The decline of rental completions in the U.S. housing market: 1970-1994.
The purpose of this paper is to build a model and to develop some null hypotheses to explain the performance of the U.S. rental housing market during the 1970-1994 period. A recent HUD report entitled America's Renters Need, (1995) underscored the need to explain that decline by factors such as tax policies, capital and credit market crises, and the low Federal Housing Administration (FHA) participation. The major premise on the tax side is that tax shelters provided by the Economic Recovery Tax Act (ERTA) of 1981 for investment in rental properties were reduced or even turned off by The Tax Reform Act (TRA) of 1986. On the capital and credit sides, we see decline in loans due to the Depository Institutions and Monetary Control Act of 1980 (DIDMCA) that eliminated interest rate ceilings on mortgage loans among other things, and the Garn-St. German Act of 1982 that allowed saving institutions to invest in areas other than mortgages. Loans outstanding at S&Ls fell by more than 20 percent between the second quarter of 1989 and first quarter of 1991. [Bernanke and Lown (1991) p. 208.] On this point the Bush administration responded with new risk-based capital requirement by passing the Financial Institution Reform, Recovery, and Enforcement Act (FIRREA) of 1989 to rid the financial institutions of moral hazard and adverse selection problems. Finally, on the FHA side we see that multifamily mortgage has declined since the 1980s, from 30 percent in 1980 to 16 percent in the mid-1980s, and to 3 percent in 1992.
The tax, capital and credit, and FHA issues are debatable. If ERTA caused an increase in unintended multifamily inventory, then rents and vacancy rates should have time to fall to a normal level before multifamily financing is resuscitated. However, when speculative builders anticipate future profit in a soft market, they will continue to build for a time, allowing the vacancy rate to increase and disengaging the demand constraint. [Rick (1972)]. Similarly, the history of capital and credit teaches us that depository institutions will innovate around regulations as Hubbard (1994, Ch. 15) predicted. Therefore, it is only a matter of time that a gap financing caused by the demise of the S&Ls will be filled. A hitch in the gap-filling adjustment process is the perception that regulators are overzealous, but according to Bernanke and Lown (1991) p. 218, it will be ". . . hard to determine whether regulators are 'excessively' tough." Finally, FHA's participation did not always target a higher homeowner ratio, or privatization over the sample period. Policies have varied over the period depending on the state of the economy, and even corruption.
The overall problem with the HUD report is that it lacks scientific corroboration. The purpose of this paper is to fill that gap. We will examine those hypotheses along with others relating to vacancy and rational expectation issues during the sample period. The paper is divided into a survey of the literature section, a model section that integrates several theoretical works, and a statistical section that confronts several null-hypotheses to the data for that period.
2. Survey of the Literature
The guidelines for FHA housing market analysis is contained in the FHA Techniques of Housing Market Analysis, 1970. However, the approach there is engineering rather than econometric. The econometric base for housing models was set by Tinbergen (1968), Klein (1950 p. 93), and others as summarized by Fair (1972) and Brady (1973). Maisel's (1963, 1965) model gives an equilibrium point of view; Fair's (May 1972), a disequilibrium point of view; Smith' s (1972, 1974) and Rosen and Smith (1983), a vacancy indicator point of view of the housing market.
We will set out explicitly the information we borrowed from those models in the Model and Specification sections below. Particularly, we borrowed the equilibrium concept from Arrow and Hahn (1971), McAllister (1990), Kreps (1977), and Anderson and Sonnenschein (1982). Consumer maximization for the rental market is modeled after Arias and Arnott (1991). Excess credit concepts are from Ricks (1972), and rational expectation concepts are from Fair (1979), de Leeuw and McKelvey (1984), Lovell (1986), Maddala (1988), and Lucas (1976).
3. The Model
The following propositions and corollaries are accepted in the literature. We begin with two definitions to be amplified in the statistical section.
Definition I. (Variables): x = services provided by a housing units, e = endowment, p = prices, Z(p) = excess demand at that price, V = market vacancy rate, VL = vacancy level, RM = mortgage rate, D is demand (occupied units), S is supply (Occupied + vacant units).
Definition II. (Constants): [Beta] is an agent intertemporal discount factor, Vn = natural vacancy rate, assumed to be fixed at 5 percent.
PROPOSITION I: [Equilibrium State: Rosen and Smith (1983), Maisel (1963, 1965)]: Assume a competitive rental market. Let D and S be demand and supply quantity vectors that depend on some predetermined variables (Xes). Then the vacancy rate will be a decreasing function of excess demand in the rental market.
COROLLARY to PROPOSITION I: [Smith (1974)] Ceteris Paribus, if the normal vacancy rate is a constant, and S is fixed for simplification, then the market vacancy rate and rent are inversely related.
PROPOSITION II. [Disequilibrium State: Fair (1972)]: Let price be inflexible; i.e., fail to adjust each period to equate D and S. Let output be Q = min(D,S), then market failure will occur, and nonprice rationing will be necessary. In the housing market, [Delta]RM = RM, - [RM.sub.t] - [RM.sub.t-1] = f(D - S) will adjust to clear the market.
PROPOSITION III. (Our Model): Given two states of the economy, [Omega] = ([[Omega].sub.1], [[Omega].sub.2]), corresponding to equilibrium and dis-equilibrium, the rental market will be more frequently in one state than in the other for the timeframe of this study, and depending on the nature of the constraint facing renters in their intertemporal decision making, the optimal rental stock will increase, decrease, or stabilize.
COROLLARY to PROPOSITION III: [Rational Expectation Implications (RE)] If renters believe that the equilibrium (or disequilibrium) model is the true model and know how to solve it. If renters have the same expectation of exogenous variables such as information values, and the error term of their model. If those forecasted beliefs and expectation solve their current maximization problem based on past and current information, then the model has some rational expectation implications.
4. Statistical Specification of the Model
We have formulated the following four equations model to test the above hypotheses for the rental market. Among their overall novelties are that they confront the Fair and Maisel models with annual data instead of the monthly and quarterly data of the original models, respectively. Also data availability, lags, and single vs. system of equations models will be determined empirically as recommended by Brady (1973). For instance, Brady results gave a superior performance of the single over the system approach. Since our system approach integrates the rental model within the entire partial equilibrium housing market, we would care to see if Brady's result is validated.
[Mathematical Expression Omitted] (1)
[Mathematical Expression Omitted] (2)
[Mathematical Expression Omitted] (3)
[Mathematical Expression Omitted] (4)
[HC.sup.DD] = New Privately Owned Housing Units Completed (In Thousands).
[HC.sup.SS] = New Privately Owned Housing Units Completed (In Thousands).
[Delta]HH = Change in households.
WH = Average weekly hours in the construction industry.
[Mathematical Expression Omitted] = Previous period housing stock (In Thousands)
T = Time trend variable.
RM = Yield of FHA mortgages in the Secondary Market.
/[Delta]RM/= + [Delta]RM if [Delta]RM [greater than or equal to] 0, or zero otherwise.
\[Delta]RM\ = - [Delta]RM if [Delta]RM [less than or equal to] 0, or zero otherwise.
[Delta]SF = Change in deposits at Savings and Loans Associations and Mutual Savings Bank ($Billions).
DHF = Change in Advances of the Federal Home Loan Bank to Savings and Loan Associations ($Billions).
MC5 = New Privately Owned Housing Units Completed in Structure with 5 units or more (In Thousands).
FHA5 = FHA's share of Residential Mortgage Origination: Five unit Buildings and greater (In $Billions).
MTGORIG5 = Residential Mortgage Origination: Five-unit Buildings and greater (In $Billions).
Rem = Removals from the housing stock.
TAX = A dummy variable with values of one in 1986 onwards.
TB = Treasury Bills rate (one year).
V = (Excess Vacancy rate in structure with 5 units or more).
AV = Trend -(Available vacant units for sale + Available vacant units for rent).
F = A dummy variable with values of one in 1989 onwards (Representing FIRREA).
R = Median Asking Rent.
RI = Rent component of the BLS Consumers Price Index.
C = Residential cost component of the GNP implicit price index.
4.1 Fair's System
The first two equations represent supply of and demand for completions. Fair (1972) has argued that in the disequilibrium state, the mortgage rate is the appropriate variable of focus, as rents tend to be sticky. We followed him in making the parameters [[Beta].sub.5] and [[Gamma].sub.6] of that variable equal in order to bring out the disequilibrium characteristics of the housing market. However, our model differs in several respects from Fair's. We substituted production for starts because production is the premise of our hypothesis. Fair's seasonal and monthly construction workday variables are not relevant to our annual model. Instead, we add a work-hour variable. Also, we introduced the Tax and FIRREA dummies in both of Fair's equations as they affect both the single and multi-family markets. The other variables remain faithful to Fair's original specification.
With Fair, we anticipate negative price and housing stock coefficients and positive work-hours, deposits and loan advances coefficients. The coefficients of the dummy variables should all be negative to reflect erosion of the tax advantage, and the tightening of capital availability. Also with Fair, we wish for the time coefficient to be positive on the demand side and negative on the supply side in order to help identify the equations, given the same dependent variable in each equation.
4.2 Our Rental Specification
Eq. (3) is our specification for the rental market. Our dependent variable is rental production. While a time trend identifies supply and demand equations, captures demographic influences in Fair's model and avoids erratic behavior of removals in Maisel's model (1963) p. 366, it examines the hypothesis of a decline in completions in our model. While a tax variable in Smith's (1974), and Eubank's and Sirmans's (1979) model captures operating expenses, it relates to regulatory reform in our model. While the change-in-rent variable is a haul over from the traditional specification, it is a surrogate for Maisel's rent index (1963) p. 372. in our model.
For variables shared with the Fair's specification, we expect the same signs, but differences in magnitude, to reflect the different time dimensions. We expect the coefficient of FHA share in mortgage lending activities to be positive, because multifamily production should rise or fall with the presence or absence of FHA participation. We expect multifamily production to vary in the same direction with change in rent-to-cost ratio because high rents usually imply that the market is tight, ceteris paribus. Finally, whenever the vacancy rate is above the normal level, as is observed frequently for the sample period, its coefficient should be negative, because excess vacancy is one of the symptoms of a soft market.
4.3 Maisel's Specification
Eq. (4) is Maisel's reduced form specification. We expect with Maisel that the coefficients of the Treasury Bill rate, vacancy, and three period lagged dependent variable will be negative. The other coefficients are expected to be positive. As the demolition data are poor, we adopted a percentage of inventory definition by taking a half of a percent of the annual stock as a surrogate measure. Maisel (1965) p. 191 advocated a smaller percent plus a constant; viz., 21.4 + .001 x Available housing units or 21,100 + 0.1 percent x available [stock.sub.t-1]. Maisel (1963) p. 383. However, our definition gives a better result, and is used as a norm in FHA type of market analysis.
4.4 Other Specifications
We need some distinctions on how to empirically examine the tax hypothesis. One approach would be to follow Rosen (1989) in examining the level of increases in rents and apartment prices to see whether the benefits of ERTA were actually eroded. Rosen argued that rents will have to increase by 19 percent, or apartment prices fall by 16 percent to level out the tax benefit of ERTA. Another approach might be to follow the "typical project model" approach of Ling (1992). A third suggestion by Aim and Follain (1994) is to build a structural model that draws upon discrete-time discounted cash flow. We find that the approach recommended by Richard Florida (1986) easily fits into our framework. Florida's (1986) p. 454 suggested that a dummy of one in 1980 alone might be enough to capture the surge in mortgage financing. However, it would not test the effect of ERTA diffusing beyond its counteracting period initiated by TRA. The latter would require a dummy variable equal to one in the post-1981 period as well. The in-between alternative is to specify ones for the 1981-1986 period. Finally, to capture the effect that TRA has reversed the momentum of ERTA, a dummy of ones from 1986 onwards would be necessary. In the results section below, we found that only the latter specification gives the best result.
For the FHA hypothesis, we can pick a date as the starting point of FHA's low or nonparticipation in the rental market. Alternatively, we can assess whether FHA's participation enhances or diminishes the level of rental completions. The former will need only a dummy variable and an F-test to check the differences between sub-periods. However, to identify such a breakpoint is ambiguous at best. Besides, FHA surveys of mortgage lending activities are available since 1970, and to ignore such generally available data will be deliberately reducing the information content of our model. We therefore will follow the latter course, where a significant FHA coefficient will assess whether FHA's participation and policies were appropriate for the time-frame of the study. The coefficient level will explain how much impact variation in FHA's participation had on completions, while its sign will show the direction of change that will enable comparative static analysis.
The vacancy hypothesis requires the concept of a normal vacancy rate, which we take to be 5 percent. If the actual exceeds the normal level more often in the sample period, we expect its relationship with completions to be inverse. We allow for symmetry between excess vacancy that depresses construction, and tight vacancy that encourages construction. The side that predominates will determine the sign of the coefficient. Maisel's model requires that vacant units vary around a trend, which we will estimate for the 1970-1994 sample period. Fair (1972) p. 213 suggested that because of the way starts are defined, only a change in vacancy and not its trend is relevant in the Maisel model. While that is true for the identity equation of housing starts to hold, it did not capture significant variations of completions in the model.
Finally, on the capital and credit side, we used a dummy variable to measure the effect of FIRREA. That variable will take values of one for 1989 onwards and zeroes otherwise, and lags will be determined empirically.
4.5 Rational Expectation (RE) Implications
We can draw some inferences of RE from the rents (R, and RI), housing units completed (HC), interest yield (RM) in FHA mortgages in the secondary market, and 1-year TB rate (TB) variables. Maddala (1988) proposed that lag variables can substitute for expected values. Rather than arguing that case, we will be testing for RE behavior based on our empirically determined lag structures. On the rental side, Smith (1972, 1974) has expectation entering more in the distributed lag tradition, using an Almon one period lag. Since we use Smith's model only to keep the causation of rent and vacancy straight, and not as an explicit equation in the model, this expectation quality is only inferential. In the Maisel specification (Eq. 4), the housing stock variable appears on both sides of the equation, and with lags. In our model (Eq. 3), R, RI and RM enter in their first difference form. In addition, RM takes on a one period lag form.
In Fair's specification (Eqs. 1 and 2), the rate of interest is based on the concept of expectation as against preferred habitat or permanent income hypotheses, and therefore has current and past rates in its term structure or yield curve solution. While the TB variable does not take on a lag dimension, it is coupled with the long-term secondary market rate, which together gives partial information on term structure. If the expected term structure hypothesis is adopted, we can make some implications of RE behavior by the path iterative method suggested by Fair (1979). However, it is not likely that the procedure will yield a unique RM solution that exactly solves the expectation model, and therefore, it has only some implications for RE.
We should also infer RE behavior to the extent that agents allow the government and other institutions to form their expectation. Tax reforms, credit crunch, and FHA decisions are based on full information, and therefore, can be considered rational. If agents know the policy rules of those institutions, then that information will affect their expectations. Agents do not follow mechanical rule but only their belief. Broadly speaking, if they believe that institutions will follow a right credit policy, low FHA participations and fiscal incentive, then they will expect low supply of production. The long-term declining trend in rental completions indicates that policy measures have been in place for a long time in order to have a permanent downward trend outlook.
4.6 Data Sources
The data sources for the Fair model are taken from the same sources as Fair's. We tabulate the sources as follows:
Variable Source Deposit and FHA Rates. Federal Reserve Bulletin. Office of Thrift Supervision. DRI., Inc., Historical Tables, 1992 and other issues. Weekly Earnings. Employment and Earnings., U.S. Department of Labor, BLS, October 1995, p. 45. Completions, median rents U.S. Housing Market and Housing data. Conditions, February 1995 mainly, put out quarterly by HUD. Vacancies for 5+ units. Statistical Abstract of the U.S. Rent index. Business Statistics 1961-1988; current from the Internet. U.S. Department of Commerce, BEA. Costs data. Economic Report of the President, 1995.
After the demise of the S&L institutions, the disaggregated deposit variables for Mutual Savings Bank were not available from 1983 onwards. The Office of Thrift Supervision published data on SAFI insured savings banks up to 1989. Surrogate data for time deposits and savings for both thrift and commercial banks are also available from DRI. Inc., "Historical Tables" (1992 and other issues), and the Statistical Abstract of U.S. Savings data from the Federal Reserve Bulletin supplemented with updates from the Internet are the ones that yielded the best results across all the models we estimated.
In Maisel's system, the TB rate is from DRI sources (1992, and more recent issues). The 6-months and 1-year rates were examined, and only the one year rate yielded significant results. Some household and available vacant data were not reported. Data for the years 1971, 1972, 1982, 1984, 1986, 1988, and 1992 were estimated. Missing data for the inventory was interpolated first using the formula: [inventory.sub.t-1] + [starts.sub.t] - [removals.sub.t]. We defined removals as 0.5 percent of the base period inventory, although Maisel advocated a slightly different methodology as described above. Missing available vacant units were interpolated
using the formula: [AV.sub.t] = Inventory x [AV.sub.t-1]/[Inventory.sub.t-1].
The vacant available units are next deviated from its average change for the 25 periods, which we estimated as 72.32. Missing household data were not interpolated but taken directly from DRI data. Finally, the rent and cost data are indexed to 1987.
5. Estimation Procedures
We derived all the estimates through iterative three stage least square (3-SLS) with a three period autoregressive corrections, AR(3). While we experimented with one and two periods ARs, it was the AR(3) adjustment that yielded the best results. It suggests that the model contains a three period relation of the error terms. The most significant three period effect in the sample period was the Kemp-Roth three-stage 30 percent tax cut between 1981-1983. While it created a sustained period of economic growth, it also ran up the deficits for the rest of the sample period, which indirectly put a strain on FHA's participation in the housing market. The Durbin-Watson statistics indicate that the single equation embedded in the Maisel system shows a lack of autocorrelation in the error terms after the AR(3) procedure was performed. That equation also posted the best t-values and [R.sup.2]. As the [R.sup.2] and [Mathematical Expression Omitted] are close together, the included independent variables do not appear redundant.
We estimated the Fair system with restrictions on the price coefficients to be the same. A variety of estimation methods were offered in the literature, see Fair (1972, 1974), Amemiya (1974). The basic technique we followed regressed the price variables \[Delta]ARM\, and/[Delta]RM/on all the instruments, which we take to be all the independent variable less prices. That gives their fitted values, which we then substituted for the original values to estimate the demand and supply equations. Finally, the AR(3) procedure then minimizes the inter-correlations of the error terms across equations. Essentially, we used an iterative 3-SLS technique with an empirically determined autoregressive structure.
In regard to the Maisel specification, Fair (1972) p. 213 reminded us that a true reduced form requires a constraint to make the net-household formation and removal variable coefficients the same. However, such a procedure gave poor results. Our estimate is therefore more faithful to Maisel's original reduced form model, however inexact it might represent his original demographic specification. Again, iterative 3-SLS with autoregressive corrections were needed.
According to Smith (1974), we should find rents and vacancies highly correlated. As we use them both as predictors, we should suspect multicollinearity. Our model used changes in the ratio of median rent to the residential cost component of the GNP implicit price index (henceforth cost), along with the excess vacancy in 5 + units as defined above as predictors. Maisel used the rent component of the BLS Consumer Price Index as a ratio to cost, along with available vacant, i.e., vacant for sale plus vacant for rent, as predictors. We regressed those ratios on their respective definition of the vacancy rate, using a constant and an Almon one-period lag and a one-degree polynomial. The result was no significant relationship between the respectively defined rent and vacancy variables.
6. Results and Conclusion
We have been looking at the decline in rental completions from different modeling perspectives. We present the results of the three models in this section with some conclusions and rational expectation implications.
[TABULAR DATA FOR TABLE 1 OMITTED]
6. I Single and System Models Performance and Implication
Overall, the results of Tables 1 and 2 suggest that our approach is useful in validating our hypotheses. Table 1 gives the results for our rental specification. It displays the three hypotheses to examine our rental specification under the single equation, the disequilibrium system environment of Fair, (1972, 1974) and the equilibrium system environment of Maisel (1963, 1965). Only the Maisel system environment generated overall significant estimates and correct signs for our specification. The synergy between the single equation model and Fair's disequilibrium model revealed insignificant rent-to-cost ratio, and sources of fund coefficients. The single equation model yielded an insignificant FHA participation coefficient, in addition to the rent-to-cost and sources-of-fund coefficients. From a statistical point of view as well, i.e., [TABULAR DATA FOR TABLE 2 OMITTED] t-values, [R.sup.2], and Durbin-Watson statistic, the Maisel equilibrium model described a more harmonious housing market environment for our model.
The results of Table 1 accepts the regulatory arguments that TRA eroded the tax benefits put in place by ERTA, and that financial reforms occasioned by FIRREA have a significant negative impact on completions in either the equilibrium or disequilibrium environment. Other across-the-board significant coefficients include the constant, a time-trend, excess vacancy, changes in mortgage rates, and average hours worked. The FHA participation coefficient is significant in both the embedded case. Finally, changes in rents adjusted for costs, and the source of fund variable are significant in the Maisel model alone. The significant negative time coefficient corroborates the downturn of completions during the sample period, and the significant positive FHA coefficient validates the FHA hypothesis.
Table 2 displays the results of Fair's and Maisel's system. The equations were fitted without a constant term as preliminary runs revealed insignificant constant coefficients. That result was not a consequence of a dummy variable trap between the first columns of ones, the tax dummy, and the FIRREA dummy in the set of independent variables. It is not surprising as well, as Fair's constant was not significant in the supply side, and Maisel did not report the t-values for his intercept. The tax dummy variable is insignificant in both the demand and supply equations of Fair's model, indicating that other forces have dominated the negative effects of TRA. In addition to the tax variable, the time trend variable is also insignificant on the supply side, showing a weak influence of demographics on completion. The models did not improve when the other tax dummies mentioned above were substituted, nor when the time variable was omitted on the supply side. The differences between our and Fair's results may be reconciled by the differing time dimensions. In short, aggregating months to a year tends to bring in other counteracting factors, or net out opposing predictors that weakens demographic influences on completion. On the tax side, because the equilibrium model of Maisel shows a significant coefficient, the disequilibrium over the sample period in Fair's model may be in a corridor that excludes the tax influence.
Regarding signs of coefficients, all of the demand coefficients match up with Fair's. However, the supply equation shows negative signs for RM and positive signs for time, contrary to Fair's. Several reconciliation factors can be cited. While the time coefficient is of the wrong sign in the supply equation, it is not significant. While we experimented with different lag structure on the RM variable, only the current time observations yielded a significant coefficient. The significant sign difference in RM is also partly due to the different time dimensions of the two models. Being longer, our timeframe allows a net of the monthly variations in the interest rate.
The result of Maisel system model in Table 2 shows that all the coefficients are significant. It shows only one incorrect sign for the change in households. However, that coefficient was extremely small, viz., -0.0003, and the data were punctuated with many interpolations to draw much attention.
6.2 Rational Expectation Results and Implication
The rational expectation implications and findings are illustrated in Figure 1, and Table 3. The results of taking expected value of completions conditioned on all information and knowledge of the model, i.e., on past and current observations and values of the parameters including a parameter for the error term, yield the plot of actual and forecasted values of Figure 1. The models track the turning points well since the 1980s, as the policy predictors relate to that period mainly. The somewhat tight forecast indicates that market participants are predicting systematic policy changes accurately in line with Lucas's (1976) theory. An implication for current FHA policy would be for HUD to use shorter time lags or better information from the market to make its systematic policies have real effects. The graph also shows that the Maisel model predicts an early turn-up in 1994, while the Fair's model tracked the data more exactly for that date. The differences, particularly as [TABULAR DATA FOR TABLE 3 OMITTED] they relate to RE, can be due to time lag which will have to play out over several years, preliminary vs. revised data, and re-benchmarking as observed by Lovell (1986). It could also be a result of the inability to predict the nature of deep parameters such as those of the participants utility function as Taylor (1993) has indicated. We will need time and further tests to tell whether the differences in current prediction among models are real.
Following de Leeuw and McKelvey (1984), and Lovell (1986), we have examined the strong aspect of rational expectation. It examines whether the forecasted errors are uncorrelated with all the variables known to the forecaster. Briefly stated, the procedure regresses the actual data on its forecasted values, with the restrictions that the constant is zero, and the slope is unity. As Table 3 shows, the restrictions are neither individually nor jointly rejected: t-values are significant for the slope and insignificant for the constant. Also, the estimated F-values for the joint test are less than the critical values at the I and 5 percent confidence level, indicating that we should accept the joint hypothesis.
To sum up, we have specified and failed to reject a rental equation that explains the downturn of rental production over the 1970-1994 period. Of the three attempts we made, the single equation did not render all the hypotheses proposed as significant. The two system results showed a dominance of significant coefficients with proper signs, and our rental specification excelled within the Maisel's rather than the Fair's domain. The system models were worth the extra effort. Without them, we could not make a significant inference about FHA's influence on completions, but can do so now as both yielded a significant FHA coefficient. The RE results indicate that agents are anticipating systematic policies from the 1980s. The results also show no significant influence of the rent-to-cost ratios and source of funds in the Fair model, but significant influence in the Maisel model.
The unbiased rational expectation results indicate that agents have full information about what their forecasted level of multifamily completions will be. We find that the agents information on FHA participation is more efficiently used within the system rather than the single equation model. This is a significant finding since the Brady model (1973) has advocated the reverse preference in a non-rational expectation context.
Despite the differences in results, we reject the temptation to pronounce one environment-equilibrium or disequilibrium better than the other, although the former yielded more significant results, overall. They are true prototypes in essence but not in substance to Fair's and Maisel's system. However, we make the generalization that our specification of a rental model seems biased towards an equilibrium environment in regard to statistical attributes. In that environment, we find validations for the tax, crises and credit, FHA, and some other hypotheses mentioned above, and unbiased rational expectation results. The present administration is in a state of transforming FHA to meet new challenges. Future statistics will allow testing of whether the FHA coefficient is still positive, and if FHA participation is at the necessary and sufficient level. It will also throw more light on whether the superimposition of a system model is necessary for future specification of rental market models. It will be interesting to find out which direction current HUD policies in regard to the "Re-invention of Government," which will be in effect for 5 plus years, is steering the rental market. The second term of the Clinton administration is well underway, demonstrating little change in housing policies from the first term.
Alm, James and Follain, James R. (1994). Shocks and Valuation in the Rental Housing Market, Journal of Urban Economics, 36, pp. 117-143.
America's Renters Need. (1995). FHA ISSUE BRIEF NUMBER 8
Amemiya, Takeshi (1974). A Note on fair and Jaffee Model, Econometrica, 42, pp. 759-762.
Anas, Alex, and Arnott, R., (1991). Dynamic Housing Market Equilibrium with Task Heterogeniety, Idiosyncratic Perfect Foresight, and Stock Conversions, Journal of Housing Economics, 1, pp. 2-32.
Anderson, R., and Sonnenschein, H. (1982). On the Existence of Rational Expectation Equilibrium, Journal of Economic Theory, 26, pp. 261-278.
Arrow, Kenneth, and and Hahn, F. H., (1971). General Competitive Analysis, [San Francisco: Holden-Day Inc.]
Bank, David M., and Winnick, Louis (1953). The Structure of the Housing Market, The Quarterly Journal of Economics, 67, pp. 181-203.
Bernanke, Ben S., and Lown, Cara S. (1991). The Credit Crunch, Brookings Papers on Economic Activities, 2.
Brady, Eugene A., (1973). An Econometric Analysis of the U.S. Residential Housing Market, in Ricks, R. Bruce, ed., National Housing Models (Lexington: Lexington Books).
de Leeuw, Frank, and Ekanem, Nkanta E (1971). The Supply of Rental Housing, AER, 61, pp. 806-817.
de Leeuw, Frank, and McKelvey, Michael (1984). The Expectations of Business Firms: Bias in the Short and Long Run, AER, 74(1), pp. 99-110.
Eubank, Arthur A., Jr. and Sirman, C. E (1979). The Price Adjustment Mechanism for Rental Housing in the United States, The Quarterly Journal of Economics, 93, pp. 163-183.
Fair, Ray C. (1972). Disequilibrium in Housing Models, Journal of Finance, XXVII, pp. 207-221.
Fair, Ray C. and Jaffee, Dwight M. (1972a). Method of Estimation for Markets in Disequilibrium, Econometrica, 40(3), pp. 497-514.
Fair, Ray C. and Kelejian, Harry H. (1974). Method of Estimation for Markets in Disequilibrium. A further Study, Econometrica, 42(1), pp. 177-190.
Fair, Ray C., (1979). An Analysis of a Macro-Econometric Model with Rational Expectations in the Bond and Stock Markets, American Economic Review, 69(4), pp. 539-552.
Florida, Richard L. (1986). Housing and the New Financial Markets [New Jersey: Rutgers-The State University of New Jersey, 1986].
Hubbard, R. Glen (1994). Money the Financial System and the Economy [USA: Addison-Wesley Publishing Company].
Klein, Lawrence R., (1950). Economic Fluctuations in the United States, [New York: John Wiley & Sons.]
Kreps, D., (1977). A Note on "Fulfilled Expectations" Equilibrium, Journal of Economic Theory, 14, pp. 32-43.
Ling, David C. (1992). Real Estate Values, Federal Income Taxation, and the Importance of Local Market Conditions, Journal of American Real Estate and Urban Economics Association, 20(1), pp. 125-139, Spring.
Lovell, Michael C. (1986). Tests of the Rational Expectation Hypothesis, The American Economic Review, 76(1), pp. 110-124, March.
Lucas, Robert E., Jr., (1976). Econometric Policy Evaluation: A Critique, Carnegie Rochester Conference Series on Public Policy, 1. [Amsterdam: Morth-Holland].
Maddala, G. S., (1988). Introduction to Econometrics, [NY: Macmillian Publishing Company].
Maisel, Sherman J. (1963). A Theory of Fluctuation in Residential Construction Starts, AER.
Maisel, Sherman J. (1965). Nonbusiness Construction, in Duesenberry, James S. et. al., The Brookings Quarterly Econometric Model of the United States, [Chicago: Rand McNally & Company] pp. 179-201.
McAllister, Patrick H., (1990). Rational Behavior and Rational Expectation, Journal of Economic Theory, 52(2), pp. 332-363.
McCallum, B. T., (1974). Rational Expectations and the Natural Rate Hypothesis, The Manchester School, 43 44, pp. 56-67.
Office of Thrift Supervision (1989). Savings & Home Financing Source Book, Office of Thrift Supervision, Washington DC.
Ricks, R. Bruce (1972). Comment on Fair, Ray C., Disequilibrium in Housing Models, Journal of Finance, XXVII, May, pp. 220.
Rosen, K. T. and Smith, L. B., (1983). The Price Adjustment Process for Rental Housing and the Natural Vacancy Rate, AER, September, pp. 779-786.
Rosen, K. T. (1989). The Apartment Market: A changing demographic and economic environment, Housing Finance Review, Winter, 63-80.
Smith, Lawrence B. (1972). Disequilibrium in Housing Models, Journal of Finance, Discussion, XXVII, p. 228.
Smith, Lawrence B. (1974). A Note on the Price Adjustment Mechanism for Rental Housing, AER, 64, June, pp. 478-481.
Taylor, John B., (1993). Macroeconomic Policy in a World Economy, [New York: W. W. Norton & Company.]
Tinbergen, Jan, (1968). Statistical Testing of Business-Cycle Theories, [New York: Agathon Press, Inc.], first published by League of Nations Economic Intelligence Service, Geneva (1939).
Lall B. Ramrattan, University of California, Berkeley, Extension Instructor, Economist at U. S. Dept. of HUD, and Lecturer at California State University, Hayward, and San Francisco State University, San Francisco. The views expressed in this paper are those of the author and not those of any other persons or institutions such as the U.S. Department of HUD.
|Printer friendly Cite/link Email Feedback|
|Author:||Ramrattan, Lall B.|
|Date:||Mar 22, 1999|
|Previous Article:||The new political macroeconomics: an interview with Alberto Alesina.|
|Next Article:||A puzzle resolved: Japan's high currency value and trade surplus.|