; Nanyang Technological University (NTU)

David Barry

Duke University

Edward Stansky

Duke University

Economic Research Initiatives at Duke (ERID) Working Paper


How successful are stock market predictions? We explore one set of easily accessible predictions by a respected firm, GMO. Specifically, we evaluate how effective GMO’s predicted stock returns have been in guiding investors from June 2000 through March 2014. We find that the predictions have been useful, although based on past history investing solely in the top one or two performing indexes would have been an inferior strategy for maximizing return to investing equal amounts in the three indexes with the top predicted returns.

GMO’s Predictions: A Useful Guide For Investors? – Introduction

Beginning in early January 2000 GMO published predicted rates of return for various stock and bond market categories on its web site. The data are predictions as of  the first of each month, and are published several weeks into the month. Initially these were predicted ten year returns. More recently they have been  seven year predicted returns. Initially GMO archived these quarterly predictions. Subsequently GMO removed the early predictions from its web site, so only the more recent quarterly predictions remain. We have saved the quarterly predictions, and we test to see whether these quarterly predictions of five stock categories are useful for investors. One must register to access the web site,, but registration is free.

The Economist has evaluated GMO’s predictions (Buttonwood, 2013), and Tower has written several working papers on the GMO family. This  paper is yet another  one in that series, (Tower 2010a, 2010b).

How High Are The Correlations Between GMO’s Long-run Predicted Returns With Realized Returns?

GMO reports predicted real returns for six categories of stocks: US Large, US Small, US High Quality, Intl Large, Intl Small, and Emerging. US High Quality were added after the series was started, so we exclude them and focus on the other five indexes. To compare predicted returns with realized returns, we use five indexes, which we access from the Bloomberg terminal. These are the S&P500 index, the S&P Small Company index, MSCI EAFE index, MSCI EAFE Small Cap Index, and MSCI Emerging Market Index. The Bloomberg codes are SPXT, SPTRSMCP, NDDUEAFE, NCUDEAFE, and NDUEEGF. The Morningstar codes are S&P 500 TR, S&P SmallCap 600 TR, MSCI EAFE NR USD, EAFE Small Cap NR USD, and MSCI EM NR USD. Initially the predicted returns were ten year predicted returns. We assume GMOs predicted returns are annualized real returns. We work solely with continuously compounded returns, as they have the desirable property that the continuously compounded return over several periods is the average of the continuously compounded returns over the component periods. Thus, all returns in the paper are continuously compounded. They are also real.

Exhibit 1 shows the correlation between the continuously compounded seven or ten year predicted return and the realized seven year return. To calculate the realized real return we use the consumer price index provided by the Bureau of Labor Statistics.

The data are the predicted and realized returns starting from the prediction at the close of the market on the last day of June 2000 through the last day of March 2007. Thus the data series for realized returns ends on the last day of March 2014.

Exhibit 1, shows the predicted and realized returns to have correlations of close to unity during the early periods and lower correlations subsequently, with negative correlations in the most recent five quarters. Over the entire period the correlation averaged 0.581.

The GMO predictions at the start of each month are not published until later in the month. We ignore this lag and pretend that the investor at the beginning of each month has access to the predictions for the last day of the preceding month. This is not too much of a stretch, as the predictions are based on the ratio of stock index price to fundamentals, and the stock index price is observable at the beginning of each month and the fundamentals don’t change much from month to month.


See full PDF below.