A number of influential macroeconomic models, from Keynes to the monetarists, assume that real interest rates (i.e. discounted for inflation) change over time and are sensitive to their political environment, other words that they are policy variant. However this assumption has not been empirically tested, should this assumption be in fact contradicted, it would have important repercussions on these models’ viability. Moreover, the policy-variant hypothesis apparently conflicts with Eugene Fama’s conclusion that the mean of the real rate is essentially constant (p.322). The literature has found to match Fama’s views more closely with reality: real interest rates are “essentially constant over long periods of times but subject to infrequent mean shifts that are not related to policy regime changes”.
The authors present an hypothesis that would reconcile both parties: real rates are indeed roughly constant, except for a few shifts that could be caused by major political or bureaucratic changes, namely new heads of states and central bankers. The study being centred on the US, the positions in question will be the chairman of the Fed, the president and the Senate (p.323). Considering, that the election of a president from the same political party as the incumbent is merely more of the same, the authors identify three breaks occurring during their period of interest 1961-1986: the end of the Johnson presidency, the end of Ford’s and Reagan’s first election. To this they add the major shift that the election of a Republican Senate represented n 1981-86. They also consider the four changes at the head of the Fed as potential shocks.
Comparing these changes to the real rate data would predict changes during or around the 4th quarter of 1969 (1969.4), 1977.4, 1979.3, 1968.4, 1976.4, and 1980.4 (p.324). A simple regression shows that the changes at the head of the Fed does not account for the real rate changes, but that political shifts do. To strengthen their results, the authors use the global optimization method developed by Bai and Perron (p.325).
These techniques, indicates three important breaks: 1967.1, 1972.4 and 1980.2. These again are not consistent with the Fed chair changes but are quite close to the model predicted by the party changes. Furthermore, a fourth – lesser – shock can be found in 1976.3, which is again correlated to a major political shift (p.326). Another method to estimate the robustness of the results (J-test) points to similar conclusions: “presidential party turnover is highly correlated with real rate shifts and is largely consistent with the breakpoints found with [Bai and Perron] statistical methods”.
… some more
An extension to the data to 1992, should reveal another breakpoint, around 1986.4 as the Republicans lost the Senate. As predicted by the model, another shock occurred in 1986.2 (p.327). The authors “consider these results to be impressive evidence in favor of the proposition that large changes, specifically switches in party control of the presidency or Congress, are significant predictors of real rate shifts” (p.328).
Reassuringly, these breakpoints, once integrated into a regression controlling for macroeconomic variable (money growth, oil price, stock returns) yields strong results indicating that they have a significant impact on real interest rates (p.330). These conclusions imply that “large political changes are good proxies for policy regime shifts in that they correlate well with real rate shifts [and that] models of political influence on macroeconomic policies and outcomes should concentrate more on turnover of elected officials and less on turnover of appointed bureaucrats.”