To determine the appropriate hurdle rate of return for an international opportunistic real estate investor, it is necessary to think in terms of a global asset allocation model.
In the purely U.S. context, investors all look at the appropriate hurdle rate in terms of spreads over the "riskless" U.S. Treasury securities.
In the international context, the global investor looks at the U.S. Treasury securities somewhat differently. The U.S. Treasury security, while riskless in a default sense, now takes on risk in terms of relative currency rate movements. These currency rate movements are, in turn, driven by inflation expectations and current account balances, interest rate differentials, actual and potential political instability and government fiscal and monetary policies, and, ultimately, investor capital flows.
In this global context, risk is defined not only in terms of "default risk" but also in terms of market volatility of returns and the cross-correlation of returns between countries. As a result, global bond investors often will accept a lower current return than is available in U.S. Treasury securities because of the lower "global risk," which includes default risk, currency risk and "cross-correlation portfolio risk reduction."
Extending this global investor perspective to the real estate arena involves the addition of risk variables specific to real estate. In this context, security of property rights and contracts, the stringency of land-use controls, the length of the standard office lease, and the cyclical volatility of the office market all must be considered. The global economic risk and return expectations all are expressed through investor capital flows. Global office capitalization rates, or cash yields, suggest the appropriate current return that an investor will receive on the real estate investment.
It is apparent that cap rates are lower in a number of European and Asian markets than in the United States. This cap rate configuration implies investors place a lower perceived economic and real estate market risk on the investments in many markets compared with the United States. The lower cap may be in part the result of economic risk factors, such as the large U.S. trade deficit and the potential for weakness in the dollar. In addition, global cap rate patterns might be the result of supply-side variables and the perceived risk of a potential oversupply in the U.S. real estate market. In the past, oversupply has led to volatile and, on average, lower total returns in U.S. markets.
Our work attempts to quantify such risk factors and then to produce hurdle rates across countries with those risk factors, building on the efforts by Jones Lang LaSalle Inc. ("Investment Strategy Annual 2000") and by Prudential Securities (Y. Liang and W. McIntosh, "Country Risk Premiums for International Investing," January 2000).
In our approach, we look at an array of variables that may affect hurdle rates of return on real estate, and estimate a two-equation model that attempts to explain both economic and real estate market risk using a subset of these variables. Using this approach, we can approximate total hurdle rates of return under the assumption that variations caused by appreciation are proportional to the variations for income.
To quantify the real estate market risk, and hence hurdle rates, in various international markets, we initially considered an extensive list of variables that we thought had some impact on both economic and real estate market risk. These variables included measures in several classes: demographic, economic, political/legal, financial, the real estate market.
Demographic/economic variables: These included population, population growth, population living in cities and change in population living in cities; share of employment in agriculture, in manufacturing and in services; government spending; gross domestic product growth; inflation; trade surplus; foreign currency reserves; current account index and changes in it; volatility of exchange rates; and the location in the economic cycle.
Political/legal, financial and real estate market variables: Those considered include political stability and changes in political stability; political effectiveness and changes in it; trade policy; security of property rights and contracts; sovereign debt credit rating, and whether it had changed; existence of a corporate debt market; existence of a collateralized mortgage-backed securities market; interest rates; spreads to U.S. Treasury yields; length of the standard office lease; land use constraints; and the location in the real estate cycle.
The economic and financial variables generally had better data availability, primarily through international organizations such as the International Monetary Fund and the World Bank. For the political variables, we relied on information published by the Economist Intelligence Unit and the Fraser Institute.
Data were gathered for 1999 for 45 countries. We also attempted to secure a history of these variables back far enough to include economic conditions unlike those in 1999. But while the economic and interest-rate data are available for earlier years on a consistent basis, the capitalization rate measure used as the dependent variable is not available for more than several years. Because of the lack of a consistent history for both the dependent variable, and many of the independent variables, the analysis is confined to a cross-section study for those 45 countries in 1999.
The first quantitative step was to compute a correlation matrix for all of the variables. There was a high degree of correlation among many of the variables.
For example, GDP per capita and the political effectiveness index have a correlation coefficient of 0.805. The trade policy index and the political effectiveness index have a correlation coefficient of 0.815. As one might guess, the political effectiveness index and the political stability index were almost perfectly correlated, with a correlation coefficient of 0.868. Statistically speaking, many of the factors we initially thought were factors in real estate market risk in fact are measuring the same thing.
As a result, from this initial analytical step, we can anticipate the number of factors that will come to bear in a statistically significant way will be very much smaller than the candidate set of factors we initially considered.
Constructing a model
We take a two-step approach to the problem of trying to quantify real estate market risks across countries. In the first step, we estimate an equation of economic/financial risk. Overall economic/financial risk comes to bear on real estate markets and product markets in general. For example, factors that create economic/financial risk such as a weak current account and a depreciating currency, also flow through to real estate market risk.
As a measure of economic/financial risk, we used the long-dated yield on domestic Treasury bonds as a spread to the long-dated U.S. Treasury bond.
The variables that might have an impact on economic/financial risk are the following: The proportion of employment in agriculture, manufacturing, services, the current account index, government budget deficit as a percent of GDP, volatility of the currency, GDP per capita, inflation rate, the trade balance, trade policy, political effectiveness and the sovereign debt credit rating. For example, the political effectiveness in Russia is relatively low, creating a riskier economic/financial environment than, say, in Germany.
To determine which of these variables contributed to the explanation of economic/financial risk, as we have defined it, we used a stepwise regression technique modified by judgment. The technique adds variables in the order of their statistical importance in explaining the dependent variable, in this case the spread to U.S. long-dated Treasury yields. The first four variables, in order of their significance, are: 1) sovereign debt credit rating, 2) political effectiveness, 3) trade policy index, and 4) volatility of the currency.
In choosing a final equation to explain the cross-country economic/financial risk, we note that the political effectiveness index is highly co-linear with the sovereign debt credit rating. By substituting volatility of exchange rates for political effectiveness, we improve the statistical significance of the other variables with very little sacrifice in goodness-of-fit (R-squared). Although the significance of the exchange-rate volatility variable is compromised (t-value of 0.21), it is no worse than the significance of the political effectiveness variable. In addition, we know from a univariate regression of the Treasury spread on the currency volatility measure that it is a highly significant predictor of economic/financial risk (t-value of 3.16). Furthermore, the overall regression is no worse with the volatility measure, and the other variables are more significant with it. As a result, we choose as a final equation for economic/financial risk one that includes the following as independent variables: 1) the sovereign debt credit rating, 2) the standard deviation of the domestic currency with respect to the dollar, and 3) the trade policy index, which includes as a factor capital controls and constraints on repatriation of capital.
The largest impact on the Treasury yield spread across countries is from the sovereign debt credit rating. This result is as one might anticipate because the credit rating is designed to measure country risk for sovereign borrowing. Given the spread that results from the credit rating, other smaller independent impacts result from exchange rate volatility and trade policy. In part, we think the small impact from these two variables may be because the Moody's credit rating already either explicitly or implicitly accounts for factors like exchange-rate volatility. Certainly a more volatile exchange rate produces risk in the repayment of sovereign debt.
These three variables together explain 60.5% of the variance in the spread. This is a reasonable goodness-of-fit, given the nature of the data and the fact that we are dealing with a cross section data set. The variables in the economic/financial risk equation not only bear up statistically, but they also have structural plausibility.
Were total returns data available, we would use it at this point. However, with a paucity of such data, we turn to the more readily available yields (or capitalization rates). These data capture most of the return to investment in real estate assets, and for this reason, we treat cross-country variation in the cash return, or yield, as approximating the variation in total return.
We approached the estimation of the office market yield equation in much the same way as the economic/financial risk equation. That is, we used a stepwise regression initially on all the candidate variables to explain the variation in real estate yields. The variables that came into the regression in order of statistical significance are as follows: 1) land use constraints index, 2) Treasury spread, 3) GDP per capita, 4) current account index, 5) political stability, and 6) security of property rights and contracts. Equity return was the seventh variable that entered the regression, but its contribution to the goodness-of-fit was virtually nonexistent and its statistical significance was borderline at best.
The final equation explaining office yields across countries included the following variables: 1) land use controls index, 2) Treasury spread measuring economic/financial risk, 3) current account index, and 4) the security of property rights index. This equation explains 71.6% of the cross-country variation in office market yields.
The intercept for the office market yield equation is 11%. This may be thought of as the starting point for calibrating the office market yield to the various risk factors. Economic/financial risk, measured by the Treasury spread, is based on the factors developed above, and it enters the real estate market risk equation. The office yield rises by 18 basis points for each percentage point increase in the Treasury spread. Additionally, the supply side of the market is captured in the land use constraint index. That is an integer index going from 1 through 5, with 5 representing the most stringent constraints. Accordingly, we would expect a negative relationship between that index and the yield. In fact, the relationship is estimated to be negative, with each one-point increment in the index resulting in a 70-basis-point decline in the yield. The current account index, likewise, is an integer index, going from 0 to 4, with 0 being the most favorable. As the current account index increases by one, the yield increases by 60 basis points. Finally, the security of property rights index is an integer index going from 1 to 10, with 10 representing the most secure property rights. Again, we would expect the relationship between this index and the yield to be negative, and the statistics bear out this expectation. For each one-unit increment in the index, the yield declines by 70 basis points. These variables not only have statistical viability in explaining the office yield across countries; they also have structural plausibility.
Calculation of the hurdle rates
Using fitted values from the economic/financial risk equation, we get an estimated Treasury spread, which summarizes the economic and financial risk in each country, given the independent variables (credit rating, exchange rate volatility, and trade policy, including capital flow restrictions). As indicated in the example comparing Indonesia and Germany, using the fitted value for the Treasury spread, we then calculate the office market capitalization rate (yield) for each country. For opportunistic office investment in the United States, we use a target hurdle rate of 20%. We then scale up or down from that hurdle rate based on the relative "fitted" cap rate. As an illustration, if the U.S. fitted cap rate is 8%, and Germany has a fitted cap rate of 6%, then opportunistic hurdle rates adjusted for risk in Germany would be 15% (or 6/8 x 20%). We make this calculation based on the assumption that the income yield (represented by the cap rate) constitutes most of the return, so variations in the cap rate based on variations in risk across countries should be reflected proportionately in the appreciation return as well. If total return data were widely available across countries and property types, we would estimate the equations directly with the total return series rather than the cap rate series.
The estimated hurdle rates are really a spread (or more precisely, a ratio) to the U.S. hurdle rate, which is assumed. Using this two-equation model, the simulated hurdle rates, with the U.S. rate assumed to be 20%, vary from a low of 15.9% for many Western European countries and Britain, to a high of 35.4% for Russia. Thus, the hurdle rates for opportunistic office investments in Western Europe are 19%. Hurdle rates in Latin America tend to be at the upper range of this spectrum.
Bradford H. Dockser is managing director, Europe, at Starwood Capital Group LLC, Greenwich, Conn. Kenneth T. Rosen and Daniel T. Van Dyke are chairman and principal, respectively, at Rosen Consulting Group LLC, Berkeley, Calif.