Skip to main content

Solving Asset Market Riddles

In three papers, Monika Piazzesi and Martin Schneider explore asset market puzzles with the help of heterogeneous and less-than-rational economic actors.

December 1, 2009


Solving Asset Market Riddles

Interrelationships among different asset types—stocks, bonds and real estate—remain something of a mystery to economists. How are their prices correlated? How do optimal allocations vary with expectations about inflation? What explains the dramatic shifts in portfolio holdings in the United States in recent decades? How do beliefs shape markets?

In a December 2007 Region article, “Masters of Illusion,” Fed economists Monika Piazzesi and Martin Schneider, now at Stanford University, explained their mutual interest in exploring the forces that drive these interrelated markets and in expanding asset price models to include real estate—a topic that demands greater scrutiny from economists given the current financial crisis.

“Housing as an asset, and the behavior of its price, is somehow not at the forefront of research in finance and macro,” observed Schneider. But “most asset pricing models do not include housing.” Integrating real estate with other asset markets has therefore been a focus of much of their recent work. As Piazzesi has written elsewhere, “We want to understand the joint behavior of asset prices and quantities of the three major asset classes—bonds, houses and stocks.”

In a series of 2009 Minneapolis Fed staff reports, Piazzesi and Schneider delve into three curious aspects of these markets. The first paper explores household beliefs during the recent boom in U.S. housing markets. The second examines portfolio shifts from stocks into homes, and declines in household wealth relative to real GDP, during the 1970s, along with differing beliefs about inflation. The third seeks to understand the role of expectations and learning in explaining investor behavior in bond markets during the 1980s.

All three papers follow a similar path: They identify an asset market puzzle, statistically confirm the puzzle’s existence and then develop a model with some pattern of subjective belief or “adaptive learning” that helps to explain the puzzle. With this method, the economists push the boundaries of representative agent/rational expectations theory by adding elements of heterogeneous and less-than-rational economic agents.

Momentum traders

In “Momentum Traders in the Housing Markets” (SR 422; also published in the American Economic Review, May 2009), the economists examine data from consumer surveys in the early-to-mid 2000s about housing price trends. They find that there was always a small cluster who believed it was a good time to buy a house and that the size of this “momentum” cluster strongly increased toward the end of the housing price boom.

More specifically, there seemed to be two phases in consumer attitudes about housing prices. From 2002 to 2003, a large fraction of those surveyed believed it was a good time to buy, peaking at 85.2 percent; their most important reason for this belief: “Credit conditions were favorable.”

During a second phase, 2004 to 2005, overall enthusiasm about home buying declined to about 60 percent and views about credit conditions also worsened. But the fraction of those who stated that “house prices are going up” or similar optimistic rationales for home buying more than doubled, from under 10 percent to over 20 percent. (The demographics of this momentum cluster, write the economists, don’t differ significantly from the population as a whole. On the other hand, housing price optimists are more optimistic than average about economic conditions in general.)

But how could such a small cluster of optimistic consumers (just 3 percent of the population) influence house prices, even though they don’t buy a large share of housing stock? Piazzesi and Schneider build what they call a “simple search model of a housing market” where, indeed, this happens. Three features are important for their result: (1) prices are set in a bilateral negotiation, so the price reflects the optimist’s valuation; (2) optimists account for a large share of transactions, so they drive the average transaction price; and (3) transaction costs are high enough to keep content homeowners from flooding the market, thereby keeping trading volume low.

Inflation and assets

In the second paper, “Inflation and the Price of Real Assets” (SR 423), Piazzesi and Schneider seek to understand why the ratio of household wealth to GDP dropped by 25 percent during the 1970s and why price trends in stocks and housing led to a 20 percent portfolio shift from equities to real estate during that same decade. Again, they start by documenting the puzzle, providing data on price and GDP ratio trends in housing, equity and net worth from 1952 to 2003.

To explain the dramatic trends in the 1970s, the economists build an asset market model in which households differ by age and wealth, and also where credit is nominal, meaning that inflation affects bond returns and the cost of borrowing. With this model, plus data on asset prices and holdings, as well as data on household inflation expectations, they develop a plausible explanation for both the drop in wealth and the shift toward housing.

The drop in household wealth relative to GDP, they show, was due to two unique events in the 1970s that reduced the propensity to save: First, young baby boomers entered asset markets, immediately lowering average savings rates. Second, unexpected inflation eroded bond portfolios, lowering financial wealth and thereby reducing saving.

The shift from stocks to housing was due to three factors, with the Great Inflation of the 1970s as a key influence. First, higher inflation expectations led to lower predicted stock returns, so investors looked to housing instead. Second, disagreement about real interest rates (young households shifted their inflation expectations more quickly than old households) led to more borrowing and lending among households and an increase in collateral prices, namely, housing. And lastly, changes in inflation expectations made housing more attractive than stocks because of capital gains taxes on stocks and deductibility of mortgage interest. Taken together, these three factors (about 50 percent attributed to lower predicted stock returns and one quarter to each of the other causes) explain the asset portfolio shift.

Bonds and learning

The third paper, “Trend and Cycle in Bond Premia” (SR 424), examines another anomaly in asset markets. Sophisticated investors in the past 50 years “could have made a fortune,” Piazzesi and Schneider write, by borrowing short-term funds and investing in long-term bonds whenever they observed a large spread between long- and short-term Treasury interest rates (right after recessions, for example) or when-ever the overall level of the yield curve was high (especially during the early 1980s). But while publicly available interest rate data clearly reveal these statistical regularities, investors never exploited the profitable opportunities. Why not?

Most economists have hypothesized that investors’ assessment of risk (either perceiving that objective risk had increased or increasing their personal risk aversion) changed over time, leading them to shun the investment opportunity—however lucrative—as too risky. But another possibility is that investors simply didn’t recognize the pattern seen so clearly with the benefit of (massive databases and) hindsight.

Piazzesi and Schneider evaluate both explanations by first looking at survey data on interest rate and inflation forecasts and then building an asset pricing model that measures the explanatory power of both hypotheses. In their scrutiny of survey data, they find that “both candidate reasons for predictability patterns are important.”

Their model then seeks to incorporate both and measure their relative importance. To do so, the model accounts for “adaptive learning”—the idea that economic actors learn over time, reacting to past and current information to form expectations about the future. They find that this model “can help understand the movements in both components.”

Because adaptive learners react slowly to new information, argue the economists, they don’t change their interest rate forecasts quickly in response to sudden, sharp rate changes. Thus, adaptive learning “provides a reason for systematic differences between statistical forecasts and survey forecasts.”

Also, adaptive learning “gives rise to changes in perceived risk” and, given difficulties in the 1970s when high inflation wiped out much bond wealth, “adaptive learners viewed bonds as particularly unattractive around 1980,” demanding high risk premia.

The economists thus substantiate the idea that subjective beliefs, as measured through consumer surveys, can help explain what conventional asset price models have long viewed as puzzling.


Douglas Clement
Editor, The Region

Douglas Clement is a managing editor at the Minneapolis Fed, where he writes about research conducted by economists and other scholars associated with the Minneapolis Fed and interviews prominent economists.