- Gold did not act as a hedge against the 2008 financial crisis.
- The price of gold is highly sensitive to interest rates. The relationship is sufficiently strong that a model can be constructed to calculate gold prices in terms of interest rates.
- If gold were acting as a hedge, a premium should be observed, indicating that other factors are pushing the price beyond what interest rates can justify.
- Analysis shows gold was trading at a significant discount to its expected value given the interest rates during the crisis.
A short time ago I started looking at the question: "Is gold an effective hedge against a financial crisis?" Having studied the question in more detail, I find the answer is no, gold is not an effective hedge against a financial crisis.
I decided to use the 2008 financial crisis for this analysis because the impact was far more broad than any other crisis since the crash in 1929. The financial crisis started in August 2007 and I chose to use June 2009 as the end date. This gives enough room for the entire crisis to have passed, but not allowing the recovery to have a major impact on the analysis.
It can be somewhat hard to define exactly what it means for something to be a hedge against a financial crisis. Put options could be considered to be a hedge against crisis, but continuously rolling put options to maintain the hedge would cost a lot of money and over the long run wouldn't likely be effective. Nor is it enough for an asset to simply increase in value during a crisis. As I see it, the asset needs to at least retain its value during the crisis and would preferably increase in value to some extent. Furthermore, the hedge should exhibit a premium that is unexplained by normal factors, or, barring that, the hedge should prove to be less volatile than the underlying factors. What I mean by "underlying factors" will become more clear later in the analysis.
GLD data by YCharts
^GSPC data by YCharts
TLT data by YCharts
In the chart above, you can see the relative performance differences between the S&P 500 (Pending:GSPC), Treasuries (NYSEARCA:TLT), and gold (NYSEARCA:GLD). The stock market hit its lowest point in March of 2009, down more than 50%. Both GLD and TLT remained above the breakeven point for the entire period. GLD finished the period with the greatest gain by a considerable margin compared to TLT.
Now, TLT holds long dated Treasuries (20+ years), so it's easy to see that the corresponding yields were falling as the prices of the bonds increased. However, of the various interest rates, GLD shows a particularly strong relationship with the 10-Year Treasury rate (TNX).
Source: Data by Yahoo Finance for GLD and TNX.
The plot above shows the relationship between GLD and TNX. This would be a good time to point out that I have a reason for using GLD rather than the actual price of gold. In short, some of my research on this matter (link) indicates that GLD could be responsible for a meaningful change in the way gold is traded. In addition, further reading indicates that the conection between gold and inflation has been very weak for the entrie time GLD has existed (since late 2004). I'm not going to speculate on the impact GLD has had on gold trading (if any) in this article, but I consider the information above enough reason to limit the range of consideration to the lifespan of GLD.
It makes sense that there is a negative relationship between GLD and TNX because GLD has no yield. This is important because interest rates were falling during the crisis. Looking for a linear fit for the two variables yields the prediction equation: GLD = 219.24-34.28*(TNX) with R2=0.7717. Note, TNX is quoted in 100*0.01% (i.e., TNX=3.5--->3.5%; this formula assumes you're using 3.5 not 0.035).
This give a simple model for estimating what the value of GLD should be given a particular interest rate. I used the model to assess the performance of GLD during the crisis relative to what would be expected.
Source: Data by Yahoo Finance. Model as specified above.
The top chart shows an overlay of the predicted value of GLD and the expected value of GLD from the model. The bottom chart shows the nominal difference between the two: GLD - model value. Using the model, I calculated the premium/discount of GLD vs. the model as shown below.
Source: Data by Yahoo Finance. This is simply a transformation of the chart above.
The blue line shows the actual percentage change of GLD during the crisis. The red line is the premium/discount percentage vs. the model. This is the heart of my overall argument that gold does not effectively hedge against a financial crisis. Even though GLD did increase in value as would be expected if it were a hedge, the increase can be more than be accounted for by interest rates. During the worst parts of the crisis, GLD had a steadily decreasing premium vs. the model. That suggests that GLD was undervalued for the second half of the crisis period.
Furthermore, GLD exhibited considerable volatility during that period relative to TLT and both are dependent on interest rates. Well, I don't think it's fair to say TLT is dependent on interest rates because the yield on Treasuries is used as the interest rate. That was the reason for my earlier statement about underlying factors for the hedge. It wouldn't make sense for gold to be considered a hedge (as I defined it) because there is a major factor that is driving its price and would make a better choice than gold itself.
Therefore, considering all of the analysis above, I have concluded that gold is not an effective hedge against a financial crisis. It was undervalued for much of the crisis, according to my model. The same point can be made by looking at the worst part of the crisis in the very first chart. Notice that TLT peaked first and then slowly started to drift down as the markets began to move past the crisis. GLD continued to rise even as the crisis was coming to a close and would rise further after the crisis. All of the evidence points to the rise in gold not being a result of it being a hedge against crisis (or a safe haven if you prefer the term).
Now, before anyone loses their mind, I'll point out a few counterpoints to my argument (believe it or not, I'm not in love with my own ideas and more than happy to accept counterarguments).
- This is a bit of single variable analysis. While the overall model seems to have a high correlation, that's not the same thing as causation. To pin the price of gold to a single factor is pushing it, in fact, I didn't want to only use single variable analysis in the first place (see next point).
- The crisis caused the usual relationships to stop working. There is very little I can do about that fact, no matter what the analysis is. It can be argued that the relationship between GLD and TNX just doesn't hold in that environment or that other factors become more important.
- The next crisis will be different than any crisis that has come before. I think that is likely to be a true statement, but in that case it's hard to see how anything can be considered a hedge against such a crisis. We cannot know the nature of a crisis in advance, so is it impossible to hedge against? If you can make an educated guess, how much time do you need?