Forecasting market returns: bagging or combining?

Abstract

This paper provides a rigorous and detailed analysis of the methods of bagging, which addresses both model and parameter uncertainty. We provide a multi-country study of bagging, of which there are very few to date, that examines out-of-sample forecasts for the G7 and a broad set of Asian countries. We find that, when portfolio weight restrictions are applied, bagging generally improves forecast accuracy and generates economic gains relative to the benchmark. Bagging also performs well compared to forecast combinations in this setting. We incorporate data mining critical values for appropriate inference on bagging and combination forecast methods. We provide new evidence that the results for bagging cannot be fully explained by data mining concerns. Finally, forecasting gains are highest for countries with high trade openness and high FDI. The potentially substantial economic gains could well be operational given the existence of index funds for most of these countries.

The predictive value of inequality measures for stock returns: An analysis of long-span UK data using quantile random forests

“…The log change in the UK Market Return Index is used to estimate stock return. 6 The other predictors used are key candidate variables drawn from valuation fundamentals and macroeconomics, which follow Goyal and Welch (2008) and Jordan et al (2017). Specifically we include the dividend-price ratio, dividend yield, earnings-price ratio, book-market ratio, T-Bill rate, inflation, and stock variance.…”

“…The existing literature on predicting stock returns in developed and developing countries, based on a wide array of models and predictors, is vast, to say the least (see for example, Rapach et al, 2005, 2013, Sousa et al, 2016, Aye et al, 2017, Jordan et al, 2017. On one hand, practitioners in finance require real-time forecasts of stock returns for asset-allocation decisions.…”

“…Given that inequality may affect stock returns through a variety of channels and in a variety of ways, quantile random forests are capable of capturing such complex nonlinear effects in a natural way. In addition, quantile random forests render it possible to capture any interaction effects of inequality with a wide-array of predictors (as in Jordan et al, 2017) generally used in the literature on stock-returns predictability. Finally, as with a more standard quantile regression model (Koenker, 2005), quantile random forests allow the predictive value of inequality for stock returns to be traced out along the quantiles of the conditional distribution of stock returns.…”

“…The data is downloadable from:https://discover.ukdataservice.ac.uk/series/?sn=200016 and https://discover.ukdataservice.ac.uk/series/?sn=2000028.5 We would like to thank Professor Haroon Mumtaz for kindly sharing the inequality data with us.6 The TOTMKUK(RI) is the data series used as in many previous studies includingJordan et al (2017).…”

Predicting the equity market with option-implied variables

“…We impose the restriction that whenever the forecast of the market excess return or of the realized variance (or of both) in Equation (22) equals zero, we set the portfolio weight equal to 1/(2γ). Further, following Campbell & Thompson (2008) and Jordan et al (2017), we impose the restriction that ω t is bounded from below by 0 and from above by 1.5. Economically, the lower bound implies that the agent does not short-sell the risky asset.…”