Christopher Geczy
University of Pennsylvania – The Wharton School, Finance Department
Mikhail Samonov
Forefront Analytics
May 18, 2015
[drizzle]
Abstract:
Extending price return momentum tests to the longest available histories of global financial asset returns, including country-specific sectors and stocks, fixed income, currencies, and commodities, as well as U.S. stocks, we create a 215-year history of multi-asset momentum, and we confirm the significance of the momentum premium inside and across asset classes. Consistent with stock-level results, we document a large variation of momentum portfolio betas, conditional on the direction and duration of the return of the asset class in which the momentum portfolio is built. A significant recent rise in pair-wise momentum portfolio correlations suggests features of the data important for empiricists, theoreticians and practitioners alike.
215 Years Of Global Multi-Asset Momentum: From 1800 To 2014 – Introduction
The academic study of price momentum has intensified since Jegadeesh and Titman (1993) found that buying winning stocks and selling losers generated significant positive returns over 3 – 12-month ownership. Indeed, there is a momentum to the momentum literature. We did a search for the ‘momentum’ term in the SSRN database and plot the number of studies per year that mention momentum (Figure I). For example, for 2013 there are approximately 300 papers containing the ‘momentum’ term. The continuing academic interest in the momentum strategy is paralleled among both institutional and, recently, retail investors via strategy and products including momentum-focused ETFs and various trend-driven rotation strategies (Antonacci (2014), Faber (2013)). In fact, the cumulative increase of capital allocated to this strategy might partly have caused the significant recent increase of pair-wise correlations between momentum returns within different asset classes.
We set out to create and examine the longest possible history of the global asset momentum effect to better understand its time-series properties, and to better explain the general characteristics already discovered in more recent history. In addition, observation of momentum “crashes” (Daniel and Moskowitz (2013)) in any given asset class would be useful since, while they appear in recent experience, they do so rarely. We find that they are relatively more frequent before the second half of the 20th century, creating a potential underestimation of the inherent risk associated with the premia if only 1950 – present be considered.
The first contribution of this study is the creation of a long-run dataset of global financial asset return histories back to 1800. Using Global Financial Data databases and additional data available through Bloomberg, we create an expanded dataset going back to 1800 which includes 47 country equity indices, 48 currencies (including Euro), 43 government bond indices, 76 commodities, 301 global sectors, and 34,795 U.S. stocks from Geczy and Samonov (2013). In addition to mapping and organizing the large quantity of time series, we also created spliced time series in cases where multiple indices were available, in order to create the longest possible histories for the individual countries and sectors. The splicing effort had an especially notable impact on the availability of long-run global sector data, extended to 1800 for 33 countries. A comprehensive map of identifiers and link dates used in this study is available in Appendix A, and asset return characteristics are shown in Table I.
The second contribution of this study is the extension of the price momentum factor history back to 1800 for this broad and extended sample of assets, both within and across six asset classes. We test for the cross-sectional momentum effect on the mostly untested monthly data from 1800 to 2014. For example, to our knowledge all studies of cross-sectional country-level momentum begin at or after 1965 (Moskowitz, Ooi and Pedersen (2012)). However, there are two recent papers that extend trend and industry momentum tests into the 19th century (Lemperiere et al. (2014), Szakmary and Zhou (2013)). We document that on average, since 1800, momentum effect appears significant in all asset classes, except in commodity spot prices where it is significantly opposite.2 We plot the average of six intra-asset class (country equities, currencies, country government bonds, commodities, global sectors, U.S. stocks) plus one cross-asset class (incorporating stocks, currencies, bonds, and commodities) momentum top- and bottom-third portfolios’ 10-year rolling and log-cumulative excess returns. Generating a 215-year history of global multi-asset class price momentum (Figure III), we document that the momentum return is consistently significant in each asset class, across asset classes, and in combination.
See full PDF below.
[/drizzle]
