http://www3.gmu.edu/departments/economics/bcaplan
Econ 816
Spring, 2000
Week 5: Monetary vs. Fiscal Policy in Real Models
I. Macro Policy in Real Models
A. Recall: "Real models" includes both:
1. RBC models
2. New Keynesian models with real rigidities
B. In these models, the economy's response to tradition monetary and fiscal policy is usually the same. For the sake of convenience, I will use "RBC models" and "real models" interchangeably until the end of the discussion (when the complexities real rigidities add will be addressed).
II. Three Versions of RBC
A. Since RBC seems so intuitively implausible to most people, we begin by discussing the empirical evidence for three RBC or RBC-like business cycle mechanisms.
B. Mechanism #1: Supply shocks. (Hamilton 1983 + others)
1. Major findings: 7 of 8 recessions 48-72 preceded by "dramatic rise in price of crude."
2. Very hard to find anything that predicts these dramatic rises. (Oil "Granger causes" real activity, but nothing Granger causes oil price movements).
3. 70's recessions plausibly connected to oil shocks as well. "[P]ost-OPEC world has more in common with its predecessor than many might suppose."
4. A drop in the bucket: Using annual 48-98 data, (or 73-98!) I find it hard to identify any connection between unemployment and nominal or real price of oil. More evidence in monthly data, although again controlling for a few other factors makes a relation hard to find.
C. Mechanism #2: Sectoral shifts. (Lilien 1982 + others)
1. Claim: Observed fluctations in unemployment are to a significant extent fluctuations of the natural rate rather than deviations from the natural rate. When optimal allocation of labor shifts, it takes time for labor to move, even with market-clearing. This raises the natural rate. "Unemployment is simply the time it takes workers displaced from contracting firms to find employment in expanding firms. The quantity of unemployment generated by shifts of employment demand will thus depend on the speed with which workers find new jobs."
2. E.g.: Move from manufacturing to services. Claim is that both employed and unemployed manufacturing employees will tend to be unhappy compared to before the decline in demand, but there is no reason for the unemployed to envy the employed.
3. Lilien finds support for SS for 1948-1980. Debate continues; probably most telling is that job vacancies are negatively correlated with unemployment, not positively as the theory predicts.
D. Mechanism #3: Seasonal cycles. (Barsky and Miron 1989)
1. Most macro data are "seasonally adjusted." What can be learned about business cycles by failing to adjust?
2. Doing regressions of percentage changes in key macro variables on four seasonal/quarterly dummies (with constant omitted) shows interesting pattern. Note:
|
Variable |
1st Quarter |
2nd |
3rd |
4th |
|
real GNP |
-8.08 |
3.72 |
-.49 |
4.85 |
|
price level |
-.19 |
.14 |
.19 |
-.14 |
|
unemployment |
1.08 |
-.67 |
-.17 |
-.24 |
|
nominal money |
-1.05 |
-.82 |
.13 |
1.74 |
3. Monetary base more sensitive than broader money supply. (Note this for later discussion of endogenous movement of the money supply).
4. Sectoral decomposition shows that Christmas matters more than weather, but both matter.
5. Rebuttal to "RBC is intuitively implausible." Weather would explain seasonal fluctuations in agricultural economy. But you wouldn't expect big seasonal fluctuations in a modern, non-agricultural economy. But: there they are.
6. Note procyclicality of labor productivity. Barsky and Miron see this as support for conventional "labor hoarding" storing rather than intertemporal labor supply.
7. Implication for "misperceptions" theories: Seasonal cycles are predictable, but still seem to matter for employment.
8. Overall: Simple but challenging findings. Argument from "parsimony."
III. Monetary Policy in Real Models
A. In real models, money is neutral; nominal movements don't cause real movements. You can essentially figure out what the real sector is doing first, then use your answers to find out what happens in nominal terms.
B. Caveat: Inflation is strictly speaking not neutral even in real models! Inflation raises nominal interest rates, which induces a shift from money holding and transacting to non-money substitutes.
1. In practice, though, this caveat is probably not too important - though it would be if inflation got high. (Note the massive rise in "shoe leather costs" during hyperinflations).
C. Puzzle: Empirically, there is a nominal/real correlation. Isn't this a decisive blow to real theories?
D. Answer: No, the money supply could be endogenous. Perhaps real activity causes changes in the money supply rather than the other way around. Two possibilities:
1. Endogeneity of the base.
2. Endogeneity of the money multiplier.
IV. King and Plosser's Model of Endogenous Money.
A. Early stages of RBC had no explanation for observed real-monetary correlation. K-P argue that this can be handled by thinking of monetary services as a intermediate good that is complementary with real economic activity.
B. The model (we’ll just look at K-P’s simplified version): 2 productive sectors, 1 intermediate good, 1 final good.
1. Goods industry: output is either consumed or used as input for future production.
2. Financial industry: output is intermediate good - transactions services. Transactions services economize on time and other resources needed to exchange goods.
C. Transactions cost minimization reveals:
1. Demand for both (real) currency and financial services is directly proportional to output.
2. Real currency demand decreases as the nominal interest rate rises, and increases as the price of financial services and the opportunity cost of time rise.
3. Demand for financial services rises as the nominal interest rate rises, and falls as the price of financial services and the opportunity cost of time rise
D. Price level depends positively on: high-powered money (currency + bank reserves) and nominal interest rates; it depends negatively on output, price of financial services and labor, and real bank reserves. (Increase in real reserve demand holding high-powered money constant reduces price level).
E. Therefore: Real activity (including real deposit services and real deposits) is neutral wrt base money changes. Since K-P assume that banks are always exactly at the legal minimum RR, fluctuations in deposits have to be the result of variations in the public’s currency/deposit ratio. (Does this make sense? Couldn’t K-P have gotten more plausible results if they allowed for excess reserves?)
F. K-P on the money-output relation: First, a correlation between output and broad money is predicted by the model, so it’s not surprising. Second, a correlation between output and base money is not predicted, but not impossible either depending on the central bank’s feed-back rule.
G. Evaluation: If you look carefully, you will discover that K-P don't prove what they are usually supposed to have proven! They do show that the real money supply (c+gd) is proportional to real output. But they fail to show that the nominal money supply P(c+gd) has any necessary connection to real output. In particular, in RBC models the price level will normally be contracyclical, so P(c+gd) could easily be contracyclical.
V. Fiscal Policy in Real Models (Barro 1981; Aschauer 1988)
A. There is no theoretical difficulty in explaining how fiscal policy affects output in real models. But the mechanism is not the traditional Keynesian one.
B. Two margins to think about:
1. How much to work: The higher the productivity of labor, the greater the real wage and the greater the incentive to work. (Caveat: Backward-bending labor supply).
2. When to work: It pays to work more (and consume less) in periods with high real interest rates.
C. Spending shocks: temporary vs. permanent. (Barro 1981)
1. [Note: Assuming Ricardian equivalence and lump-sum taxation!]
2. In general, spending shocks raise the demand for current output, increasing real interest rates and total real output.
3. For temporary shocks, it seems like the supply of real output should be more interest-elastic than for permanent shocks. Intuitively, if the real interest rate is temporarily high, even people who won't work more overall will work more now.
D. Tax shocks: temporary vs. permanent.
1. With lump-sum taxation and Ricardian equivalence, taxation is just a negative wealth shock. Since leisure is a normal good, greater lump-sum taxation implies an increase in labor supply!
2. Insofar as the lifetime wealth effect of temporary increase is less than permanent increase, the labor supply response to a permanent tax increase is greater than to a temporary one.
3. Distortionary taxation tends to reverse the (counter-intuitive?) conclusion that taxation spurs work. With distortionary taxation, the real after-tax wage falls; labor supply acts as if multi-factor productivity were lower.
4. Similarly, with distortionary taxation, the labor supply response to a temporary tax hike exceeds the response to a permanent one. People just enjoy leisure and wait for taxes to fall back to their standard level.
VI. Does RBC Make Sense?
A. Even sophisticated RBC models of monetary policy like K-P have a lot of trouble fitting the basic facts. Aside from the impact of money growth on nominal interest rates, money simply doesn't matter for these theories.
B. RBC fiscal policy channels are similarly problematic.
1. Is labor supply in general actually sensitive to temporary changes in the real wage?
2. Is labor supply actually sensitive to changes in the real interest rate?
C. Consider Braun and McGrattan (1993), which tries to explain the U.S. and U.K. experiences during the world wars in an RBC framework.
D. Monetary policy barely mentioned at all. No effort made to connect rapid money supply growth to the real economy's behavior, even though (for the U.S.):
|
Year |
M2 growth |
Year |
M2 growth |
|
1913 |
3.3% |
1939 |
8.5% |
|
1914 |
3.0% |
1940 |
10.4% |
|
1915 |
13.8% |
1941 |
9.1% |
|
1916 |
16.7% |
1942 |
19.8% |
|
1917 |
12.5% |
1943 |
23.8% |
|
1918 |
11.2% |
1944 |
16.1% |
|
1919 |
14.8% |
1945 |
15.6% |
|
1920 |
3.8% |
1946 |
7.1% |
|
1921 |
-6.0% |
1947 |
4.3% |
E. Fiscal expansion used to explain all unusual movement of real variables. But this differs completely from the dinosaur Keynesian version of "G creates booms."
F. How can labor productivity go up? If labor supplied is responding to high real interest rates, or if the labor supply curve shifts due to the negative wealth effect, then observed labor productivity should fall. Braun and McGrattan appeal to GOPO, which as the commentators point out, is highly implausible.
G. Is this a reductio ad absurdum of RBC?
VII. RBC and the Theory of Unemployment
A. What does "market-clearing" even mean for labor markets? Presumably it doesn't mean that zero labor is devoted to searching for better job matches.
B. This makes search theory a natural complement of RBC; note sectoral shifts take on RBC.
1. Interesting feature of many search models: Unemployment rate is inefficiently low! Employer search costs decrease as the total number of searching workers rises.
2. Unemployment would on the other hand tend to be too high if the search costs of individual workers increase as the total number of searching workers rises.
C. Bigger puzzle for RBC: Why does the number of people employed vary along with the total number of employee-hours? If everyone is on their labor supply curve, why doesn't everyone cut back their hours slightly? In reality, it is much more common for a few workers to be laid off, while the rest continue to work full time.
D. Rogerson (1988): Just add a fixed cost of working.
1. In equilibrium without lotteries, U(w+r)-m=U(r), where w is the wage, r is the rental price of capital, and m is the fixed cost of working. Clearly those who don't work consume less.
2. The trade: All workers enter a lottery (usually interpreted as unemployment insurance). Whoever wins gets to quit - saving the fixed costs of working - and then receives a normal salary courtesy of the losers. The losers are the people who remain employed!
a) Question: How does this differ from a "labor cartel"?
3. This is (expectationally) Pareto-improving. Not surprising when you look at the allocations (c, n, k) in the example:
a) Normal: Workers: (1.107,1,1); Non-workers: (.31, 0, 1)
b) Lottery: Workers: (.63, 1, 1); Non-workers: (.63, 0, 1)
4. Moral hazard makes this precise contract difficult, so in practice such a lottery would give less than perfect insurance.
5. What about the opposite sort of moral hazard: What prevents the unemployed from re-entering the labor market if markets clear and work is better than the dole?
E. Extending Rogerson. Risk-aversion drives the preceding results, and indivisible labor supply is merely assumed.
1. Suppose that the disutility of worker looks like: K+h2. This implies a U-shaped AC of working. This creates gains from trade vis-a-vis a situation where everyone cuts back their hours proportionately.
2. But without risk-aversion, it seems that the uninsured result of free competition would be to set w=minimum AC.
F. General problem with Rogerson-type models: Don't psychological factors and/or stigma make the "fixed cost of working" low or negative?
VIII. From RBC Rigidities to New Keynesian Real Rigidities
A. Returning to Lucas (1978), adding imperfections to an RBC model opens up the possibility of efficiency-enhancing intervention.
B. But will these interventions resemble traditional macro policy? No, they will look much more like Pigovian corrections for externalities. E.g. subsidy to unemployment to correct for search externality.
C. Adding real rigidities to RBC models makes them difficult to distinguish from NK models with real rigidities. E.g. Subsidies to industries where monitoring is costly to compensate for the social inefficiencies of "efficiency wages."
D. Note further that the sign of the optimal Pigovian correction depends on the model. In e.g. the efficiency wage model, a tax on unemployment is efficiency enhancing.
E. Do actual patterns of taxes and subsidies exhibit any plausible connection to macroeconomic concerns?
F. Almost all of the complaints about RBC models can thus be almost automatically applied to NK models with real rigidities.