Trump's Taylor Rule

Trump’s Taylor Rule

A Jul 28 commentary for Project Syndicate draws on this research note.  

Jeffrey Frankel and Sohaib Nasim

Donald Trump has now gone beyond his usual norm-busting habits in his criticism of Federal Reserve Board Chair Jay Powell.  He has harassed him to cut interest rates by 300 basis points, has heaped abuse upon Powell personally, has trumped up accusations regarding the remodeling of the Fed’s building, and has gone so far as to draft a letter to fire him.   [That is, to remove him from the chairmanship. Even Trump must realize that he doesn’t have the legal power to remove Powell from the Federal Reserve Board altogether.]

It is revealing to look statistically at what determines whether Trump criticizes the Fed for interest rates that are too high versus too low.   Does he have a “Taylor Rule” of his own, guiding desirable monetary policy in response to current economic conditions?

First, let us recall the standard argument that central banks should be independent.  Then we will look at Trump’s Tayor Rule.

1 The importance of Fed independence

Most presidents observe the independence of the Fed scrupulously.  In 1997, soon after I was Senate-confirmed as a Member of Bill Clinton’s Council of Economic Advisers, I was asked by a TV interviewer how the Administration viewed the latest monetary stance of the Federal Reserve.  (Alan Greenspan was Chair at that time.)   I dutifully replied, “We don’t comment on the Fed’s policy, because we have confidence that they will do a good job.” I was soon corrected by a superior: “the first part of your answer was good, but you should have omitted the second part.”  The danger was that, at some future time, one might be asked, are you no longer willing to say that you have confidence in the Fed?  From then on, I did what all Administration officials did and stuck to simply “we don’t comment on the Fed.”

Perhaps the most important insight from monetary economics in the last 50 years is the proposition that a country with an independent central bank, shielded from political pressure, can achieve better economic performance than a country where the central bank is under the direct control of the government.  The latter set-up results in an inflationary bias, where the authorities cannot resist the temptation to stimulate the economy, and everybody knows this ahead of time, with the result that money has to expand enough to offset and validate expectations of inflation, without even reaping a bonus of higher real growth.

The Federal Reserve was granted independence at its founding.  The Federal Reserve Act of 1913 says that the Chair cannot be removed without cause [i.e., malfeasance] and also that, for example, the Board’s freedom to arrange its own building cannot be second-guessed.  Trump has recently asserted the power to violate both of these provisions.  [This even though the Act was recently reaffirmed by the Supreme Court, in recognition of the importance of shielding the Fed from political pressure.]

Even Richard Nixon, the last president to pressure the Fed chair (Arthur Burns) toward an easier monetary policy [accompanied by wage-price controls], with disastrous results, at least had the sense not to do it in public, let alone amidst insults to the Chair’s competence. 

Recognizing the value of central bank independence, most other major countries have, in recent decades, also adopted this reform.

2.  Trumpian behavioral hypotheses

The position of Trump and his supporters is that he is a better judge of what is the appropriate interest rate at any given time than is the Fed.  At Davos on January 23, 2025, he claimed that he understands monetary policy better than Powell.

 Even those who might not believe Trump’s claims to know more about everything than anybody [more about the tax code than tax lawyers, more about technology than the technologists,  more about ISIS than the CIA, more about the military than the generals, etc.,] might reason that he was a successful businessman, and thus must know more about monetary policy than the intellectuals.  So, does Trump use his experience and wisdom to judge when inflation is high or unemployment low, calling for tighter money (higher interest rates), and when inflation is low or unemployment high, calling for looser money (lower interest rates)? That is, does he have a “Taylor rule” of his own?   Another possible hypothesis: Does Trump believe that interest rates are always too high, regardless of contemporary specifics [perhaps as a result of his real estate background]?

3. When does Trump criticize the Fed for interest rates too high?

 We looked at those days on which Trump criticized Fed policy, as posted on his platform TruthSocial, or Twitter before that.  We have also looked at his spoken remarks as reported in the media.  There were 145 such events between January 2013 and June 2025.  (There may be other occasions that this source misses.)  Of these 145, he viewed interest rates as too high and called for easing on 129 occasions, and interest rates too low, calling for tightening on 16 occasions.

We used Firth penalized logistic regression and OLS with robust standard errors to examine whether Donald Trump’s public calls for looser monetary policy follow a macroeconomic logic akin to the Taylor Rule. In addition to including the contemporaneous Fed funds rate, we examined the effects of the unemployment rate and CPI inflation rate on the likelihood of him criticizing monetary policy as too tight. A dummy variable indicates whether Trump was in office—defined to include the transition period following an election victory—and helps test whether his critiques are politically motivated or self-interested. The logistic form is more appropriate linear OLS because of the limited dependent variable, while penalized log-likelihood model avoids convergence issues due to separation.  (The 0-1 nature of the dependent variable calls for the probit technique. But meaningful estimation was not possible due to multicollinearity.)

    In the linear probability OLS model, InOffice is highly significant (coefficient = 0.60, p < 0.001), while the interest rate is significant but small, and the inflation term is marginally significant (p = 0.051) with the wrong sign. 

The Firth model shows somewhat similar results, but the only statistically significant predictor of whether Trump calls for looser monetary policy is whether he is in office. The coefficient on the InOffice dummy is large (6.62) and highly significant (p = 0.002), while the macroeconomic variables—Fed funds rate, unemployment, and inflation—are all statistically insignificant. 

  These results indicate that Trump’s monetary criticisms are not guided by economic fundamentals: he is far more likely to advocate for looser policy when he holds office, regardless of inflation or unemployment. This one variable explains over 76% of the variation in the linear model (R² = 0.76), underscoring the extent to which his position on interest rates appears driven by political context rather than macroeconomic indicators.

Put simply, whenever Trump is in the White House, he  calls for lower interest rates.    (The six data points that we classified as Trump wanting lower interest rates even though he was not in office at the time are essentially claims by him that if he were to become president, his skill would reduce inflation and allow lower interest rates.)

 Evidently, Trump has no Taylor Rule, but rather pushes for easier monetary policy whenever he is in office. Central bank in dependence never looked so good. 

==

.Appendix: Data, Regression Output, and Diagnostics

THIS STATISTICAL APPENDIX IS UNDER CONSTRUCTION

1. Data Construction and Event Classification

We constructed a dataset of all public criticisms by Donald Trump regarding Federal Reserve monetary policy between January 1, 2011 and July 15, 2025. These observations were drawn from Trump’s statements on Twitter (before 2021), Truth Social (after 2021), and additional remarks reported in major media outlets. In total, we identified 145 events where Trump explicitly called for either tighter or looser monetary policy. Each statement was coded as either “ask for easing” (e.g., lowering interest rates) or “ask for tightening” (e.g., raising rates), based on the substance of his message.

The table below summarizes the distribution of these statements by the party in power:

2. Correlation Matrices

These tables report pairwise correlations among the main variables. The dependent variable `WantsLooser` is strongly correlated with `InOffice`, and `InOffice` is also moderately correlated with macroeconomic indicators.

a. WantsLooser vs. InOffice

b. InOffice vs. Macroeconomic Variables

3. Linear Probability Model (OLS with Robust SE)

An OLS regression was run to estimate the effect of macroeconomic variables and political incumbency on the likelihood of Trump calling for easier monetary policy. The model explains 76% of the variation, and the `InOffice` dummy is highly significant.

Dependent variable: WantsLooser | N = 148 | R² = 0.761 | Prob > F = 0.000

4. F-Test: Contribution of InOffice Variable

An F-test was performed to compare the full model to a restricted version that omits the `InOffice` dummy. The test confirms that `InOffice` significantly improves the model's explanatory power.

Test of: InOffice = 0

F(1, 143) = 25.20

Prob > F = 0.0000 

A second F-test was conducted to evaluate whether the macroeconomic indicators (Fed Funds rate, Unemployment Rate, and CPI Inflation 12-month) jointly improve the model over a restricted model using only the `InOffice` dummy.

Result: F(3, 143) = 13.37; Prob > F = 0.0000

5. Nested F-Tests: Contribution of Individual Macroeconomic Variables

To determine which macroeconomic variables most strongly contribute to the explanatory power of the model, we conducted a series of nested F-tests. In each test, one macro variable was removed from the full model to assess whether its exclusion significantly reduced model fit. The results are summarized below:

6. Firth Penalized Logistic Regression

To correct for quasi-complete separation in the binary outcome, we used a Firth penalized logistic model. Only the `InOffice` dummy is statistically significant.

Dependent variable: WantsLooser | N = 148 | Wald chi²(4) = 21.12 | Prob > chi² = 0.0003