Several Pieces of “Evidence” (This theory became more creditable when all alternatives were exhausted).  No one came to the RBC theory because of real evidence, or because it was overwhelmingly persuasive.  (This is opposed to the accepted validity of Keynesian economics to correct macro downturns). 

First Criteria: Logical stories of rational behavior models should not include systematic mistakes being made on the part of the rational actors (internal consistency).  Keynesian economics had fooling.  Bob Lucas’ model – “Islands model” allowed for people to be tricked by surprises in the movement of the monetary variables. Even he admitted that his model contained inconsistencies, i.e. that people could just look at the wall street journal and know what direction the macro variables were moving, leading logically to no fooling. 

Second Criteria: The unit root test (Long and Plosner first introduced).  To explain the unit root test: Think about unemployment.  If someone told you that the unemployment rate was 14% and asked for your best forecast in 10-15 years a normal reaction would be to posit that the new rate would be somewhere around 4-6% (the possible natural rate), before you would suggest that the rate would stay at 14%.  This is not unit root.  A unit root would be more like stating that the current growth rate is 3.3% and your best estimate for the future is that the growth rate would stay the same.  An example is to flip a coin to create a time series.  (In class we started at a 2% growth rate and each time the coin was flipped heads meant a rise of 1% and tails meant a fall of 1%).  The result was suitably random looking.  At the end of this process the best prediction for the rate 10 coin flips from now would be the last observed rate, it would be absurd to measure a mean of flips and predict that based on past observation.  This is truly a unit root.  The hypothesis of a unit root could not be rejected in the long piece (when null testing you set up a thesis you would like to prove wrong, thereby giving the alternative higher creditability).  The result of this essay was to suggest that the business cycle could be thought of as the equivalent of a random coin flip.  People stopped thinking that every move in the data was a business cycle move which could be explained by high theory.  This is like the equivalent in basketball, that a shooter could have a “hot hand” and should not be taken out of the game.  A shooter is no more likely to hit a basket after hitting 7 in a row than he was when he had missed 7 in a row.  When this theory came out it forced people to think more about the assumption that the error term on the Robert Solow model could be completely explained.  Real Business cycle theory posits a way to look at this problem even in the presence of randomness.  “When this first came out, I was held captive by trying to explain every little movement in the data with an explanatory variable, Now this randomness is taken for granted.” 

3) Litterman and Weiss (1983).  Vector Autoregression (1983) VAR- No structure (in terms of a hypothesis and model) this was a rather large set of vectors, or matrix, of data useful in talking about the macro economy.  A huge correlation matrix was computed on the data.  This was much like data mining for significant relationships.  One of the big insights is that the real interest rate had no statistically significant relationship with the aggregate measure of money (contrary to theory, and suggesting that money doesn’t matter in determining interest rates).  This contrasts with the theories of the Keynesians and the monetarists.  “If we can’t even find a pattern between money and interest, we really don’t understand what is going on in the macroeconomy.”  McCallen suggested that maybe the interest rate moves before money, meaning that the money wouldn’t be as important in describing the changes in interest rate, this was a changed mindset. 

4) Simulations were offered as “Evidence”  The Solow model is a dependant variable of growth rate as a function of (technology shocks or new ideas, labor, and capital).  Solow’s original decomposition said that three-quarters of the growth variable could be explained by the new ideas variable.  (this leaves labor and capital explaining only a quarter and the residual being counted as all tech).  The tech variable is never measured it is only the residual term for the model which measures capital and labor contribution to explaining growth (the revised models which include better measures of capital and labor do a better job of predicting growth). 

The internet is a perfect example of a productivity shock.

Real Business Cycle – says that Solow was right all the time.  Technology was really the driving force; with some randomness.  Growth (trend component) – take this portion out of the data.  Cyclical part is left.  The residual can be described as the “pure cycle” or the randomness.  The RBC theorists would think this is the worst intellectual move, the cycle is caused by some underlying components of the total data, to divorce these would be to loose explanatory potential.    Solow wanted to keep the decomposition and show accept the random part of the cycle. (TC does not like the Solow model).

Long-Plossner:  How much could, would, did, this number technology measurement change and then output vary?  Katrina (was this a negative shock?) we will see what happens in the evidence, if no negative shock then this provides evidence against.

This model IS complete with the error term (for whatever else bad you can say about it, it is one of the more simple and powerful models in explanation).  We can look at Prescott and the others who defend the RBC model and see them posing the questions for additional microeconomic research.  These programs should focus on further developing the productivity, technology components of the error term in the Solow model.  We should continue to tease any other variables out of the residual.  During the 90’s the work on understanding the technology effect of computers on output could be looked at in the broadest sense as promoting the same program of study as these RBC theorists.  Remember that we are useful in explaining much of history with the business cycle model, looking at agricultural society, we can say that rain is the exogenous shock which then helps determine the labor and capital effects on output, similarly this technology exogenity is to be considered. 

RBC helped to reestablish market clearing rationality, rational expectations (the converse was a formula for more government, like the Keynesian framework).   Now it is the most fundamental of the theories, the building block for discussing the macroeconomy and for the last 25 years has been the newest developing theory in macro.  Two ways of thinking about RBC:  1) enough brainpower applied to these models and we will eventually crack the theory (Prescott) 2) useless b/c there are always going to be phantom variables which we cannot predict the effect of on output (black). 

Big Business cycle countries: Tailand and Argentian (recently).  The U.S. and other developed economies seem relatively unaffected by the business cycle factors.  Big questions include things like: why Japan has experienced slow growth.  In a way the RBC model is the best model that we have for understanding the recent history of the science. 

Any good Business cycle model must contain and explain these 3 things:

1)      Persistence (lasts for awhile)

2)      Co-movement (many sectors go up and down together)

3)      Changes in the labor supply (employment)

To be a serious model you should generate these three things.  Only model that does “almost fit” the 3, but really only satisfies the first two.  Needs to prove.  The 1^st assumption – one person (a dramatic assumption since it assumes away problems like coordination failure).  And output = input.  The example: A farmer.  Two animals, ducks and chickens.  They both lay eggs and the farmer can either consume the eggs or let them grow into additional chickens.  A farmer smoothes consumption of eggs over time.  The technology shock in this case is the decreased fertility of the ducks.  Fewer eggs are available, so that the farmer feels poorer, has a lower flow of outputs.  He will then start eating fewer duck eggs, and more chicken eggs (substituting into chicken eggs).  The farmer lowers total consumption due to lower production.  The process of substitution accounts for the business cycle.  A farmer who “takes a bigger lump up front” does not experience the business cycle, but the farmer who substitutes does.  Models are rigged so that the substitution effect is the dominant effect, the labor supply effect (or the weaker the farmer gets with fewer eggs), is going to be small in comparison (except in Long and Plossner where the cancel out). 

Long and Plossner: Define technology shocks that are big enough to match the actual movement in the data from the real economy, then they model this movement.  This mimicked movement looks remarkably like the real world data due to the embedded propagation mechanism between sectors (2/3 of the time).  Model, however, fails to capture movements in the labor supply (the movement is significantly muted).  There is no accounting for intertemporal substitution, and when comes to labor market this is not a side show (it is a central issue of importance in policy).  Working more when you are paid more does not happen in the real world data. 

Few core facts:

1)      Real world changes in the hours worked are much greater than the changes in productivity.

2)      Hours worked and productivity are not correlated

3)      When you have a downturn it is not that people work less, it is that some people get fired.  These people do not work at all, but the people with jobs still are working about the same amount that they always did.

4)      Growth (for example China) is not actually working the way it would in the model.  People do not work less now because they think that productivity (and therefore wage) will rise in the future; in fact they were hard now.

There is a theory of indivisible labor which tacks on a fixed cost associated with supplying labor.  These include the cost of the commute, the baby sitter, or any other number of forgone things that have facilitated your time at work.  This adds a bit of lumpiness to the cost of labor.  Two cases when the MPL falls:

A)    8 hours becomes 7 hours for all employees across the board

B)     Fire 1/8 of all the people employed at the company

In the case of more fixed costs the employer is going to opt more often for the case of B. 

(people could contract for a lower wage)

Keynesians would complain that this model does not explain the labor market.

Katrina, if it is viewed as a capital shock, we could see a unit root effect where the output stays permanently lower.    We could also imagine a case where the MPK rises such that new capital is brought online.  If it were a labor shock, we would see that this has to be recovered in some way.  If we just think of it as a technology shock than we see the same unit root effect.  This feeds the question:  Should we rebuild New Orleans.  This model’s major failure is to leave out social and moral capital.  All the looting has fundamentally changed the way that N.O. people behave.  Will the same mixture of workers return to the city?  The poor?  The middle class?  What would be the effect of a large government subsidy to the city?

A preview for next week:  Optimal clustering (think about a city in these contexts: space and time).  Cities do exist due to some agglomeration, there is clustering in time (this is why the classroom is empty on the nights and the weekends.