For the cooper price data, we have relied as usual on the USGS data (which we expanded back to 1850 also from USGS). For the S&P 500 data we have used Robert Shiller's data (Yale) going back to 1871. For the calculation of total return we took the dividends plus index valuation for any annual period.
Overall the picture is rather disappointing in the sense that there is really no dependency structure between the two variables as can also be seen in the graphic below:
A left tail dependency might be suspected visually (especially the 1931 data point with -37% copper price change and -40% S&P 500 total return), but corresponding statistical tests don't confirm such relationship.
Obviously the hypothesis of the high co-dependency between copper price changers and stock market return dates back to the Great Depression, where there correlation for the time period 1925 to 1935 was 65%.
However, in the recent episode surrounding the Great Recession, this co-dependency was much lower with a correlation coefficient of 31%.
In a future post we will look at monthly time series to better evaluate lagging behavior (for above episodes, it seems that S&P 500 was somehow leading copper price).
