Arnold Kling  

The Macro Doubtbook, Installment 9

Marketing Liberty to Immigrant... Why We Have Freddie and Fannie...

The previous installment was here. This installment discusses expectations.

[UPDATE: This installment happens to contain what I would say is my answer to the question posed today by Tyler Cowen.|


When people make economic decisions, how do they form expectations about the future? In my view, there are two types of markets to consider. First, there are centralized, national financial markets. Second, there are decentralized local markets.

Centralized, national markets include the stock market, the bond market, commodities markets, and derivatives markets. They are centralized in that you know where you need to go to trade in those markets. If I want to buy wheat futures, I go to Chicago. If I want to buy stocks, I go to New York. These markets are national in that what people buy and sell is the same no matter where they live. IBM stock is IBM stock, whether you are in Peoria or Topeka.

Decentralized, local markets include markets for labor, haircuts, and property. In fact, most nonfinancial markets are decentralized and local. The prices and characteristics of goods and services that are exchanged can differ from place to place. When I need a haircut, I do not think in terms of Chicago (say) as the place to go to get it. I do not expect the style and price to be identical regardless of where I live. If I work in marketing communications, I do not expect that the salary and work requirements will be identical working for a cattle feed company in Wyoming and working for a social media startup in Silicon Valley.

Next, let me posit two ways that people can form expectations. These two ways are habit-based and model-based, respectively.

Habit-based expectations involve simple extrapolations of the recent past. If I saw that orange juice sold for $3 a gallon at the grocery yesterday, then I expect it to sell for $3 a gallon next week. If college tuition has been going up about 5 percent every year, then I expect it to go up 5 percent again next year. Among economists, the term "adaptive expectations" is commonly used where I am using habit-based expectations.

Model-based expectations involve coming up with a theory of the determinants of long-run equilibrium and of the process for getting to that equilibrium. For example, I might have a theory of the exchange rate that in the long run the exchange rate will be one at which there is a trade balance, with a process of adjustment that is gradual. If the dollar is overvalued by 15 percent, then I might expect it to depreciate at 3 percent for five years.

What I call model-based expectations is closer to what economists mean by rational expectations. Rational expectations means that the model that market participants are using is in fact an accurate representation of the true processes that determine long-term equilibrium and the path of adjustment. If expectations are rational and there are no surprises, then going forward the price path expected by market participants should coincide with the true path. I use the term model-based expectations to include the case where individuals are forward-looking (not habit-based or adaptive) but might be using the wrong model.

The local-national distinction and the habit-model distinction can be put into a matrix that yields four combinations, as shown below..

Local MarketsNational Markets
Habit-based expectationsPolicy has discretionBubbles
Model-based expectationsGame-theoretic PolicyEfficient markets

Inside the matrix, I have put what I see are the most important characteristics of each configuration. For example, when there are local markets with habit-based expectations, policy has discretion. To put it simply, when workers and firms set wages based on local conditions and habits, then policy makers have the discretion to affect real wage rates. In the textbook model, when the money supply expands, prices rise and real wages fall.

When even local markets have model-based expectations, policy has a more difficult time. We are in a game-theoretic environment, in which market agents choose their strategies on the basis of how they think policy makers are going to behave. This means that workers will take steps to ensure that their real wages are insulated from monetary policy changes. That means trying to anticipate such changes and factoring anticipated monetary policy changes into wage bargains. This line of theory takes us down the road of "rational expectations" macroeconomics, about which we will say more in subsequent chapters. As we will see, this path was considered promising to an entire generation or more of economists, and yet I see little or no value in it.

Turning to national markets in which financial assets are traded, if expectations are habit-based then it is easy to see how bubbles could form. People expect prices to go up because they have been going up, without regard to long-term fundamental economic factors that should determine those prices.

On the other hand, suppose that many investors actually do try to determine the fundamental values for asset prices. If their model is objectively correct, then markets will be strongly efficient. That is, any information that affects the long-term value of the asset will be reflected in its current price.

I believe in a weaker form of market efficiency, as reflected in the phrase There Ain't No Such Thing As A Free Lunch for speculators, meaning that there is no way to easily beat the market using known information. I believe that there are enough investors who are trying to use models to assess fundamental long-term values. However, they do not necessarily know the correct model. If you have a different model, your model may turn out to be correct, in which case you can make a huge profit by betting on your model. But you never know. TANSTAAFL.

Today, we speak of the housing bubble that took place from approximately 2003 through 2006. However, housing falls closer to my definition of local markets than national markets. Because of that, I think it would be more precise to speak of multiple housing bubbles taking place simultaneously.

In general, I think it is most useful to think of expectation formation in local markets as habit-based. I think it is most useful to think of expectation formation in national markets as model-based. This has the following relationship to my views of macroeconomics.

1.I think that the asset markets in which the central bank operates, including the foreign exchange market and the bond market, are model-based. Accordingly, I am skeptical about the ability of central banks to make discretionary changes in these markets. Instead, markets will be measuring central bank targets against feasible long-run equilibrium values. Think of the exchange rate, for example. If the central bank has an exchange rate target that conforms to the feasible value, then speculators will help the central bank hit the target. On the other hand, if the central bank has an exchange rate target that is out of line with the feasible value, then at some point speculators will start betting against the central bank, and if they are correct, then the central bank will lose.
2.I think that habit-based expectations prevail in local markets. However, I do not subscribe to the textbook model of the transmission mechanism for macroeconomic policy, in which policy affects demand, demand affects real wage rates, and real wage rates affect employment and the supply of output. Thus, for me, it is of little or no consequence whether labor market expectations are habit-based or model-based.

Comments and Sharing

CATEGORIES: Macroeconomics

COMMENTS (7 to date)
paul writes:

Arnold, I'm not sure why you are making a distinction in kind. why aren't all expected prices formed rationally? If a storm is about to hit I expect either the bread to be gone or the price to rise. For many things the information set doesn't change that much. even in financial mkts there is probably too much price volatility give the actual change in information.

why the distinction?


Rob writes:

Another thing to keep track of is how institutions and history affect the formation of expectations. A central bank with a good reputation is more likely to be able to "cheat" and realize big gains in employment and output when expanding the money supply. Of course, then they lose their reputation and the old fashioned aggregate supply curve becomes steeper. Also, the credibility of promises from politicians and central bankers seems to be historically important in dampening or exasperating the sacrifice ratio. See Volcker, Paul.

Various writes:

I agree with all that you say. I think your description of these markets is rather elegant and effective. Obviously your matrix is a simplification of the real world, and there are various combinations and permutations that one can derive from your basic model. But I like your model! Thanks

Lord writes:

I would say goods markets are largely national since only the cost of shipment really stands in their path. Sometimes national markets lack models, the tech boom being an example where there are no fundamentals to estimate it by though in the end a return to the old models, earnings, dividends, growth, reasserts itself. Enough investors must follow them though or there won't be sufficient means or courage to overcome momentum investors and the model can change from one of fundamentals to momentum timing where it is not a matter of what you believe but what you believe others believe along with keeping an eye out for when a reversion to fundamentals may take place. Timing is always risky so many fundamental investors will avoid the overpriced rather than try shorting it leading to inefficiency.

While model based national markets may make uncredible bank actions difficult it makes credible ones easier and no one doubts the ability of the bank to create inflation, only the willingness and correctly so. The market will respond only when the bank has convinced them they are serious, and they haven't yet and appear not to feel any need. The bank can't overcome expectations of falling home prices without creating considerable inflation. They can raise asset and commodity prices and lower the exchange rate some of which will leak into demand and jobs.

Hyena writes:

All markets are necessarily "habit based" in that regardless of the model, it requires inputs derived from past realized values on the expectation that the future will stay the same in important respects. Alternatively, all decisions are "model based" precisely because you cannot directly translate past experience into expectation. That requires a model, however simple.

The only differences between "habit based" and "model based" are the level of mathematical complexity, formality and work performed by the actor in each case. The assumption set underlying "habit based" approaches is very small but it's still a model; proof of this is that many of these local markets respond to knowledge about local conditions in predictable, expectant ways. If the Department of Commerce states that housing prices in Los Angeles are expected to fall, individual Angelenos will (if that expectation is sensible) help push the LA housing market that direction.

This outcome is no different than what you seen in the "model based" scenario you provide. In fact, I would wager that the dividing line between your "habit" and "model" based approaches is the return to research. In most cases it just won't do to carefully research your decision: it is senseless to plan your orange juice purchases based on commodity markets, cold-calling farmers in Florida and the Central Valley, using contacts to obtain an inside track, etc. There's not enough there; your savings would be, what, $6 over the course of a year?

Your "model" based trader, however, does all these things because they're moving many years worth of orange juice at a time and the tiny margins are a significant income at that level.

The latter model is more complex because the potential returns justify data collection and hence the acquisition of information necessary to defining the variables in a more complex model. If you aren't going to do the research to drive the model forward, it makes no sense to use a complex model with lots of noisy guessed values. In the end, however, there is no difference in basic behavior whatever and so both sides are at risk (perhaps equally) of bubbles, etc.

Yancey Ward writes:

Tyler's question seemed silly to me. I hoped he asked it rhetorically. A government mandated minimim wage only fools unsuccessful entrepreneurs- if they pay too much for labor, they suffer losses and the jobs disappear (the other way employment can fall due to the price floor). The wage guarantees nothing about the actual output- the government is only guaranteeing the profit of the worker, not the entrepreneur, for the time frame of the job. A government mandated cost of loans can be profitable for a time for the borrower/entrepreneur- the time that the government maintains the price ceiling.

Norman writes:

"[...]I think it would be more precise to speak of multiple housing bubbles taking place simultaneously."

This does seem to be a necessary consequence of your distinction, but I think you create just a big a problem for yourself on the other side. If housing is truly a local good, why would the errors across local markets for it be so highly correlated with each other, but not with other local markets like haircuts or marketing jobs? This doesn't really get us any closer to understanding the relationship between expectations and the housing bubble(s).

Comments for this entry have been closed
Return to top