Chapter dzǵ: Tacit Preachments are the Worst Kind ȁȃȂcommonly employed, “Granger-causality” tests, and particularly vector-
autoregression studies, for the way they handle correlations among the
“independent” variables, because of processing of the data (prewhiten-
ing, trend removal, and other purifications of time-series data in ways
that throw away some of the association that may exist), and because
these methods are testing for specific (e.g., linear or log-linear) relations
and rigid relations among the variables, whereas money exerts its effects
with “long and variable lags.” Ļe filtering techniques employed remove
much of the cyclical movements in money, and monetary influences are
masked by innovations in interest rates, in turn reflecting monetary pol-
icy. (Michael Bordo, editor of the volume, adds that observation.) Ļe
VARtechnique for dealing with spurious correlation eliminates impor-
tant monetary changes. By removing all serial and cross correlations from
economic series,VARin effect removes all but short-run blips in money,
losing the influence of relatively sustained monetary changes that do tend
to affect business activity.
Cagan also criticizes the treatment of money’s endogeneity. (Mone-
tarists know that connections between monetary changes and business
activity can run and evidently have run in both directions.) If the Federal
Reserve could override the endogeneity of money and thereby make out-
put behave differently than it behaves in fact, then money does count.
Ļere is a difference between being endogenous with no independent
effect and a mutual dependence which policy can affect. Ļose who deny
monetary effects on output may be aware of this point but continue to
neglect it. Even if money had been in some sense completely endogenous
inȀȈȁȈ–ȀȈȂȂ, the Federal Reserve could have overridden that endogeneity
and saved the economy from devastation.
I can only raise, not answer, a few further questions about supposed
econometric evidence. Is it really informative to run correlations with
time-series figures taken not only from periods of cyclical or “abnormal”
change in output, money, prices, and so forth but also from periods of
steadiness or relatively steady growth (or relatively undisturbed money-
supply-and-demand relations), as if all these figures, taken indiscrim-
inately, constituted observations on a single universe? In other words,
can one really examine and compare the effects of monetary and non-
monetary disturbances by jumbling together numbers from periods both
experiencing and not experiencing such disturbances? Ļe issue is not
really what calculated parameters describe ill-defined average-over-time
relations among various macroeconomic variables. Ļe issue, instead, is