Prof. Bryan Caplan

bcaplan@gmu.edu

http://www3.gmu.edu/departments/economics/bcaplan

Econ 345

Fall, 1998

Week 12: Simple Econometrics and Monetarism
  1. Monetarism vs. Keynesianism
    1. Monetarists and Keynesians both accepted the AS-AD framework, and agreed that AS was not vertical (in the short to medium run).
    2. The disagreement: What shifts AD?
    3. In other words: What determines Nominal GDP?
    4. Three possible positions:
      1. Position #1: Only fiscal policy matters (early Keynesians)
      2. Position #2: Both fiscal and monetary policy matters (later Keynesians and some monetarists)
      3. Position #3: Only monetary policy matters (other monetarists, + the modern consensus?)
    5. Most modern policy-makers act as if #3 is true. Monetary policy is now almost the sole tool of stabilization. But is this for theoretical or practical reasons?
  2. GDP and The Equation of Exchange
    1. Definition of GDP: GDP=C+I+G (plus NX; we'll ignore that).
    2. Most important monetarist identity: P*Real GDP=M*V
    3. P is the price level (so P*Real GDP=Nominal GDP).
    4. M is the money supply (preferred measure: M2)
    5. V is the velocity of money.
      1. Mathematically, it is simply defined to make the equation true.
      2. Economically, it has a natural interpretation as the percentage of income people hold as money. If my annual income is $40,000, and I typically have $10,000 in M2, my personal velocity is 4.
    6. Note: Accounting identities do NOT prove causation, EVER.
  3. Velocity and Monetarism
    1. Question: If Position #3 is true, what can be said about velocity?
    2. Answer: It must be constant.
    3. Question: If Position #2 (or #1) is true, what can be said about velocity?
    4. Answer: It must be a positive function of fiscal policy.
    5. Empirical Evidence:
      1. If you regress velocity on government spending as a fraction of GDP, you get a significantly negative coefficient.
      2. If you regress percent change in velocity on percent change in Nominal G, the coefficient is insignificant.
    6. Bottom line: No obvious impact of fiscal policy on velocity. This tends to support Position #3.
  4. Nominal GDP, Nominal G, and M2
    1. Several ways to check for the impact of fiscal and monetary policy on Nominal GDP.
    2. Method #1: Try regressing Nominal GDP on Nominal G and M2.
    3. Result: Only Nominal G seems significant; M2 doesn't seem significant at all. This is consistent with Position #1.
      1. Adding a trend doesn't change matters.
    4. Method #2: What if you re-do, using percent changes instead of levels? Also, try adding 1 lag for each independent variable. Regress Percent Change in Nominal GDP on % change in G, lagged % change in G, % change in M2, and lagged % change in M2.
    5. Result: The only clearly significant variable is lagged % change in M2. Lagged % change in G is insigificant, and has a negative sign; % change in G is (barely) insigificant, and has a positive sign. This is consistent with Position #3, although with a small sample size you might still think that Position #2 is right.
  5. Nominal GDP, G/GDP, and M2
    1. Does it make sense to use Nominal G as the measure of government spending? Maybe G just has built-in inflation adjustments.
    2. Suppose that the government just spends 20% of income on G. Then regressing NGDP on G will give a perfect correlation, even though it isn't causal.
    3. But: If this scenario is correct, then regressing NGDP on G divided by NGDP will give zero correlation. Let's try it.
    4. Method #3: Regress NGDP on M2, G/NGDP, and a trend.
    5. Result: Only money matters! The sign on G/NGDP is insigificant and negative. (The trend is also insignificant).
    6. This suggests that it is a mistake to use Nominal G as a regressor, and provides some more support for Position #3.
  6. Real GDP and Fiscal and Monetary Policy
    1. Whatever influences Nominal GDP should also influence Real GDP in the same direction. Does it?
    2. Method #1: Regress % change in Real GDP on % change in nominal G and % change in M2.
    3. Results: Neither G nor M2 seem to matter. G's sign is negative; M2's sign is positive.
    4. Method #2: Add first lags of the prior regression's independent variables.
    5. Results: Only lagged M2 matters (again!). Negative signs on G and lagged G, but not significant.
    6. Again supports Position #3.
  7. Conclusion
    1. The evidence is somewhat mixed, but overall there is a surprising amount of support for Position #3: only money matters for nominal or real GDP.
    2. Money seems to work with a 1-year lag.
    3. Government spending, if anything, often seems to depress real GDP. This is particularly striking since G is a definitional component of GDP.