1.1 Introduction Real estate, Stock ow models chapter focuses on price movements according to change on stock in housing market. It also briey analysis lack of models and its biases. 1.2 Model The paper House Price Dynamics: The Role of Tax Policy and Demography focuses on price movements in U.S. between early 70s and late 80s. Author tries to base reasons of price uctuations on two factors which are changes in tax code and demography of households in US. Figures belong to housing market indicates how it is important for overall economy. In 1990s households (abbreviated as HH from now on) net worth is $17.1 trillion where $4.6 trillion in gross values of it belongs to owner occupied homes. The fact that corporate equity owned by HH, which is around $2.4 trillion, is far less than owned house. Three dierent data has been used for regressions. A data from Census Bureau Data has been used for aggregated information according to Census Regional criteria. 1 More disaggregated data is available from the National Association of Realtors (NAR) and reports quarterly median house prices for 115 Standard Metropolitan Statistical Areas (SMSA). While Census Bureau constant quality prices index for valuing only structure, NAR median house prices captures combines cost of land and structure. Third and last data set belongs to Urban Land Institute (ULI) which provides semi decadal data based on survey of land cost and available since 1975. The survey asks experts to price a standard improved 10.000 sq. foot in 30 dierent cities. In order to explain house price movements, we need to indicate that Housing market consist of two sub-markets: Stock for existing house and ow of new construction. Price of house determined by existing stock house whereas level of new investment is determined by ow of new constructions. Thus, any shock to those sub-markets will eect house price. Equilibrium of owner occupied house market requires house owner should get some return 1 See Appendix 1 1 2 CHAPTER 1. REAL ESTATE, STOCK FLOW MODELS from investing to houses as they could get some return when they invest on any other asset. Furthermore, this requires equation 1 which is user cost equation. where R H denotes the marginal value of the rental services per period on owner-occupied homes, P H the price of existing houses, the investors marginal tax rate, i the nominal interest rate, p the property tax rate as a share of house value, the depreciation rate on housing capital, the risk premium required on assets with the risk characteristics of housing, m the maintenance cost per unit value, and e the investors expected rate of nominal house price appreciation. 2 The fact that the issue with this equation is that it assumes all house has same quality, location and other characteristic. Nevertheless it help us determine what investor will do after looking at cost of investing via this equation. Housing stock is determined at beginning of every period by past investment and housing stock then determines RH. In the equation 1, everything except e are endogenous. So, expected housing ination shapes rational expec- tations on housing investors. Then, it becomes a connection between housing price today and housing investment in future. This relation is demonstrated on equation 2. Under the assumption of expectation which is formed with future prediction, e t = (P H,t+1 P Ht )/P Ht , equation 1 and 2 becomes pair and initial house price should be deter- mined in order to solve forward. In this framework, real price can increase by supply shock which can cause an increase on price of current or future construction cost or by current and anticipated demand shocks which can cause an increase on value of rental services of owner occupied house stock. On the other hand we can shape framework with extrapolative ex- pectations rather than rational. This allow systematic overbuilding of houses in the housing market and also allow to predict what excess return will be on houses. House price move- ment during 70s and 80s showed that investor in housing have extrapolative expectations therefore there are not rational on investment decision making process. There are three popular way of explanation of price raise on 70s. First one is shock to construction cost which caused systematic change. Afterwards, it resulted with an increase in housing price relative to GNP deator. Second one is favorable and unexpected demand shock. As it is demonstrated on equation 1, U.S. tax code allows households to deduct nominal interest rate from taxable income. If nominal interest rates increase 1 by 1 according to ination rates, then [(1-)i- e ] which is after tax of borrowing declines as ination rate increases. The tax reform during 80s reduced marginal tax rates, and this increased marginal 2 taken from Poterba, J.M., Weil, D.N., and Shiller, R. (1991). House Price Dynamics: The Role of Tax Policy and Demography. 1.2. MODEL 3 cost of housing and it stressed the price of homes, especially hold by higher income of house owners. Third one is based on demography. The forecast is that people between 25-35 years old will demand a house more than other part of community due to fact that marriage, becoming independent etc. So, Mankiw and Weil(1989) argue that entry of large cohort into 24-35 age period should be forecasted earlier than they start demanding new houses. However, they found that price increase in during 70s mainly caused by baby-boom generation. They fall into 24-35 age bracket during 70s and high demand increased house prices. In addition to that, next generation was not populated as much as baby-boom generation and this could be one of the reasons of price fall during 90s. In order to estimate price change on relative appreciation of dierent sized houses, re- gression has run based on hedonic approach. Big-size houses that can be bought by high income household and small size houses can be bought by low income households are dened as a trade-up houses and starter houses respectively by author. Findings shows that broad patter of price change on dierent sized houses has a same patter with a change on real user cost as it can be seen on table 2. But, hedonic approach has a shortcomings. A better housing index should include analyzing repeat sales of houses. Because repeat sales method 4 CHAPTER 1. REAL ESTATE, STOCK FLOW MODELS has advantage of capability of controlling for the heterogeneity in housing characteristic and the less data requirement in the estimation. For that, Karl E. Case and Robert J. Schiller discuss about repeat sales in their article, The Eciency of the Market for Single-Family Home, which is published on The American Economic Review on March, 1989. Their paper tries to nd an answer to eciency in housing market for years from 1970 to 1986 for four dierent cities; Atlanta, Chicago, Dallas and Oakland/San Francisco. They use repeat sale data on individual homes which havent have apparent characteristic change. On their paper, they modied a model based on Martin J. Bailey, Richard Muth and Hugh Nourse (Initials BMN) and called it as Weighted Repeated Sales. It is three-step weighted procedure. For rst step, BMN model followed, which produces estimation and standard errors for housing price index by using ordinary least square regression method on change in log price of each house. In second step, squared residuals from rst regression has regressed again on a con- stant and the time distance between sales. Constant term is estimate of variance of noise in price and slope term is estimate of variance of individual housing value through time. In the last step, they rst divide each observation in the regression in step-one by square root of tted value in second step regression, then they run regression again. They found out that citywide real hosing prices in given certain year tends to anticipate a change on same direction for the following year. Also, forecasted change in real interest rates does not seem to be incorporated in prices. Additionally, taxes and rental rates reveal that city wise after-tax return excess are predictable. However, Man Cho also indicates that there is various potential biases on repeat sales which comes from the structural changes between sales or by some time-varying structural qualications. There are 5 dierent type of bias in repeat sales indices: renovation bias where is caused by structural changes between two sales; hedonic bias which is a bias comes from values of some structural and locational characters changes over time; trading- frequency 1.2. MODEL 5 bias where trading frequency causes some bias on measured rate of appreciation; sample- selection bias where occurs when there is a bias on estimates of the changes in the value of the stock of housing and aggregation bias where time interval causes bias. In order to solve these issues, 3 dierent methods are suggested: Hedonic Repeat Sales Model which models the eects of the age of structure, a vector of time dummies in the hedonic model, and a set of time dummies in the Repeat Sales Model on the covariance between repeat sales and hedonic residuals; Intercept Repeat Sales Model which adds a xed eect item to Repeat Sales Model and estimate mean and variance of xed and temporal items of repeat sales by using maximum likelihood method , and Distance Weighted Repeat Sales Model that decompose the price appreciation as per-period expected return, jurisdiction-specic return, and random changes private property to a house, then assumed distributional characteristics of error terms designate covariance between housing returns of those two jurisdictions. It is also important to analyze price change in city level over time according to change in demography, construction cost, incomes and tax rates. Equation 3, reduced form cross section model gives us the relation in empirically. where p it is the logarithm of the real median house price in city i at period t, c it denotes the logarithm of real construction costs, d it is the logarithm of a demand measure based on population structure, y it is the logarithm of real per capita income, u it is an indicator of real user costs of home-ownership, and v it are residuals. 3 The demand pattern is result of holding a house, household formation decisions and decision about how much to purchase based on holding conditions. All of those are eected by price level. Migration decisions are also aected by price level. Thus, in order to prevent endogeneity , d it is calculated by using national index of aged specied housing demand. Another variable is a measure of user cost of housing in dierent cities. Including nominal interest rates and expected ination, most of components of user cost doesnt change among cities. However, federal marginal tax rates vary among cities where households can deduct mortgage interest and property taxes. What about the question of forecast of price movements. Recent studies convey that prices can be anticipated. Equation 4 demonstrates this arguments empirically. where r t in a return on 90-day Treasury Bills and dependent variable is excess on housing investment. 3 taken from Poterba, J.M., Weil, D.N., and Shiller, R. (1991). House Price Dynamics: The Role of Tax Policy and Demography. 6 CHAPTER 1. REAL ESTATE, STOCK FLOW MODELS 1.3 Conclusion As a conclusion, demographical changes might be a possible explanation for price decline during 80s. There is a strong relation between level of real house price and housing demand related with age prole of population for United States. Caution is needed in extrapolating historical trend of hosing prices and demography far into future. Because, even if there is relation, on SMSAs level demography-house price link doesnt always hold. Investors in owner occupied house market does not have a rational expectations but extrapolating for estimating future capital gains on housing. This could be a reason of price movement during 80s due to fact that investors hadnt realized yet that user cost is higher than 70s. Same intuition can be make price decline late 80s. Investors had extrapolated price deductions starter mid-80s and shaped their decisions according to recent price decline. References Poterba, J.M., Weil, D.N., and Shiller, R. (1991). House Price Dynamics: The Role of Tax Policy and Demography. Brookings Papers on Economic Activity, Vol. 1991, No. 2, pp. 143-203. Case, K.E. and Shiller, R.J. (1989). The Eciency of the Market for Single-Family Homes. The American Economic Review, Vol. 79, No. 1, pp. 125-137. Cho, M. (1996). House Price Dynamics: A Survey of Theoretical and Empirical Issues. Journal of Housing Research, Volume 7, Issue 2, pp. 145-172. 1.3. CONCLUSION 7 Appendix 1