Labor Income Uncertainty, Residential Real Estate, and Asset Allocation

Thus far, our asset allocation review has focused on financial assets. This month, we're going to take an in-depth look at how two non-financial assets can have a substantial impact on this process. The first of these is your future labor income. To convert this into an asset, you can discount your expected future labor income back to its present value, which you can think of as your "human capital." The second important non-financial asset is, for many people, the value of the home they own. In the next two articles, we will start by looking at what theory says about how each of these should affect the allocation of financial assets. We will then move on to what different empirical studies have found. Finally, we will summarize the implications of the theory and the data for how people should allocate the financial assets in their portfolio to achieve their long term goals. Before we begin, we should mention one important caveat: the academic study of these issues is still, relatively speaking, in its infancy. While our results are quite interesting, you shouldn't regard anything that follows as the final word on these subjects. There is still much to learn, and we will continue to research and report on new findings in these areas as they become available.

The Impact of Human Capital on Financial Asset Allocation

As we just noted, you can convert annual labor income (a flow) into human capital (an asset) by discounting it to its present value. The present value of your human capital therefore depends on a number of different variables. The first is your current expectation for the size of your annual future labor income flows. All else being equal, the higher your expected future annual labor income, the higher the present value of your human capital. The second consideration is how many remaining years of labor income you expect to receive. All else being equal, the more remaining years of income, the higher the value of your human capital. The third variable is the riskiness of your labor income. This is a function not only of the different factors that directly affect it (e.g., global, national and regional/local economic conditions, conditions in your industry and company, and personal factors, such as your ability to hold your tongue in the face of clueless comments by your boss…), but also of the flexibility you have to supplement it (e.g., by working a second job). All else being equal, the less risky your labor income, the higher the value of your human capital.

To get a better understanding of how risky labor income actually is, we looked at national accounts data from the United Kingdom and the United States. In the former, between 1971 and 2001, the average annual nominal change in compensation of employees was 9.95% per year (based on annualized quarterly data), with a standard deviation of only 3.20% per year. The United States data covered 1971 to 2002, and was broken down into three categories: wages and salaries paid to private sector employees, wages and salaries paid to public sector employees, and proprietors’ income for non-farm businesses (essentially, self-employment income). The average annual rates of change, and associated standard deviations, are shown in the following table, along with comparable data for the returns and risk for domestic bonds and equities over the same period.

This table makes two interesting points. First, labor income is less risky that domestic bonds or equities. Second, within the broad category of labor income, public sector employment is the least risky, while owning one’s own business is the most risky.

Now let's look at the theoretical linkage between the value of your human capital and the allocation of your financial portfolio between different asset classes. Assuming people seek to avoid big swings in their current consumption of goods and services, the higher the value of your human capital, the more risk you can afford to take in your financial portfolio (because temporary setbacks in the value of the latter will be much less likely to affect your current consumption).

In their paper "Investing Retirement Wealth: A Lifecycle Model", Campbell, Cocco, Gomes, and Maenhout note the impact of aging on the labor income/financial asset allocation relationship: "A typical individual starts adult life with little financial wealth; initially, as labor income increases, human wealth may grow faster than financial wealth, but fairly early in adult life financial wealth starts to accumulate faster than the present value of remaining labor income. This implies that most younger investors with relatively safe labor income should concentrate their portfolios heavily in equities, and gradually shift toward less risky investments as they approach retirement." It also implies that younger investors with relatively riskier labor income should allocate less of their investment portfolio to riskier asset classes.

These authors also add detail to the relationship between labor income risk and financial portfolio risk: "the [current] theoretical literature on [the relationship between labor income and financial asset allocation] can loosely be summarized as follows…If labor income is [low risk], then [holdings of low risk financial assets] are crowded out and the household will tilt its portfolio strongly toward [higher risk] assets…If labor income is risky, but uncorrelated with [the returns on] risky financial assets, then [low risk] financial asset holdings are still crowded out, but less strongly; and the portfolio tilt toward [higher risk] assets is reduced. And if labor income is positively correlated with risky financial assets, then [higher risk] assets can be crowded out, tilting the portfolio toward [lower risk] financial assets."

A critical empirical issue is therefore the extent to which changes in labor income are correlated with returns on different asset classes. In the past, the standard theoretical models assumed a relatively strong positive correlation between changes in labor income and changes in the rate of return on a broad equity index. This conclusion was based on the observation that many investors held relatively low percentages of their financial portfolios in higher risk asset classes such as equities. However, recent research has found that this assumption was incorrect.

The following table shows the correlations between nominal changes in different types of labor income in the United States between 1971 and 2002 and nominal returns on different asset classes.

The analysis presented in this table shows that changes in labor income generally have very weak correlations with the returns on different asset classes. It also suggests that, if one were trying to hedge changes in one’s labor income, one might want to consider the use of domestic or foreign bonds. However, given that labor income appears to be much less risky than financial assets the need for this seems to be minimal. However, this analysis is also based on aggregate level figures. Could people in different occupations face a higher degree of labor income risk?

In their paper "Occupation Level Income Shocks and Asset Returns", authors Davis and Willen examined changes in labor income across ten different U.S. occupational groups, and found that none had a statistically significant relationship with U.S. equity market returns. Moreover, they also note that "several [other] studies that consider a variety of countries, time periods, and income components [also] find zero or small correlations between aggregate equity returns and the value of human capital." As a result, they conclude that "the market equity portfolio has modest value as a hedge instrument for the average investor's labor income." However, a later study by Campbell, Cocco, Gomes and Maenhout found that the relationship between labor income and equity market returns was positive for people who are self-employed, and for college graduates (who are more likely to hold positions where compensation is partly tied to the performance of a company’s stock). They concluded that "privately owned business risk is an important substitute for stock market risk." In other words, all else being equal, you would expect to find a self-employed person or someone whose compensation was tied to the performance of her company’s stock holding a lower percentage of risky assets in his or her portfolio than someone with an equivalent amount of labor income from less risky sources.

Finally, Davis and Willen also found stronger correlations between changes in labor income and returns on industry level equity indexes. This suggests two possibilities for hedging labor income risk. You could maintain a permanent short position in the equity index for the industry in which you work, or, alternatively, one could tilt one’s investment portfolio toward asset classes with returns that have historically been negatively correlated with returns on your industry’s equity index. Practically, it is very difficult for most investors to continuously maintain a short position in a stock index (either directly or via put options). Given this, we took a closer look at the second strategy. In the following table, we show the correlation of nominal returns (we used nominal since labor income is denominated in nominal currency units) between ten different sector indexes and the equity market as a whole. We also show the three asset classes (including other equity indexes) with the highest negative correlation with each equity index.

This table makes two points. First, to the extent that labor income is positively correlated with the returns on an industry equity index, the equity market as a whole does not provide a good hedging vehicle. In fact, in every case but one when the industry index (and also, we assume in this case, labor income) declines, the aggregate equity market index will likely do the same. Second, it would seem that real return bonds, foreign bonds, residential real estate, and, to a lesser extent, other asset classes provide opportunities for hedging some amount of labor income risk in different industries.

There are, however, two important caveats to these conclusions.

Most importantly, they are based on a relatively short data set (the broad-based Dow Jones industry sector index returns have only been available since 1992), which makes them tentative at best. The second caveat is that the crucial link in this argument between labor income and the return on sector equity indexes is subject to some uncertainty. For example, Davis and Willen noted that the sign of the relationship between the two (which one would normally expect to be positive) was sometimes found to be negative. As an example of the latter, these authors cited "the introduction of labor saving technology, [which] may generate higher returns on industry equity index, but lower earnings for workers." They concluded that "the usefulness of industry level equity portfolios as hedging instruments for [labor income] is…an empirical issue… which no single study can definitively settle."

The Impact of Owner-Occupied Housing on Financial Asset Allocation

A recent report in Australia ("A Primer on a Proposal for Global Housing Finance Reform" by Caplin and Joye) succinctly summarized the current situation with respect to the treatment of residential real estate in most asset allocation analyses: "When Australians plan for retirement, they must consider how a disparate collection of assets combine to form the overall portfolio of 'household wealth.' Yet surprisingly one of their most valuable assets is often excluded from these planning decisions entirely -- the owner occupied home…One risky and highly illiquid asset dominates almost all holdings in most household portfolios: owner-occupied housing. To cast this into relief, suppose your financial planner proposed this investment strategy: You place between 60% and 80% of all you wealth in the stock of a single company for a minimum of, say, five to ten years. The company makes one product and has one plant. It operates in an industry which has grown rapidly and which is favored by federal tax laws. This industry is, however, also characterized by large fluctuations in value, arising as a result of both macroeconomic and idiosyncratic conditions, high transaction costs, and extremely low liquidity. You would probably decline the opportunity to embark on such a risky endeavor. Nevertheless, most people make exactly this commitment when they decide to acquire their own home. And while dwellers can insure against damage from natural disasters, it is almost impossible to insulate oneself from market meltdowns or regional economic swings."

Moreover, "the home ownership experience evolves strikingly over the course of the life-cycle. When young, families scramble to scrape together funds for a down payment so that they can graduate from the difficulties of rental accommodation to the supposed nirvana of owner occupation. This period of intense saving often induces a considerable consumption squeeze and may severely constrain lifestyle choices. In the middle years, dwellers frequently struggle against the specter of relentless mortgage payments -- the so called "house poor." Finally, in later life many manage to pay off all their debts and live in the home clear and free. Unfortunately, by this time retirement beckons and the majority of households have precious little income other than their pensions. They are now 'asset rich, but cash poor." [Indeed, for most homeowners aged 60 and above, the dwelling represents more than 80% of total non-pension assets]. Upon death, most leave their homes as (unintended) bequests to heirs and/or the public authorities."

In a subsequent paper ("Innovative Approaches to Reducing the Costs of Home Ownership, Volume 1", a Report published by the Menzies Research Center in June, 2003 for the Prime Minister's Home Ownership Task Force by Caplin, Joye, Butt, Glaeser, and Kuczynski), the authors noted that, "regrettably, the risk properties of residential real estate receive very little attention…Two key facts regarding house price risk are worthy of comment. First is its scale. By any reasonable measure, real estate risk is of immense importance to the typical homeowner. Secondly, this hazard is multifaceted: there is no single statistic that adequately summarizes it. Instead, to appreciate the many dimensions of housing risk, we must tell a complex story. In a nutshell: (a) Property prices are volatile and positively related to labor income; (b) Most families are highly leveraged to real estate; and (c) The dwelling is often the dominant asset in the household's portfolio…It is commonly accepted that house price risk is not at all well understood by academics. The small number of studies that do exist tend to conclude that fluctuations in the real value of real estate constitutes a serious economic threat to the average household's standard of living. Furthermore, this hazard is exacerbated by three intertwined factors: (1) The indivisibility of the dwelling asset, which compels home owners to bind together their consumption and investment decisions; (2) The high proportion of wealth that is, as a direct result, held in the form of housing, and (3) The absence of instruments that would enable occupiers to hedge the financial risks associated with this investment."

The main reasons residential real estate is so difficult to incorporate into an asset allocation analysis is because it differs from traditional financial assets in so many ways. To begin with, it is both a consumption good and an investment good. In the former sense, it replaces rental payments (and hedges against their future increase), while also providing intangible benefits such as social status display (e.g., the "McMansion" phenomenon) and an enhanced sense of security (see for example, the paper "A New Kind of Gold?" by Trimbath and Montoya of the Milken Institute).

As an investment good, residential real estate has identifiable risk and return characteristics. Unfortunately, these are not easy to measure. Virtually all published residential real estate indexes are based on price changes, rather than the total return calculations one finds in the case of financial assets. The reason for this is easy to understand: while house price changes are (relatively) easy to measure, the "dividend" or "annual income payment" on this asset is not. For example, any such calculation should include not only the value of avoided rent, but also maintenance expense, tax benefits, and possibly some valuation of the intangible benefits of home ownership cited above. To say the least, the measurement challenges involved are daunting.

Four other factors further contribute to the difficulty of comparing residential real estate to financial assets. Unlike financial assets, most residential real estate is bought on margin -- that is, using a combination of mortgage debt and owners equity. Closely related to this is the unique tax treatment given to residential real estate in many countries (e.g., mortgage interest and property tax deductibility, the exclusion of avoided rent from taxable income, special capital gains tax, treatment, etc.). On top of this, residential real estate is a relatively inefficient market: transaction costs are very high in the residential real estate market (e.g., the typical brokerage commission in the United States, including search costs, has been estimated at ten percent of the transaction's value), no two assets are alike, and informationally, the market is not very efficient (e.g., individual houses do not sell very often, and real estate agents business practices are regulated more loosely than those of security salespersons). Finally, owner-occupied housing is, so far, an indivisible asset: once you have purchased a house, you cannot sell off a portion of your equity interest in order to achieve a better risk/return trade off in your overall asset portfolio.
Because of these unique characteristics, a number of authors have suggested that in practice, people choose the level of real estate that is optimal from a consumption point of view, and thereby "back in" to their allocation to it as an investment. The net result is usually an allocation to residential real estate that is too high in terms of asset allocation theory (see, for example, "Owner Occupied Housing and the Composition of the Household Portfolio" by Flavin and Yamashita). Some authors have estimated that this over-allocation to residential real estate reduces net worth at retirement by as much as fifteen percent or more, because the heightened risk caused by mortgage borrowing reduces the allocation to risky assets such as equity. Most often, substantial investment in risky financial assets (and the benefits of their potentially higher compound returns) is deferred until mortgages have been paid down (see, for example, "Home Ownership as a Constraint on Asset Allocation" by Cauley, Pavolov, and Schwartz).

This last point raises obvious and important questions about the nature of residential real estate as an asset class. The authors of the limited number of studies which have been done all seem to agree, despite differences in the data and methodologies they use, that the returns on residential real estate have a relatively low correlation with the returns on domestic bonds and equities (which potentially makes this asset class an attractive addition to a portfolio). The following table summarizes the findings on correlation from a number of recent studies.

Unfortunately, it is much more difficult to compare across different studies the estimated returns and risk on residential real estate as an asset class, because they all use different methodologies to estimate these statistics (e.g., some impute an annual income value, while others just measure price changes). As a result, we have conducted our own analysis, using data from the United States.

Our residential real estate return data are based on the quarterly price change index produced by the Office of Federal Housing Enterprise Oversight since 1975. This index is based on repeat sales of the same property over time, which has some theoretical advantages over different approaches to constructing a residential real estate index (for more on different index approaches, see "Prices of Single Family Homes Since 1970:New Indexes for Four Cities" by Case and Shiller). We have not made any adjustments to the quarterly returns to reflect imputed net income. Rather, we have assumed that depreciation is exactly offset by maintenance spending, and the total consumption benefit is equal to the mortgage payment plus tax benefits. As a result, we assume that the investment return on residential real estate is entirely due to price changes (a conservative assumption that probably understates the returns on this asset class). We have, however, made two adjustments to the national level data to make them better reflect the situation faced by most investors

First, because the OFHEO data are unleveraged, we have assumed an average mortgage debt level of forty percent over the holding period of the home. Second, we have had to adjust the national level data to reflect the fact that the risk faced by an individual homeowner (like an investor in a single company's stock) is much higher than that faced by an investor in a broadly diversified market index. A number of studies suggest that this difference in risk (as measured by the standard deviation of returns) is quite significant in the case of housing. A number of studies have found that the correlation of residential real estate returns between different cities is quite low. For example, "The Portfolio Implications of Homeownership", by Eicholtz, Koedijk, and de Roon compared returns between 1980 and 1997 in Los Angeles, San Francisco, Chicago, New York and Boston. The highest correlation between residential real estate returns in these cities was only .32, between Los Angeles and San Francisco. Case and Shiller reached the same conclusion in "Prices of Single Family Homes Since 1970", which compared Atlanta, Chicago, Dallas and San Francisco, as did Chinloy and Cho in "Housing Returns and Restrictions on Diversification," which was also based on city comparisons.

It is clear that the risk faced by an investor whose allocation to residential real estate takes the form of a single home is very different from an investor who owns a well diversified aggregate index for this asset class. As Caplin and Joye note (in "A Primer on a Propsoal for Global Housing Finance Reform"), "the volatility of a broad real estate return series is likely to materially underestimate the true level of risk at the individual home owner level. The underlying prices used to estimate representative returns are averages, which one would expect to realize only when holding a well-diversified portfolio of property. In reality, households own a single residence, which is subject to far greater idiosyncratic price variability (a good analogy is the difference between owning a broad based equity index versus stock in a single company). Case and Shiller recommend multiplying aggregate housing market risk (defined as the standard deviation of returns) by a factor equal to the square root of five to approximate the homeowner's exposure…Using Swedish data, Englund, Hwang and Quigley concluded that one should multiply regional data by a factor equal to the square root of 6." In our analysis, we have followed the latter recommendation, and multiplied our national level standard deviation of returns by 2.45 (which is approximately the square root of six) to obtain a risk/return assumption for residential real estate which approximates the actual situation faced by a typical investor.

Based on these adjustments, and using quarterly OFHEO data covering 1975 Q2 to 2002 Q4, we estimated the real return on U.S. residential real estate to be 4.85% per year, with a standard deviation of 7.57%. This is quite close to the 4.7% average annual real return and 8.1% standard deviation estimate for Australia in the Caplin and Joye paper, which uses a similar methodology. It is also close to estimates for the Euro area (4.3% return, 3.6% standard deviation for the aggregate market; 8.82% for a single house using our adjustment approach) and the United States (4.6% return, 5.39% for a single house) developed for the 1982-2001 period in another recent study ("The Role of Wealth in the Economy: The 2002 Annual Meeting Papers of the Royal Netherlands Economic Association"). However, it should also be noted that this study found that in the U.K., both residential property returns (7.9%) and the standard deviation for a single house (20.8% using our approach) were considerably higher than in other countries.

Before moving on to the correlation of residential real estate returns with those on other asset classes, two other points are in order with respect to return and risk. First, in countries (such as the U.S.) where taxes are not indexed to inflation, and where real estate receives special benefits, its relative return (compared to other asset classes) tends to improve when inflation is high. In effect, high inflation increases the effective tax rate on other assets, while reducing it on residential real estate (see, for example, "Inflation, Income Taxes, and Owner-Occupied Housing" by James Poterba). Given that inflation was declining during much of the sample period covered by the OFHEO data (1975 to 2002), our returns estimate may understate the attractiveness of real estate as an asset class. That being said, on the risk side of the equation we have to guard against a tendency to see housing as a one-way bet on which you can't lose money. For example, in their Australian study, Caplin, Joye, Butt, Glaeser and Kuczynski compared 40,650 same house repeat sales in the states of New South Wales, Queensland and Victoria between 1984 and 2002, and calculated the real returns realized by the sellers. In the bottom 25 percent of this distribution, they found that the median seller suffered a real loss of (10.9%). In the bottom 10 percent, it was even worse, with a median real loss of (20.5%). In contrast, the median for the overall distribution was a gain of 14.2%, while the top ten percent of the distribution realized a median real gain of 140.6%. Unfortunately, when it comes to housing as an investment, too many people focus the last number, and ignore, or are unaware, of the possibility of a significant real loss.
The correlation of real returns between residential real estate and other asset classes is shown in the following table. Please note that it is based on quarterly data from 1988 to 2002, which is less accurate than the monthly data we usually use in our analyses.

Like other studies, we also find a low level of correlation between real returns on residential real estate and those on other asset classes.

The next step in our analysis was an examination of the benefits of residential real estate as an asset class from the perspective of a U.S. dollar based investor (unfortunately, we did not have comparable residential real estate data series for other currencies). Our first approach was to see how the inclusion of residential real estate changed the composition of the minimum variance portfolio. As you may recall from last month’s issue, the MVP is the combination of asset weights which minimizes the portfolio standard deviation (that is, it minimizes risk) for a given set of asset classes. We looked at residential real estate from three perspectives. First, we considered the impact of the asset class as a whole, using the lower 3.09% standard deviation for the aggregate index. Next, we considered the impact of residential real estate from a homeowner’s point of view, and used the higher 7.57% real standard deviation. Finally, we calculated the MVP from the perspective of an investor who has already invested fifty percent of her or his assets in residential real estate, but still wants to minimize the riskiness of her or his financial asset portfolio – a not uncommon situation. The following table shows these three minimum variance portfolios, as well as the MVP without residential real estate.

This table makes four important points. First, as an asset class (column two), residential real estate is very attractive, and can bring significant diversification (risk reduction) benefits to a portfolio. However, the diversification benefits associated with an investment in a specific house (column three) are far lower than those associated with an investment (were it possible) in the asset class as a whole. Third, as shown in column four, the inseparability of the consumption and investment benefits associated with purchasing a house often forces people to over-invest in residential real estate, and involuntarily raise the riskiness of the overall asset portfolio. Finally, it is interesting to note how the inclusion of residential real estate in the Minimum Variance Portfolio has the greatest impact on holdings of real return bonds. This seems to be in line with most people’s intuition that residential real estate is a good hedge against inflation.

The next step in our analysis was to include residential real estate in our 3%, 5%, and 7% real target return portfolios, and rerun our simulation optimization analyses for them. As was the case for the Minimum Variance Portfolios, we did this for (1) residential real estate as an asset class, (2) for a specific house, and (3) for an investor who has already committed fifty percent of his or her assets to a specific house. In the first two cases, we allowed the investor to rebalance her portfolio annually to her target asset weights. In the third case, the investor starts off with fifty percent of his assets in his house. While his financial assets are rebalanced annually to their target weights, housing is not (in other words, when it comes to housing, our investor follows a simple "buy and hold" strategy).

The results of this analysis mirror those in the MVP analysis, with declining percentages of each portfolio allocated to real estate as we moved from case (1) to case (2). However, as case (3) is potentially the most interesting for many of our readers, we will present it in more detail below. As we have done previously, we ran different simulation optimizations using assumptions based on both historical data and future estimates for the returns on each asset class. For residential real estate, we followed the approach used in other studies, and used the returns on the asset class as a whole, and a higher standard deviation to reflect the higher risk associated with owning a specific house (obviously, given the low correlations of returns between different cities – and houses within cities -- found in other studies, this assumption is at best a very rough approximation of the actual situation faced by a homeowner). Finally, as we have done before, we combined the two portfolios using a weight of .67 for the one based on historical data, and .33 for the one based on future return assumptions. The results are presented below:

In an interesting twist on this argument, Eicholtz, Koedijk, and de Roon (in their paper "The Portfolio Implications of Homeownership") approached the question from the opposite direction, and asked what level of net annual consumption benefit would be necessary (given the expected price change based rate of return) to justify the high proportion of household assets invested in housing. Where fifty percent of assets are invested in residential real estate, the breakeven net consumption benefit was ten percent per year (note that their study only included domestic bonds and stocks as the other assets in the household portfolio). Unfortunately, no researcher that we know of has been able to estimate whether the net tangible and intangible benefits of homeownership approach this level! So for now, we’ll stick with the conclusion that most people are over-invested in residential property.

Beyond this, the tables also raise another interesting question. Many studies have shown that as the percentage of assets invested in residential real estate increases, people tend to hold a greater percentage of low risk assets (e.g., domestic bonds) in their financial portfolios. Conversely, as mortgages are paid down, the proportion of the financial portfolio invested in risky assets increases (before decreasing again as retirement approaches). However, as shown above, our simulation optimization analysis suggests that this may not be the best approach, and that many people probably could improve their chances of achieving a given target rate of return by holding a different mix of assets.

We believe there are two alternative explanations for why people do not hold a more optimal mix of assets. The first is that people are simply unaware of the fact that they could improve their expected long -term returns by changing the asset allocation in their portfolios. As we said at the beginning of this article, life-cycle asset allocation is a topic still in its infancy. Most asset allocation articles written to date have been, and still are, based on the single-period mean/variance optimization approach. And very few of them include labor income or residential real estate (much less both at the same time).

However, another explanation also comes to mind: the missing variable may be people’s inaccurate perception of the riskiness of their labor income. Our guess is that the logic (implicit or explicit) used by many people is similar to the following: "I own a house with a mortgage. If the economy encounters bad times, not only will equity and housing market returns decline, but so too may my labor income (out of which I must make my mortgage payments). I therefore should invest my financial portfolio in relatively low risk assets like short term bonds and bank deposits." As we have noted, the limited number of studies which have been done have generally found that most people’s labor income has relatively low risk (measured in terms of the standard deviation of its average percentage change from year to year). Hence, most people’s need for investments to hedge their labor income should be low. Second, it is not clear that a downturn in labor income should automatically result in a downturn in housing values. At the aggregate level, our analysis has shown that, rather than having the strong positive correlation assumed by the conventional wisdom, residential real estate itself may be an effective hedging vehicle for labor income. The recent boom in U.S. mortgage refinancings to provide funds for consumption spending seems to lend support to this point. On the other hand, if one works for a company that heavily depends on local market conditions, then a shock to the local economy could result in reductions in both labor income and local housing values. This point really seems to depend on specific circumstances. (For further discussion of how labor income, residential property, and financial assets are interrelated, see "Hedging House Price Risk in Incomplete Markets" by Joao Cocco).

We should also mention a much-discussed financial market innovation that has the potential to radically change the situation we have just described. The underlying objective is to separate the housing and investment aspects of home ownership. This would be accomplished through the establishment of a mechanism that would enable homeowners to sell off a portion of the equity in their house to special purpose funds which would pool these interests and sell index fund-type securities whose risk/return profile matched that of the aggregate residential real estate asset class. The paper "Innovative Approaches to Reducing the Costs of Home Ownership, Volume 1", by Caplin, Joye, Butt, Glaeser, and Kuczynski provides an excellent overview of the how such a market in "housing partnerships" might work, as does "Household Asset Portfolios and the Reform of the Housing Finance Market" (TIAA-CREF Research Dialogue #55).

Based on our analysis, the potential benefits from the introduction of this type of security would be very, very large indeed. However, with one exception, we have yet to see the introduction of these vehicles for making housing equity liquid and transferable. The exception is in the U.K., where instruments have been introduced that allow one to hedge exposure to changes in London area house price indexes from two to seven quarters forward in time. To date, these instruments have not proved to be as popular as their sponsors had hoped. Many observers have attributed this to the short time horizons involved, and noted that, as homeowners are long term investors, they would be far more interested in longer term instruments.

So where does this leave us? We have seen that an expanded view of the asset allocation problem includes not only your financial assets, but also your labor income and residential real estate. We have also seen that, all else being equal, the present value of your future labor income (i.e., your human capital) will decline as you approach retirement. Assuming economic theory is correct, and people maximize their satisfaction by minimizing the ups and down in their consumption over time, this leads to the conclusion that one’s investment portfolio should gradually be shifted toward less risky assets as one grows closer to retirement.

The more interesting questions involve the interrelationship of labor income, housing, and financial assets. Here, the accurate perception of the riskiness of one’s future labor income appears to be critical. In general, for a given set of long term financial goals, the riskier your labor income, the less risky your asset portfolio should be. However, most people’s labor income is, in a statistical sense, probably much less risky than they assume it to be. Consequently, they have the capacity to hold a higher percentage of relatively riskier assets than they actually have in their portfolios. Moreover, even when your labor income is riskier than average, this still does not automatically mean that you should avoid all apparently "risky" financial asset classes. Keep in mind that it is the relationship between portfolio (not asset class) risk and labor income that matters. However, also keep in mind that the potential results from using financial and housing assets to hedge labor income risk are far from guaranteed. The correlations between asset class returns and different types of labor income are generally relatively weak.

So far this seems relatively straightforward. As we have seen, it is our taste for consuming housing that tends to get our asset portfolios in trouble. In short, our consumption desires lead many of us to allocate more of our assets to a single house (usually on a leveraged basis) than is prudent from an investment point of view. In the medium term, the solution to this problem is surely the separation of housing consumption from housing investment. Unfortunately, while much discussed, the housing partnership securities that would do this are not yet in existence. In the short term, we must take another route.

One alternative is to reduce our consumption of housing (and therefore the portion of our asset portfolio we invest in it). This would seem to be particularly appropriate when a person’s labor income is closely tied to economic conditions in the same region in which their house is located.

A second alternative is to offset some of the investment side affects of our housing consumption by paying careful attention to the way we allocate the assets in our financial portfolio. This requires us to overcome a certain amount of conventional wisdom, such as the false notion that real housing (price) returns are strongly correlated with real bond and equity returns.

However, we suspect that most people are actually using a third alternative: mentally separate your house from your financial portfolio, and count only on the latter to finance your retirement. Explicitly or implicitly, this also means that you are treating your house as a potential bequest to some combination of charity, your heirs, and the tax authorities.

Objectively, none of these three strategies is preferable to the others -- it really comes down to a matter of personal choice. Our strong view, however, is that such choices should be made consciously, and with full understanding of the other alternatives that have been rejected. Unfortunately, this is too often not the case. We hope this article will help to change that.

| Equity Market Valuation Update | Global Asset Class Returns | Updated Economic Outlook | Model Portfolio Update | Labor Income Uncertainty, Residential Real Estate, and Asset Allocation |