President Trump has been pressuring Fed Chair Powell to cut interest rates. Back in January, I made the following comment:

The markets are now scaling back their estimates for future rate increases, due to slower expected NGDP growth. Is that a victory for Trump’s pressure tactics, pushing the Fed toward lower rates? Or a failure–pushing the Fed toward tighter money?

Karl Smith has a fascinating article that discusses some research on this question:

The researchers, from Duke University and the London Business School, focused only on Trump’s tweets since June 2015. They ignored tweets that also contain information about the economy more broadly. They reasoned that there are at least two ways the president could influence the Fed’s choices. First, he could pressure the Fed directly, urging it to alter its response to the macroeconomic environment — by, say, cutting interest rates. Second, he could alter the macroeconomic environment itself — by, say, launching a trade war.

The researchers wanted to isolate the first effect. So they focused only on tweets that were critiques of the Fed, and to measure the response they looked not at the Fed’s actual moves but used high-frequency market data. (Investors and large banks trade futures contracts that are essentially bets on the whether the Fed will raise or lower interest rates.)

What did they find? Within five minutes of a critical tweet by the president, according to the paper, the Fed futures market lowered its expectation of future rates. Each tweet had only a miniscule effect, but the researchers estimate that the cumulative effect has been about one-tenth of a percentage point over the last year. And that effect is growing over time.

The real question is why did interest rate expectations fall.  Was it the liquidity effect, or the income and Fisher effects?  Do Trump tweets lead to expectations of easier money, or expectations of tighter money that will slow growth and inflation and lead to lower interest rates in the future?

Smith discusses some additional evidence in the paper that is not decisive, but certainly suggestive:

They show that Trump’s tweets have little to no effect on expected interest rate changes at the next four Fed meetings. But the tweets have significant and persistent effects on interest rate changes expected nine or more Fed meetings into the future.

That is, the real impact of the tweets is on what investors think the Fed will do a year or more from now.

I don’t want to make too much of this, as the evidence is not strong enough to pin down the mechanism.  And the size of the change (10 basis points in total) is rather small.  But I’d lean slightly toward the contractionary interpretation.  If I were Trump, the fall in fed funds rates far out in the future would make me nervous.  That’s usually not good.

Future research should look at how stock prices react to these tweets.  That would help to pin down whether the effect is expansionary or contractionary.  And yet I don’t want to be too negative, as this sort of event study is exactly what macroeconomists should be doing.  It’s a good start.

I know that readers get bored with my tiresome crusade against using interest rates as the indicator of monetary policy.  But this study is a perfect example of why we need to look at other indicators too.

PS.  The paper makes the following claim:

The estimated revision in expectations might appear small, but it is important to keep in mind that the typical change in the FFR target is 25 bps. A back-of-the-envelope calculation based on two scenarios shows that a decline of 0.5 bps corresponds to a 2% increase in the probability of a 25 bps FFR target cut, which is a relevant change in the probability assigned to an expansionary monetary policy change. Furthermore, the reported coefficient is the average effect of each tweet. The cumulative effect is quite large when taking into account the total number of tweets. Finally, it is worth emphasizing that the typical size of a monetary policy shock is also small, especially in a period of near zero interest rates.

I’m not an econometrician, but that doesn’t seem right to me.  The total effect of the tweets should occur as soon as the market recognizes that there will be a series of tweets criticizing Powell.  In that case, the cumulative effect might have been little more than a basis point.  Here’s another way of making my point.  Markets quickly figure out that there will be periodic Trump tweets bashing Powell.  The expected change in yields on those days is slightly negative.  That means the expected change in yields on tweet-free days is slightly positive.  The cumulative effect requires adding up both the tweet days and the tweet-free days.