On Friday, the headlines were flashing: “Sahm rule triggered”. The reason was that with the US unemployment numbers for July coming out at 4.3%, the 3-month moving average was now 0.5 percentage points higher than its 12-month low.
The Sahm rule has historically been a strong coinciding indicator for recessions in the US – hence the triggering of the rule has clearly sparked recession fears.
While I am personally somewhat sceptical about whether the US economy is in recession, I am nonetheless intrigued by the Sahm rule. In the process of thinking about the Sahm rule, I noticed a comment on X.
A commentator on X wrote something like: “The recession is now a done deal as unemployment has risen for 4 months in a row, and that always means that the US is in recession.”
Unfortunately, I have not been able to find the comment again, and I would like the person who wrote it to get full credit. So if you are that person or somebody who has seen the comment, please drop me an email (LC@paice.io).
Nonetheless, the comment made me think – was it really true that if US unemployment rose four months in a row, it would be a good recession indicator?
I set out to investigate it.
Historical Unemployment Patterns
In the graph below, you see US unemployment since the Second World War, and we see a very clear pattern. It’s hardly surprising that US unemployment has spiked during recessions, marked in the graph with the official recession dating done by the National Bureau of Economic Research (NBER).
In the graph, I have also marked the “trigger dates” from the two rules based on unemployment – the Sahm Rule (red lines) and the 4m-RUIN (blue lines).
4m-RUIN is the name I have given to our X-inspired indicator – it is short for 4-Month Rising Unemployment INdicator.
We see that both the Sahm rule and 4m-RUIN are basically always triggered in the first months of the NBER recessions, but they are rarely leading indicators.
During the 1950s and 1960s, the 4m-RUIN seemed to be a bit more “leading” than the Sahm rule. In 1959, it even “predicted” the 1960 recession. It also sent a recession alert earlier than the Sahm rule in 1973.
Overall, both the Sahm rule and the 4m-RUIN have historically had a very high hit ratio and both have identified nearly all post-Second World War US recessions.
What Happened in 2020?
However, in 2020, when US unemployment spiked more than ever before to nearly 15%, that triggered the Sahm rule, but 4m-RUIN was silent.
Why was that? The reason was that US unemployment only rose two months in a row – March and April – and hence did not trigger 4m-RUIN. One can say that the recession was nearly over before it started.
According to NBER, the recession lasted from February to April 2020. And it was certainly not a “normal” recession. It was a lockdown – and once the lockdown ended, the economy returned swiftly to normal.
Back in May 2020, I wrote the following in my post “When Americans vote in November unemployment will be below 6%”:
“However, where most commentators are wrong is assuming that this has to be seen as a normal recession. I, on the other hand, would argue that this has little to do with a normal recession. In fact, I am increasingly thinking that the use of the term ‘recession’ is a misnomer in relation to this crisis.
…Most people don’t really think about it, but most industrialised economies in the world every year go through large “recessions” in the form of a major drop in economic activity. This happens both on the supply side, as for example Southern Europeans go on Summer vacation typically in August, or with private consumption as it fluctuates wildly before and after Christmas.”
I then went on to argue that unemployment would drop very fast – and it did.
One can, of course, discuss the reason for the 2020 “recession” and whether it was a recession or not, but that’s not really important – 4m-RUIN didn’t get that one “right”, but it got the other “normal” recessions right.
What Normally Happens After 4m-RUIN is Triggered?
Both the Sahm rule and 4m-RUIN are coinciding indicators and as such should not be seen as having any predictive power.
However, we also know that prices and wages tend to be sticky and as a consequence, unemployment is sticky – once it starts to rise (or fall) it tends to do so for a period, and as such, if unemployment has been rising recently, we might take this to be an indication of a further increase in unemployment.
The graph below is an illustration of that.
The graph shows the development in US unemployment prior to and after the 4m-RUIN has been triggered for all the episodes after 1945.
We also see the median of these episodes, and in the “median” episode, unemployment continues to rise around 1 percentage point and remains elevated for around 30 months.
It is, however, also notable that the spread is very large – there are mild recessions (for example 2001) and there are deep recessions (for example 2008-10).
It is also notable that once the 4m-RUIN has been triggered, historically we have already seen a rise in unemployment by more than 1 percentage point prior to the trigger date.
Compared to that, the present episode (the red line) is somewhat milder, and even though unemployment has increased for four months in a row, the accumulated increase in unemployment has been somewhat smaller than during most previous episodes.
And that is, of course, also why this doesn’t really seem like a recession. At least not yet.
The Lucas Critique to the Rescue?
This can, of course, be interpreted in two ways – either this is a mild recession and there is nothing to worry about, or this is just the beginning and we have yet to see the big move in unemployment. But since I am uncertain whether we are in a recession or not (despite the signal from both the Sahm rule and the 4m-RUIN), it makes little sense to speculate about this.
However, there is one thing that makes this even more complicated, and that is that both the markets and policymakers fully understand the message from the US labour market. This is, of course, what we have seen in the market in recent weeks, and compared to, for example, during the 1950s-1970s, the Fed is now forward-looking both in its communication and in the way interest rates are set, and the Fed tends to listen to the markets.
So even though market monetarists like Scott Sumner, David Beckworth and myself are critical about the Fed’s monetary policy framework (we would like to see NGDP level targeting and a market-based approach to implementation of monetary policy), the Fed has nonetheless become increasingly market monetarist in the way it conducts monetary policy.
This means that the Fed is not primarily concerned about the present level of unemployment and inflation, but rather is looking ahead and, among other things, looks at what markets are telling them about the outlook for the US economy. And the signal is pretty clear from the markets – inflation pressures have eased and recession risks have increased.
Consequently, the Fed should cut interest rates.
Markets, of course, also understand this – and therefore, the markets are presently helping the Fed implement monetary easing by pushing down market rates to reflect the expectations of future interest rate cuts from the Fed.
This, however, also illustrates part of the problem with indicators like the Sahm rule or 4m-RUIN. Once the Fed starts to understand them and markets understand that the Fed understands them, then they will become less reliable indicators of future unemployment. This is of course the Lucas Critique – once the Fed using an indicator to conduct policy the indicators stops being a reliable indicator.
Therefore, I am less reluctant to call the recession and fundamentally still believe it can be avoided and that the Fed has the tools to avoid it.
That, however, does not mean that the Fed should ignore the signal from the Sahm rule or the 4m-RUIN, but rather that the Fed should indeed cut rates – and do it soon.
And again, I didn’t come up with the idea for 4m-RUIN – a guy on X did. Could you help me find him?
Note: Illustration created with ChatGPT 4o/Dall-E.
If you want to know more about my work on AI and data, then have a look at the website of PAICE — the AI and data consultancy I have co-founded.