Detecting reversals in the US equity market.

Summary The aim of this thesis is to build a model to detect phase changes - from bull to bear and vice versa - in the US listed stock market, using a relatively large number of variables both fundamental (macroeconomic and microeconomic) and derived from technical analysis.The statistical model used is static logistic regression, with a lag for the explanatory variables ranging from zero to three months. The eight most significant variables out of twenty candidates were selected from monthly S&P500 data over the period 1963-2003. The resulting model was tested over the period 2004-2013 and outperformed the Buy & Hold strategy and a univariate model using the variable with the highest detection power - the latter model having been studied in the literature. It was also shown that variables not yet considered in the literature - the six-month moving average of non-farm net job creation, the monetary base and the OECD Composite Leading Indicator - have significant detection power for our problem. On the other hand, the binary variable indicating the position of the S&P500 in relation to its moving average over the last ten months - a technical analysis type variable - has a much higher predictive power than the fundamental variables studied. Finally, the two other most statistically significant variables are macroeconomic: the spread between the 10-year T-bond and 3-month T-bill rates and the moving average of non-farm payrolls.