Vignette: Commodities for the Long Run

[JustinMShea]

Paper & Data

Link to NBER paper Commodities for the Long Run Parser for long run commodity index data Parser containing long-term equity index data

Figure(s) 1: Cumulative Returns on Commodity Index, 1877–Present

Convert the AQR data into an xts object and then use the plot functions on the series accordingly. Check the plot.xts documentation for help, such as the use of the legend.loc argument.

Table(s) 4: Commodity Index and Aggregate Asset Returns

Table A: Compute commodity index equal-weight & long-short portfolios, stocks, and short-term risk free rate. AQR data & Shiller provided. Use PerformanceAnalytics for useful functions like the arithmetic & geometric mean arguments in Return.annualized, etc.

Table 5: Commodity Index Returns and Aggregate States

Similar to Table 4, with various return & volatility grouped by return state.

Table 6: Commodity Index Futures Returns and Aggregate States

Perform multivariate regressions of the return of the commodity index on type factor variable for the business-cycle, average ex ante carry, and inflation. Inflation and business cycle variables are demeaned. The sample period covers 1877-2015. The regressions are run at 1-month, 1-year and 5-year horizons. This isn't panel data.

Table 7: Commodity Index Futures Returns and Investment Styles

A series of multivariate regressions of the return of the commodity indices on the investment styles of momentum, value and carry. Again, use PerformanceAnalytics functions to compute momentum & value.

• momentum measured by the previous 12-month (annual) return of the index
• value measured as the negative 48-month return 12-months ago (i.e., long-term reversal),
• carry measured as the backwardated/contango value of the commodity index.

The sample period covers 1877-present. Panel A reports the regressions run at 1-month, 1-year and 5-year horizons. Panel B reports regressions of realized monthly commodity index returns across six states of nature (average backwardation, average contango, high inflation, low inflation, expansions, and recessions) on the monthly prediction model estimated in panel A.