Applying the DCF model in China's transitional market.

With a transaction volume of over 400 pieces of land (or more precisely land use rights) between 1988-1993 and an additional 400 pieces between 1994-1995, Shanghai is, beyond doubt, a very active land market even though it may not be operating on a basis similar to a pure capitalist market mechanism.(1) However, because most transactions are conducted through negotiation, it is doubtful whether a developer or an investor can actually obtain reliable market data with which to assess the value of land that they own or are planning to purchase. To prevent losing an important source of revenue for the state, the local authority should reconsider whether the method of assessing land value reflects the highest and best use principle.

A comparison is made between the factors affecting the current market prices of land charged by the Shanghai government and the appraised land values simulated in the same market environment. ("Market price of land" refers to the price of land that a developer would be asked to pay. It is therefore quite different from the traditional definition of "market land price" in a capitalist society, which defines it as a residual element that depends on the income generated from economic activities carried out on and above land.) Appraised residual land prices are simulated by the discounted cash flow (DCF) model, one of the best appraisal techniques to estimate land value in a market where finding reliable market comparables is difficult.(2) However, a number of assumptions need to be made in using the DCF model in this study.

First, market prices of various properties are taken from developers' asking prices in the general real estate market. This means that properties are developed to be sold to foreigners, Chinese expatriates, or wealthy local citizens. Transactions are normally conducted in Hong Kong dollars. This assumption is made because all the land use rights used in this analysis were sold to joint venture entities with foreign developers as partners. These joint ventures develop primarily external market properties because of the large profit margin they can obtain in this sector compared with the internal property market. Since the secondary market in China overall is not very active, developers' asking prices are almost as good as the market prices. Moreover, because the land sales were conducted in 1992, these property prices are (pre-sale) prices for development that would come to China and Hong Kong's markets in 1994-1995, hence confirming that development takes two and a half years.

The second assumption is that the interest rate on the cost of borrowing is assumed to be 9.5% because most of the development involves investors from Hong Kong. While the prime interest rate in Hong Kong was about 6.5% (the standard mortgage rate is 6.5% + 1.75% = 8.25%) during the test period, a higher interest rate of 9.5% is assumed for the risk involved in China's projects.

Third, the discount rate is fixed at 15% to reflect the risky nature of the projects, as well as the relatively high inflation in China's major cities. This is different from the borrowing rate in that the discount rate reflects the riskiness of both cash inflows and outflows.

Fourth, the developer's profit is set at 30% on the gross development value (GDV) or total sales proceeds of the project, or 45% on total development costs. This is set in consultation with professional firms in Hong Kong. Moreover, this is regarded as reasonable at that period although the profit ratio has, in fact, been dropping.

Fifth, construction costs are total construction costs, including all materials and labor costs. This figure is estimated by a property consultant in Hong Kong to be at least H.K. $350 per square foot, or about U.S. $500 per square meter.(3) (Note that 1 square meter is equivalent to 10.76 square feet.) This figure also coincides with the official indication of building costs between U.S. $373 per square meter to U.S. $585 per square meter for high-rent or luxury residential housing in Shanghai.(4) Consequently, an average of U.S. $520 per square meter is applied for residential development, and U.S. $550 per square meter for mixed-use development projects in 1992, and U.S. $550 per square meter for residential and U.S. $600 per square meter for mixed-use projects in 1993.

Sixth, professional fees reflect the fact that most Hong Kong developers employ their own architects and cost engineers from Hong Kong. However, the fees do not apply to costs for projects in China. As a lump sum, the construction costs are fixed at 5% for all projects.

Seventh, a special column of taxation is added in accordance with Article 48 of the Regulations on the Transfer of Land Use Rights for Valuable Consideration in Shanghai.(5) According to this article, a tax of 3% on GDV is levied when the development comes on the market.

Eighth, an agency fee of 1.5% on the GDV can be exacted at the point of sale. Ninth, the local authority would levy a 33% profit tax on the development profit at the completion of the development. Finally, although the cash payment pattern varies with projects, standardizing the DCF models in the simulation process is necessary. Hence, all construction costs are assumed to be payable quarterly while bank interests on costs are payable half-yearly.

Based on these assumptions, 180 pieces of land sold in Shanghai from 1992-1993 are then reappraised while the actual market prices of these lots are recorded from government publications. These land prices are computed in the format of accommodation value, or AV format, which is basically land price divided by the gross floor area of the development. This is so that the use of the [TABULAR DATA FOR TABLE 1 OMITTED] [TABULAR DATA FOR TABLE 2 OMITTED] unit land price based on per square foot of floor area is consistent.(6) The DCF model applied in the simulation process (see table 1).

To test the validity of the appraisal model of land from which land values are simulated in Shanghai, appraisers need to look at the objectivity of the DCF model. A sensitivity analysis can identify and examine the factors in the model that have the greatest effect on the assumptions used here. Table 2 shows the means and standard deviations of the changes in the appraised land values according to changes in the major assumptions.

In the sensitivity analysis, the value of one of the major assumptions is changed while the rest of the variables are kept constant. The resulting effects on the appraised land values are then recorded. When the most sensitive variables are identified, new sets of residual land values are compiled and the regression analysis is applied to them again. If the results are similar to the results of the original assumptions, then the DCF model is sound.

EMPIRICAL RESULTS

In deciding whether land values simulated by the DCF model represent a proxy for the highest and best use basis of land values in Shanghai, appraisers should check whether the appraised land values correlate better with Shanghai's market environment than the market price of land. Two sets of multiple regression analysis accomplish this.

Multiple analysis has been widely used in estimating the relationship between environmental factors and real estate prices.(7) The regression analysis applied is not intended to show the prediction power of the DCF model over the actual market prices of land, although it could be applied in this way. The purpose of the analysis is to show that it is reasonable to assume that the market environment in Shanghai generates land prices close to the DCF model's theoretical land values. This model's dependent and predictor variables are defined in the following:

Dependent variable

Market land value:

AV = Market land value for land use rights/Gross floor area allowed on site

RAV = Residual land price for land use rights/Gross floor area allowed on site

AV is the actual market price of land in accommodation format and RAV is the appraised land values in accommodation value format.

Predictor variables

Ten predictor variables were identified and drawn from similar research studies. They are as follows:

1. The first predictor variable, Comland used for commercial purposes, is derived from a survey conducted by the Planning Science & Technology Office of the Shanghai Land Administration Bureau in 1992. It shows the amount of land in each of 10 urban districts.(8) Including this variable in the regression analysis makes it possible to examine the effect of the old land use pattern created under the planned allocation of land, a decision developers made under the land use rights reform.

2. The second, Employ, shows the increase between 1991 and 1992 in the number of people employed in private enterprises in the urban districts in which the sites are located. This variable relates to the local economy (district by district) in terms of private enterprises being market oriented. The higher the number of people employed in private enterprises in a particular district, the higher the demand for premises by these enterprises and their employees.

3. Grade represents the land grade as classified by Shanghai's city government within an urban district. Five grades of land are applied to the entire urban Shanghai municipality. To avoid possible distortion (such as two choices available for one dummy variable), five dummy variables are used instead of just one variable for the predictor "land grade".(9) Hence, there are grades 1, 2, 3, 4, and 5. When a particular site is on grade 3 land, the variable for grade 3 is 1, and the variables are all Os.

4. Hirise represents the quantity of high-rise buildings (those buildings over 20 floors) in terms of gross floor area (GFA) out of the total built-up GFA in an urban district. This variable makes it possible to examine how the density of the developments affect land price in a particular district.

5. Indland represents the quantity in each of 10 urban districts. It was developed for industrial use by the Planning Science & Technology Office of the Shanghai Land Administration Bureau in 1992. The inclusion of this variable in the regression analysis makes it possible to examine the effect of the old land use pattern created under the planned allocation of land, a decision developers made under the land use rights reform.(10)

6. Popden represents the population density per square kilometer in the respective urban district in which the site is located.

7. Because of the relatively short study period of land price behavior, the 1992-1993 land sale program in Shanghai is divided into four half-yearly subperiods (T1, T2, T3, and T4). This method shows the effect of time on land prices in Shanghai.

8. Size is the size of the site in square meters.

9. A dummy variable is used for Use Type 1. It represents mixed-office and upscale residential buildings, projects that attract foreign developers and investors.

10. Distance in kilometers of the site from the Bund (a riverbank along the Huangpu River and once the central business district of Shanghai) where a financial street will be redeveloped.

In addition, it should be noted that in the 1992 land use survey of Shanghai's Land Administration Bureau, at least 10 different land uses were classified in the survey, but only those with substantial effects on land prices were selected for the regression model. Originally, the percentage of residential land was also selected for the model, but was eliminated because it correlated very closely with another predictor variable, population density. This multicollinearity effect, i.e., the correlation effect that results from the interrelationship between two between predictor variables, is undesirable.

Moreover, in the different types of development to be put on sites, only the Use Type 1, or the mixed-office/high-rent residential development is included as a dummy variable. The reason for this is to establish the highest and best use of land which, given the current market, is office and upscale residential.

Finally, the test results show that a relatively higher adjusted [R.sup.2] (0.65) is obtained in the regression analysis of the appraised land values compared with a low adjusted [R.sup.2] (0.30) in the market prices of land. The function of the adjusted [R.sup.2] is to show the degree that the predictor variables can explain the appraised land values. In this case, about 65% of the determination of the appraised land values can be explained by the predictor variables chosen in the analysis. Generally, the higher the adjusted [R.sup.2], the more useful the predictor variables and vice versa. It is, therefore, apparent from the test that the market in Shanghai relates more to the hypothetical appraised land values than to the actual market prices of land charged by the local authority.

Further Validation

The sensitivity analysis already examined shows that three components in the DCF appraisal model are influential in determining appraised land values: sales prices of the development, construction costs, and developer's expected rate of profit. Systematic changes to these three components are made one by one, and the resulting appraised land values are recorded. These new appraised land values under the new assumptions are then regressed and the new test results compared with those obtained before. Table 3 summarizes the effects of these changes in assumptions.

For a 50% fluctuation in both directions in any of the major components in this DCF model for appraising the residual land value, consistent regression results are obtained from and support the conclusion drawn in a later section. The only exception to this observation is a 50% drop in property prices - rather unusual in a growing property market such as Shanghai, in which international companies have substantial interests.(11)

As a result, it is quite reasonable to state that the standardized assumptions applied in the DCF model are realistic and the appraisal model can estimate land values that could have been realized if the market mechanism had been a more competitive one.

In addition, table 4 shows that the level of market prices of land charged by the city government in Shanghai is much lower than the appraised land values. With the promising economic growth of this Chinese metropolitan city and the substantial growth in real estate values in the past few years as well as the future,(12) it is reasonable to say that the DCF model's land values do represent the market value of land use rights in Shanghai.

CONCLUSION

As with other financial analytical techniques, the DCF model is not without drawbacks. The relationship of risk as it applies to the discount rate as opposed to the capitalization rate in the DCF model,(13) must be considered, along [TABULAR DATA FOR TABLE 3 OMITTED] [TABULAR DATA FOR TABLE 4 OMITTED] with the lack of substantial proof that the use of the DCF model actually improves an investor's performance,(14) and the reasonableness in setting the various assumptions in the DCF model.(15) However, this model remains a robust analytical tool for assessing the value of land in a market where limited information is available and where land prices are distorted because of the immature market mechanism of China's transitional economy.

A more open and competitive mechanism to price land use rights is needed to achieve market land prices close to those obtained in the DCF model. These market land prices are more closely correlated to the current market in Shanghai and are higher than the actual prices of land currently charged. While it is this article's objective to speculate the purposes of the current land pricing mechanism in Shanghai, when a secondary market of land is gradually developed among private developers later on, it will become possible to estimate market land prices using the DCF model.

This study provides a framework for analyzing the reliability of existing market data by comparing it with a set of hypothetical appraisal values. The appraisal values are a sort of audit system for estimating land prices in a new and emerging real estate market. Normally, in a developed real estate market, appraisal data does not provide this function since market prices reflect highest and best use. But when the market itself, though very active, does not function efficiently, the appraisal data obtained from the DCF model may provide a reference point for both the state and investors.

1. Ling Hin Li, Privatization of Urban Land in Shanghai (Hong Kong: Hong Kong University Press, 1996), 44-58.

2. B. R. Weber, "Market Value Without a Market," The Appraisal Journal (October 1990): 523-532; T. V. Grissom, "Appraisal Models and the Feasibility Process," The Appraisal Journal (April 1986): 214-224; and T. J. Keith, "Applying Discounted Cash Flow Analyses to Land in Transition," The Appraisal Journal (October 1991): 458-470.

3. Review Publishing Company, Ltd., "And Now for China," Far East Economic Review (December 1992): 59-60.

4. Office of the Shanghai Municipal Committee for Land Use System Reform/Municipal Statistical Bureau of Shanghai, Shanghai Real Estate Market '93 (Shanghai: China Statistical Publishing House, 1993), tables 4-10-1 and 4-10-2.

5. Regulations on the Transfer of Land Use Rights for Valuable Consideration in Shanghai, promulgated on November 29, 1987, by the Shanghai Municipal People's Government.

6. G. D. Jud, "The Effects of Zoning on Single-Family Residential Property Values: Charlotte, North Carolina," Land Economics (May 1980): 142-154.

7. Maurice H. Yeates, "Some Factors Affecting the Spatial Distribution of Chicago Land Values, 1910-60," Economic Geography (1965): 57-70; David M. Grether and Peter Mieszkowski, "Determinants of Real Estate Values," Journal of Urban Economics (1974): 127-146; Michael S. Johnson and Ward R. Ragas, "CBD Land Values and Multiple Externalities," Land Economics (November 1987): 337-347; Richard B. Peiser, "The Determinants of Nonresidential Urban Land Values," Journal of Urban Economics (1987): 340-360; Chien Han et al., Analysis of the Changes in Urban Land Prices in Taiwan (Taipei, Taiwan: National Science Committee, 1992), 40-61.

8. Planning Science & Technology Office of the Shanghai Land Administration Bureau, Report on Land Use Survey (Shanghai, China: Planning Science & Technology Office of the Shanghai Land Administration Bureau, 1992), 18-35.

9. Richard F. Gunst and Robert L. Mason, Regression Analysis and Its Application (New York, New York: Marcel Dekker, Inc., 1980), Chapter 3.

10. Report on Land Use Survey, 7, 12,19.

11. Anthony Walker, Kwong-Wing Chau, and Wai-Chung Lai, Hong Kong in China: Real Estate in the Economy (Hong Kong: Brooke Hillier Parker, 1995), 53-57.

12. Privatization of Urban Land in Shanghai, 94-113.

13. Martin A. Skolnik, "Comments on Discounted Cash Flow Analysis," The Appraisal Journal (July 1993): 394-398.

14. W. B. Martin, "Direct Capitalization or Discounted Cash Flow Analysis?," The Appraisal Journal (July 1993): 390-393.

15. Mackenzie S. Bottum, "Discounted Cash Flow Analysts: Tests of Reasonableness," The Appraisal Journal (January 1993): 138-143.

REFERENCES

Li, Ling Hin. "The Official Land Value Appraisal System Under the Land Use Rights Reforms in China," The Appraisal Journal (January 1995): 102-110.

Needham, B. "A Theory of Land Prices When Land Is Supplied Publicly: The Case of the Netherlands," Urban Studies (1992): 669-686.

Walker, Anthony, and Ling Hin Li. "Land Use Rights Reform and the Real Estate Market in China," Journal of Real Estate Literature (1994): 199-211.

Ling Hin Li, PhD, is an assistant professor of real estate economics at the University of Hong Kong. He received his BSc in land management from the University of Reading in England, and his PhD in real estate from the University of Hong Kong. His work has appeared in many real estate journals published in the United States and abroad. Contact: Dept. of Real Estate and Construction; University of Hong Kong; 5/F Knowles Building; Pokfulam Road; Hong Kong. 011-852-2859-2146. Fax [phone], Email [e-mail].