“When stocks go down, bonds go up” has been a foundational assumption in just about everyone’s portfolio construction decisions for the last 20 years.
But it might be about to break down.
There are no ‘laws’ in investing, because nothing is ever provable and there can be no counterfactuals. But the negative stock/bond correlation feels so enshrined in conventional wisdom at this point it seems more like a law than any other relationship in investing.
And for the last 20 years or so, this law has held true. But it hasn’t always.
In fact, for most of the history of financial markets, bonds and stocks have been positively correlated. Meaning when stocks have fallen, bonds have fallen too.
The negative correlation between stocks and bonds is a relatively recent phenomenon which has only really occurred since about the year 2000. According to Man Group, the stock/bond correlation was positive for almost all of the last 200 years.
- What causes a positive correlation between stocks and bonds?
- Man Group: ‘Fire, then Ice’
- UBS: ‘Stock/bond correlation in the coming decade’
- AQR: ‘Stock-Bond Correlations’
- Magnus Andersson, Elizaveta Krylova & Sami Vahamaa (AKV): ‘Why does the correlation between stock and bond returns vary over time?’
- Lingfeng Li (LL): ‘Macroeconomic factors and the correlation of stock and bond returns’
- DE Shaw: ‘Positively Negative: Stock-Bond Correlation and Its Implications for Investors’
- PIMCO: ‘US stock-bond correlation: What are the macroeconomic drivers?’
- Summary: What causes a positive correlation between stocks and bonds?
- Who cares?
What causes a positive correlation between stocks and bonds?
Because it’s such an iconic relationship, there’s been a huge amount of great research on the topic. So I’ve collected some of my favourite pieces together, to see if there’s any commonality which might help guide the way when constructing our own portfolios.
I’ll briefly summarise each of the articles here, along with any interesting charts and overall conclusions of each of the papers. If you’re feeling like a real millennial and want to just skip to the overall conclusion, then that’s OK too.
Man Group: ‘Fire, then Ice’
This is a great one to start with because it puts today’s negative stock/bond correlation into context, and importantly looks at things from a UK perspective. Most of the other evidence we’ve got here looks at the US, which is worth bearing in mind as we go through.
It definitely looks like the last 20 years have been an outlier rather than the norm.
In terms of what Man Group consider to be the driver of the stock bond correlation, the table below digs a bit deeper into each of the three ‘Episodes’ from the chart above, and compares the start and end correlations in each Episode to the start and end inflation rates and bond yields:
The table paints a pretty clear picture of what they consider to be the main driver of the stock/bond correlation. Falling correlations seem to be driven by falling inflation and falling bond yields, and vice versa. Here’s their conclusion in their own words:
‘Furthermore in all three cases [the three ‘Episodes’] the peak positive stock bond correlation roughly coincided with the peak of the yields, but importantly the trough in the stock bond correlation occurred almost exactly on the trough in the bond yield. In other words, as bond yields fell from high or relatively high levels, the stock bond correlation fell and as bond yields troughed out the stock bond correlation troughed. Now these are just three episodes but it does seem crystal clear from the trends that, historically, higher bond yields equated to a positive stock bond correlation and lower bond yields to a lower or zero or even negative stock bond correlation.’
A nice short article, UBS look like they’re in agreement with Man Group. They also try to figure out whether the stock/bond correlation increases linearly alongside increasing inflation, or whether there’s a ‘tipping point’.
The chart below shows their findings – that core inflation below 2.5% tends to support a negative stock/bond correlation, and anything above that results in a positive correlation.
“All in all, 2.5% core inflation seems to be the threshold of whether stocks and bonds cross over from negative/zero correlation to a positive correlation. With high inflation, stocks and bonds suffer and benefit together, as higher inflation hurts bonds as investors factor in higher interest rates, and equities suffer from pricing uncertainties and cost pressures. In low inflation environments, the discount rate for equities is more stable and it is earnings growth expectations that are at risk.”
More on this idea later…
AQR: ‘Stock-Bond Correlations’
A bit of an older one (2003), but anything from AQR is 100% worth reading. The more you read from them, the more you realise that every interesting and relevant topic was already covered by Cliff over 20 years ago, and far more rigorously.
When it comes to the stock/bond correlation, their conclusion is:
“We have argued that stock–bond correlation tends to be lowest when inflation and growth are low and when equities are weak and volatile — flight to quality episodes”
Their findings are summarised in the chart below. It’s interesting to see that because the paper was written in 2003, none of the stock/bond correlations are negative – they’re only varying degrees of positive:
Here we have a few academics writing for the Journal of Applied Financial Economics:
“The results indicate that stock and bond prices move in the same direction during periods of high inflation expectations, while epochs of negative stock–bond return correlation seem to coincide with subdued inflation expectations. Furthermore, consistent with the ‘flight-to-quality’ phenomenon, the results suggest that periods of elevated stock market uncertainty lead to a decoupling between stock and bond prices. Finally, it is found that the stock–bond return correlation is virtually unaffected by economic growth expectations.”
So they’re in agreement with AQR on the ‘flight-to-quality’ subject, and also agree with others that positive correlations are observed during periods of high inflation expectations.
Lingfeng Li (LL): ‘Macroeconomic factors and the correlation of stock and bond returns’
This is a useful paper, as it analyses the stock/bond correlation of the G7 countries, not just the US/UK.
From the chart below, you can see that all G7 countries (bar Japan) have had very similar trends over the last 60 years for the stock/bond correlation. A positive correlation between 1965-2000 was the norm, not just in the UK and the US, but among the other major world economies too – with all of them showing declining/negative correlation post-2000.
Interestingly, Japan’s post-1990 lower stock/bond correlation coincides with their slashing of interest rates following the stock market bubble which peaked in 1989. This lends further support to the idea that higher inflation expectations are a major cause of a higher stock/bond correlation.
As for the cause of a high stock/bond correlation, the authors believe the main culprit to be uncertainty around long-term expected inflation:
“Our main findings are, first, that uncertainty about long-term expected inflation plays an important role in determining the major trends of stock-bond correlations. Greater concerns for future inflation are likely to result in stronger comovement between stock and bond returns. Secondly, we demonstrate that the uncertainty about other macroeconomic factors, such as the real interest rate and unexpected inflation, also affects the co-movement of stock-bond returns, but to a lesser degree.”
This paper takes a deeper dive into the causes of the stock/bond correlation. Interestingly, they start with the assumption that it’s rates which determine the correlation, but argue that not all rate shifts have the same effect on the correlation. They outline four scenarios for the stock/bond correlation, all of which result in interest rate shifts, but only two of which result in a positive correlation.
Scenarios one and two show a positive correlation for stocks and bonds, and scenarios three and four show a negative correlation.
If the central bank is expected to raise rates to combat an unexpected rise in inflation or inflation expectations (scenarios 1 and 2), stock prices may fall alongside bond prices. Stocks would likely fall on the back of a combination of: rates rising faster than inflation, a higher discount rate, lower economic activity, and a higher equity risk premium in a more volatile environment.
But the situation is different if the increase in yields comes from improving prospects for economic growth (scenario 3). Although this suggests higher rates, investors would also expect higher earnings, potentially outweighing the drag of higher interest rates and boosting equities.
Finally, scenario 4 shows investors becoming more willing to hold risky assets – either because they perceive less overall risk, or simply become more tolerant of risk, they will push equity prices higher and bond prices lower.
The paper contains a couple of interesting charts to show the effect of inflation expectations on the correlation.
The first chart shows how longer-term inflation expectations were high and variable over the last 50 years, which the authors argue reflected the poor credibility of the Fed in delivering low inflation and considerable market uncertainty regarding the Fed’s policy approach. Under those circumstances, the factors reflected in the first two columns of the table (higher inflation expectations/hawkish policy surprise) were critical drivers of market fluctuations, which induced a positive correlation between stocks and bonds:
But in the late 1990s, the Fed managed to stabilise inflation expectations at c.2%. Over this period, the Fed has become clearer on its policy framework for maintaining that outcome, boosting its credibility among investors and making its policy decisions more predictable. This policy shift has lessened the importance of the first two scenarios, leaving the market to be driven by shifts in growth prospects and risk preferences (scenarios 3 and 4), causing the negative correlation we’ve seen since.
The second chart looks strikingly similar to the UBS chart, but plots inflation expectations against the stock/bond correlation, rather than the UBS paper which charts CPI vs the correlation:
Again, it’s pretty clear that the more stable 2% inflation expectations in the post-1997 era have resulted in a lower correlation. Before 1997, inflation expectations seemed all over the place, and saw a higher correlation as a result. This supports their argument that the correlation is driven by inflation expectations, and the Fed’s ability to anchor expectations at around 2% has resulted in a lower correlation.
They’re also very confident that inflation is the culprit, because, as we saw in the Li paper, this relationship hasn’t just held in the USA:
“There is also support for our hypothesis from patterns observed across other countries. The shift in the stock-bond correlation has occurred in almost all advanced economies, and it generally lines up with inflation expectations reaching low levels in those cases. A particularly informative example is Japan, where the correlation turned negative earlier than observed in other countries, coinciding with an earlier decline in inflation expectations in that country.”
And the authors on whether it will persist:
“If central banks were to lose their grip on inflation or allow policy to become more erratic, we would expect a significant and painful reversal of the favourable market dynamics that have characterized the past two decades. The term premium would likely shift higher, and investors would face a substantial increase in aggregate portfolio risk, as stocks and bonds would begin to move together.
However, we believe that a dramatic shift in this direction is unlikely. Our view is that develop economy central banks will continue to achieve success in keeping inflation expectations low and stable, allowing the negative correlations between stocks and bonds to persist for the foreseeable future.”
This is obviously a US-specific study, but it’s still worth reading through. It’s probably the most complex of the lot, and come to some slightly different conclusions to the others, too.
Ultimately, the authors find that the stock/bond correlation is driven by three factors: (1) the volatility of interest rates, (2) the covariance of growth and rates, and (3) the covariance of stock and bond risk premia.
But first, here’s a great colourful chart from the paper showing the history of the correlation:
We’ve seen that one a few times before, so not much to add to it at this point.
Diving into the detail, this table provides some high-level summary data behind the three factors the authors believe affect the correlation: the volatility of interest rates, the covariance of growth and rates, and the covariance of stock and bond risk premia:
Starting with looking into the first factor in a bit more depth – rates volatility.
The table above shows the 3m yield volatility was 2x to 3x higher during the positive correlation regime of 1965-2000 than during the two negative correlation regimes.
It’s hard to disentangle the effect of high rates from the effect of rate volatility, as rates are most volatile when they’re high. But the charts below attempt to clarify things by plotting the stock/bond correlation against both the 3m treasury yield and the 3m treasury yield volatility:
The charts show that higher and more volatile rate periods are associated with positive stock-bond correlation, while lower and more stable rate periods are associated with negative correlation. In other words, stock-bond correlation is positively correlated with the level of 3m yields (left chart) and with yield volatility (right chart).
Digging into the second factor which the authors believe contribute to the correlation – the covariance of economic growth and interest rates.
Their logic is that if periods of stronger economic growth (with stronger equity cash flow growth) coincides with periods of higher rates, then stock and bond prices will tend to move in opposite directions, all else equal.
This is exactly the same idea as scenario 3 from the DE Shaw paper above, who also argue that higher rates caused by economic growth are likely to result in a negative stock/bond correlation.
The chart below shows there’s a negative relationship between the stock/bond correlation and the growth/rates correlation when looking at sequential, non-overlapping 5y periods. Periods of higher and positive growth-rates correlation correspond to periods of lower and often negative stock-bond correlation, and vice versa:
So again, we’re seeing results consistent with what we’ve seen in other papers – that higher rates caused by strong economic growth are likely to result in a negative stock/bond relationship.
Turning to the third and final factor – the covariance of stock and bond risk premia.
The authors argue that if the stock and bond risk premia move together – regardless of direction – then stock-bond correlation tends to be positive, and vice versa. For example, if investors become generally more apprehensive about the future (say, worries about slow growth and higher inflation), they may more heavily discount all assets that promise future cash flows and prefer cash, producing positively correlated changes in both bond and stock risk premia. The chart below shows a shared upward trend for BRP and ERP from 1965 to around 1980, followed by a clear shared downward trend to 2000 reflecting either lowered perceived risk, greater risk tolerance, or both:
However, depending on the macroeconomic environment, changes in stock and bond risk premia can diverge as well. For example, investors may suddenly view stocks as having high risk and bonds as low risk, causing ERP and BRP to become negatively correlated. These negatively correlated movements, related to mean-reverting “risk-on, risk-off” periods are generally short-lived, but can be a dominant source of risk premia co-movement.
So, the authors argue, whether the stock/bond corelation is positive or negative depends on the competing forces between longer-term trends in the risk premia correlation and the shorter-term “risk-on, risk-off” periods where one of the two asset classes is clearly in favour.
The presence of a common trend in risk premia (first upward from 1965-1980 and downward until 2000) contributed to the 1965- 1980 positive stock-bond correlation regime. However, the short-term “risk-on, risk-off” movements were more dominant, with ERP and BRP negatively correlated as shown in Figure 5. Nevertheless, this component was not sufficient to prevent stock-bond correlation from being positive.
Beginning in 2000, the bond risk premium continued to trend lower, while the equity risk premium began to trend higher for about a decade. This divergent trend movement combined with divergent short-term “risk-on, risk-off” movements have produced a negative ERP-BRP correlation that has, in turn, helped to support the recent prolonged period of negative stock-bond correlation.
Overall, I’m not sure how convinced I am by the third factor, but the first two factors agree to what we’ve already seen in other studies. Higher rates and volatile rates suggest a positive stock/bond correlation, but higher rates caused by strong economic growth are still supportive of a negative correlation.
Here’s a handy graphic summarising their views on the factors affecting the stock/bond correlation, and their conclusion in their own words:
“Linking stock-bond correlation regimes to economic conditions and policy provides investors with a roadmap but is no panacea. Anticipating changes in correlation is an exercise in foreseeing how policy makers will likely behave and how economic data respond. Although the current negative stock-bond correlation regime has coincided with persistently falling and low interest rates, continued low interest rates alone are not enough to support a negative correlation. Focus ought to be on the broader policy backdrop and its implications for: interest rate volatility, the co-movement of bond and equity risk premia, and the co-movement of economic growth and interest rates. As such, we advocate that CIOs vigilantly monitor the key economic and policy developments that could lead to a shift in stock-bond correlation.”
Summary: What causes a positive correlation between stocks and bonds?
- Man group: Rising inflation/bond yields
- UBS: 2.5% core inflation
- AQR: When inflation and growth are high, and outside ‘flight-to-quality’ episodes
- AKV: High inflation expectations
- LL: Uncertainty about long-term expected inflation
- DE Shaw: Unstable inflation expectations, or perceived shifts in the U.S. Federal Reserve’s policy reaction function
- PIMCO: 1) high/volatile rates, 2) low covariance of growth and rates, and 3) high covariance of stock and bond risk premia
Nobody knows for sure what causes the stock/bond correlation to turn positive. And that’s not surprising.
As I mentioned at the beginning of this post, there are no investing laws.
One of the things I find fascinating about portfolio construction is that it’s all about creating a portfolio based on inconclusive data from the past, to be implemented in an evolving present, for the benefit of an unknowable future.
So knowing anything for certain is impossible.
But there are a few common threads appearing in the research which can help guide our portfolio construction decisions. The evidence seems to mostly agree that rising stock/bond correlations are caused (at least in part) by the combination of high levels of inflation, and uncertainty around future inflation.
So where does that leave us today? Is it likely the stock/bond correlation will turn positive?
As for the raw level of inflation, nobody knows what future inflation is going to be. We can no better forecast inflation than future stock prices. And although we can’t make portfolio construction decisions based on some vague forecast for future inflation, it’s clear that as at the time of writing (December 2021), inflation seems to be rearing its head both in the UK and the US. So it’s very possible that inflation could move higher in the future, which would be bad news for the stock/bond correlation.
And as for the second component which may cause the stock/bond correlation to rise – uncertainty about future inflation – the research suggests that since central banks have been clearer in letting the market know their thoughts around future rates movements, this has resulted in uncertainty around inflation being structurally reduced. While inflation can come and go, a more permanent reduction in inflation uncertainty caused by a more transparent and enforced inflation target may exert a more lasting downwards pull on the stock/bond correlation.
The adoption of the implicit 2% inflation target by the Fed in 1997 marks the boundary between the prior period of (mainly) positive stock-bond correlations caused by high and uncertain inflation, and the current period of (mainly) negative stock-bond correlations caused by lower and more certain levels of inflation. Is this a coincidence? The evidence suggests otherwise.
So assuming the Fed (and other) central banks have learned from their past mistakes, and remain adept at managing market expectations and maintaining the market’s confidence that inflation can be kept at low and sustainable levels, then that will encourage the stock/bond correlation to remain low.
But here’s the thought I want to end on: should we even care about the stock/bond correlation?
Do we care if our bonds go up when our stocks go up? Not really – we’re happy everything’s going up. Do we care if our bonds fall when our stocks go up? Not really – we’re happy because our stocks have gone up. Do we care about our bonds going up half a percent when stocks fall 3%? Not really – a 3% fall in stocks isn’t a big deal.
What we really care about is whether bonds are able to provide meaningful protection when our stocks fall 20%+. And that’s not measured particularly well by long-term correlations. That’s because these correlations tend to be measured over periods which include both crashes and periods of strong equity markets (so are averages), and also because correlations assume equity returns are normally distributed – which doesn’t seem particularly helpful when we’re worried about our bonds’ ability to hedge a left-tail event.
So I think the question most investors really care about isn’t whether the stock/bond correlation will turn positive, but whether bonds will still provide the same level of crash protection a) right now, with yields so low, and b) in a future environment where inflation is higher and rates are rising. In short, will the stock/bond correlation remain negative when stocks crash?
To help answer that question, I thought it would be interesting to look at whether, in any of the major countries going back as far as we have reliable data, bonds have not managed to provide diversification benefits when stocks have fallen – i.e. in all the past market crashes going back to the 1920s-ish, have bonds ever not gone up when stocks have suffered a large drawdown? And if not, why not?
This post is getting pretty long now, so although I wanted to include this analysis here, I think it deserves its own post. Given it’s a natural follow-on from this one, that’ll be my next article.
But to answer this article’s original question of whether the long-term stock/bond correlation will increase: on the one hand we have the looming spectre of higher inflation which suggests an increasing correlation, but on the other hand we have the more explicit and enforced central bank policy of maintaining a 2% inflation rate, which reduces inflation uncertainty, and so suggests a lower correlation. We have two competing forces pulling the correlation in opposite directions.
So yes, it’s very possible that the stock/bond correlation could increase. But whether this happens or not depends on central banks’ responses to rising inflation. If they do a good job in convincing the market they have inflation under control, and they’re able to pull inflation back to the long-term 2% target, then the correlation is likely to stay negative. If not, they it looks like a more positive correlation is on the cards.
As I said before, knowing anything for certain is impossible. But it’s worth considering now more than ever the effect of rising correlations on our portfolios, and whether our portfolios are adequately positioned in the event the historic negative stock/bond correlation reverses.
We’ve covered a lot of theory in this post, and hopefully it’s shed some light on the factors affecting the stock/bond correlation. But what we really want is practicalities. Is the role of bonds as an equity hedge impaired going forwards? Do we need to adjust our portfolios if we can’t rely on a negative stock/bond correlation in the future? What does all this mean for our portfolios?
And that’s a post for next week.