- Inflation expectations and market conventions imply gold should be rising.
- CPI and gold price data tell another story.
- It appears a few outlier months in 1979 have created a market misconception.
We have long heard stories that rising inflation leads to dramatic increases in commodity prices. The late 1970s/early 1980s and related moves in inflation and gold (NYSEARCA:GLD) are the prime examples given. We have also heard that all of the quantitative easing and stimulus in the world ends in massive inflation which, in turn, should send gold to the moon. The arguments are compelling.
Given the potential for runaway inflation, we would expect gold to be rising by leaps and bounds, but it isn't. One explanation is the inflation so many expect is not coming. Another possible explanation is the relationship between gold and inflation is not what convention suggests. Given this, we decided to examine the relationship between gold and inflation (as measured by CPI) historically to see what we could find.
We obtained CPI data from the reporting body, the Bureau of Labor Statistics, back to 1913 from their website located here. Next, we went to the St. Louis Fed's website and downloaded gold prices using the "GOLDAMGBD228NLBM" symbol. Gold prices went back to 4/1968, which served as the starting point of the analysis. All data were monthly, due to the reporting frequency of the CPI numbers.
We began by looking at the month-over-month CPI percentage changes against average gold prices in the same month.
In the chart, we would expect to see an upward sloping regression line. This is because high CPI changes should correspond to higher gold prices (buy gold in inflationary periods) and vice versa. In the case of gold, we actually see a downward sloping line, implying the relationship is opposite of expectations.
Next, we did the same month-over-month CPI percentage change, but compared to the average gold prices in the month following the CPI report. We did this because we wanted to see if the CPI effect was felt more in the next month (instead of current month).
Here we found about the same relationship as the prior chart.
Given the lack of expected relationship, we decided to check the CPI change against the average month-to-month percentage change in gold. So if CPI went up 2% from last month to this month, we wanted to see what gold prices did in percentage terms comparing the average of this month to the average of next month (we used average monthly gold prices). The thought was, maybe the relationship would be stronger viewed in these "normalized" terms.
Even here, there was no relationship. We were very surprised. One caveat is our method of using average gold prices for the month instead of prices on the precise dates could be masking any true relationship, but even if that were true, we would expect a stronger relationship than we are witnessing. We decided that perhaps the month-over-month data wasn't the best way to examine things, and chose to move to year-over-year (the inflation metric generally used by central bankers in targeting price levels).
We did the same three analyses from above, this time using the year-over-year CPI change numbers in place of the month-over-month used above. The following chart looks at year-over-year CPI percentage change against average gold prices in the same month.
Amazingly, while stronger than month-over-month, there was still no real relationship here either. One thing is worth mentioning: in the gold chart above, we used a polynomial regression, which includes a quadratic term in the regression equation. This is a fancy way of saying we used another form of regression that attempts to find the best fit to the data when the relationship is obviously not linear (the gold data above has a tilted "U" shape). Regardless, we see that the highest gold prices correspond to lower inflation levels, which goes against traditional expectations.
As above with the month-to-month analysis, our next step was to examine the year-over-year CPI changes against the one-month lagged average gold prices. The chart was so similar to the chart directly above that we decided to save the real estate. In essence, there was no real relationship there either.
The final chart looks at year-over-year CPI percentage changes against average gold price percentage changes (similar to step 3 in the month-over-month analysis above).
Here too, even "normalizing" by using percentage changes, we found no relationship.
We were very surprised by the results. We expected at least some relationship to support conventional norms, but it didn't exist as we structured our analysis. We decided to examine two last things to see if we could find more clues.
First, we wanted to see how gold prices fared after 3 consecutive up months in year-over-year inflation. For example, if the year-over-year number was +0.5% in December, +1% in January, +2% in February, and +3% in March, we would check the average gold price percentage change from March to April. We then found the average of all of these observations to see if it was higher than the all-weather (non-conditional) average. The thought was perhaps an upward "trend" in CPI would unveil some stronger relationship to gold. We found each instance where the phenomenon described occurred, and found the associated average and median returns on the set of qualifying observations. We then compared this to the all-weather average and median returns. Rounding to the nearest 2 decimals, the numbers were almost identical. We take this to mean a trend in rising CPI does not significantly change the average return in gold.
Next, we examined the 1978 to 1981 period to see if anything stood out that might explain why the commonplace convention that inflation brings gold price increases wasn't showing up in the data. We found the average of monthly year-over-year CPI changes in that period to be +10.65% (median was +10.78%), compared to the average since 1968 of +4.39%. In other words, the '78-'81 period experienced high relative inflation. Next, we looked at the average monthly change in gold prices in the '78-'81 period and compared that to the long-run monthly average change. The '78-'81 period average monthly change in gold prices was +2.09%, while the long-run monthly average was +0.78%. In other words, gold was performing incredibly well in that period, implying high inflation led to higher gold prices. But there's more.
We then looked at the median of gold returns in the '78-'81 period, and found it was +0.19% compared to +0.14% in all-weather, which raised an eyebrow (oftentimes, median versus average discrepancies point toward outliers skewing results). Looking further, we found that November and December of 1979 experienced gold returns (again, comparing average monthly prices, not closing prices) of +18% and +45%. Stripping these two outliers out of the calculation, we found the average of the period dropped from +2.09% to +0.79% (almost equal to the long-run average return of +0.78%). Our point is simple, 2 outlier months (November and December) appear to have driven the majority of the gold price appreciation in the '78-'81 period, and outside of those months, the returns in that period were approximately average (despite the higher-than-usual inflation).
The Iran-Iraq war of the era is often said to be responsible for skyrocketing gold prices, so it could be that played a part, but it technically didn't begin until September 1980, according to Wikipedia. Paul Volker assumed his role as chairman of the Federal Reserve in August of 1979, and subsequently raised the Fed Funds rates some 30% in a few months. It could be his presence caused the market to test his inflation-fighting resolve in the gold market. Having not been involved in the market in the late 1970s, we aren't entirely sure what was happening qualitatively. Regardless, the fact remains that based on this analysis, it certainly appears a few months of outliers and the associated psychological trauma caused have gotten expectations regarding the relationship between gold and inflation away from reality.
Perhaps it is inflation expectations, not actual inflation, which causes gold prices to rise -- but this doesn't make a lot of sense given the volume of people out there expecting inflation to grow exponentially. Perhaps there is some other quantitative combination not considered here that makes the relationship stronger. Perhaps it is an expectations-relative thing, meaning if the actual CPI report came in above or below the expected CPI number, there would be a stronger relationship. Perhaps some other measures (money supply) have a tighter relationship to gold prices. Or perhaps, there really is no relationship at all, it is just perceived to exist given the "traumatic" events of late 1979.