Lead-Lag Relationship between the Real Estate Spot and Forward Contracts Markets

Introduction

Direct investments in property have been found in many previous studies to be superior to investment in stocks, as summarized in Norman, Sirmans and Benjamin (1995). Chiang and Ganesan (1996) found supporting evidence in the Hong Kong real estate markets. Despite their superior performance, including direct property investments in asset portfolios may incur very high transaction costs (Chua, 1999). One possible way to lower these transaction costs is to use direct property derivatives, such as property futures. However, the use of derivatives in direct property investments has received little attention in the literature, except for Chau, Wong and Yiu (2003) and Lai, Wang and Zhou (2004) who established pricing models for real estate pre-sale contracts. Their ideas are that although a formal futures market for direct property does not exist, developers can carry out preoccupation sales (pre-sales1) of new developments well before occupation, so as to transfer their financial risk and gauge the market value of properties. Such pre-sale arrangements are typical in many cities such as Hong Kong and Singapore, and are effectively forward contracts in the sense that the contracting parties have agreed on the price, but the subject property, which is still under construction, will be transferred to the assignee at a later date. In the light of this interpretation, this paper aims to contribute further to the understanding of the pre-sale market by examining the lead-lag relationship between the returns on real estate spot and forward contracts.

There are several reasons why it is of great interest to identify any lead-lag relationship between real estate forward and spot markets. First, practitioners or investors require this information to make their decisions about arbitrage. second, regulators need to know the impact of the real estate forward market on the spot one (or vice versa) to formulate more informed housing and land policies. Finally, existing research largely ignores futures or forward contracts in the real estate market. This is because (1) there is no centralized exchange for trading real estate futures, and (2) the turnover of pre-sales is relatively small in many markets. To tackle these problems, this paper takes advantage of the vast amount of housing pre-sales data in Hong Kong,2 which enables a quantitative study of the lead-lag relationship between the pre-sale market and the spot market of real estate. We hypothesize that their lead-lag relationship is dictated by the level of price information available to flow between the pre-sale market and the spot market. The proportion of the number of transactions in the two markets is taken as the proxy for measuring the information flow.

This paper is organized as follows: The next section reviews the literature concerning the lead-lag relationship between the future and spot prices in the financial market. This is followed by a discussion of the flow of in the information on prices between the forward contracts and spot markets in real estate. Next, the results of the Granger causality tests on the lead-lag relationship are presented. The final section presents concluding comments.

Literature Review

The lead-lag relationship between the spot and futures (also forward) markets has drawn considerable attention. Theoretically, both spot and futures contracts reflect the same aggregate value of the underlying assets and, if instantaneous arbitrage is possible, spot and futures prices should be perfectly contemporaneously correlated. In other words, the two markets simultaneously reflect new information and there should not be any lead-lag relationship between the two markets.

In financial markets, however, some studies support the futures-lead-spot argument (reviewed in Abhyankar, 1998), while others support the feedback relationship (Kawaller, Koch and Koch, 1987; Tang, Mak and Choi., 1992; Wahab and Lashgari, 1993; Abhyankar, 1998; and Brooks, Rew and Ritson, 2001). Abhyankar (1995) puts forward three reasons why the price of stock index futures should lead the underlying spot index from the perspective of transaction costs. First, trading is more infrequent in the spot market. Second, liquidity is higher in the futures market. Third, market frictions are smaller in the futures market because of lower transaction costs, lower capital requirements and no short-selling restrictions.

In general, Abhyankar's (1995) contention is hard to test using financial data because most financial futures markets enjoy the above three advantages in transaction costs over financial spot markets. Interestingly, two of these advantages may not be applicable in the real estate market. First, transactions in the spot market of real estate are often more frequent than those in the forward market, because of the size of building stocks and the filtering-up motivation. second, market frictions in the two real estate markets depend very much on government policy. For example, in Hong Kong, there have been statutory changes in transaction costs and re-sale3 restrictions on pre-sales.

Furthermore, direct property markets are very different from financial markets. The heterogeneity of assets and the high transaction costs of direct property investments make financial models not directly applicable. The problem of liquidity in direct property markets makes them different from financial markets (Stein, 1995; Berkovec and Goodman, 1996; Hort, 2000; and Fisher et al., 2003). Basically, direct property markets are considered as search markets, while financial markets are trading markets. Interestingly, their exposition of how search markets function and affect real estate pricing does not apply in the forward market of real estate in Hong Kong. For example, Fisher, Gatzlaff, Geltner and Haurin (2004) assumed that potential buyers and sellers will negotiate or search for better deals until the reservation prices of the buyers meet those of the sellers. Nevertheless, they assume an all-or-nothing mode of transaction, which is commonly observed in direct property markets. Sellers of forward sales in Hong Kong, however, strategically put small batches on the market to determine market prices by trial and error. The mechanism of determining prices in the forward sales markets will be elaborated upon in the next section. In any case, the divisibility of their portfolios makes their volume of transactions highly different from those in direct property markets, especially during downturns in the market.

Flow of Price Information between Markets

In the real estate spot markets, negotiation or searching is commonly adopted to determine prices, and decisions are very often all-or-nothing owing to the indivisibility of their portfolios (search model). Buyers and sellers collect information on prices (through property agents, etc.) not only from the spot market but also from the forward market. On the other hand, in the primary forward market of real estate in Hong Kong, buyers rely on a developer's price list. Due to the lack of price information for their new products, developers seldom release all of the housing units in a project to the market at one time. Instead, they usually divide the sale into many small batches, and then adjust the price level in each batch based on both the market responses of the previous sales and the changes in the price levels of the spot market. This sale strategy also allows purchasers to choose a unit in the same development, at either the primary and the secondary forward markets, or the primary and the secondary spot markets.4

Exhibit la shows the general price index for private domestic property constructed by the Rating and Valuation Department (1991-2001); Exhibit 1b shows the number of transactions of forward sales and spot sales; and Exhibit Ic shows the volume ratio of forward sales to the total number of sales from January 1992 to December 2000.

The study period is divided into three windows, namely, window 1: Jan-92-Sep-95; window 2: Oct95-Jun-97; window 3: Jul-97-Dec-00. Exhibit Ic shows that all the volume ratios recorded during the second window were below the median. Exhibit 2 further shows the results of the three f-tests for the difference between two means of two samples assuming unequal variances. They clearly reveal that the volume ratios during the second window were lower (and statistically significant at the 5% level) than the other two windows. Note that the median volume ratios lie at about only 20%. This implies that most of the housing transactions took place in the spot market.

When the volume ratio is high, more price information flows from the forward to the spot market, which helps determine market prices. A bidirectional relationship of causality is thus expected between the two markets when the volume ratio is above the median. However, when the volume ratio is low, the flow of price information from the forward to the spot market is relatively reduced. Consequently, price signals mainly flow from the spot to the forward market. In this case, the spot market is expected to lead the forward market in terms of price formation.

Granger Causality Tests

The data for F^sub t,t^ and S^sub t^ are derived from the nominal transaction prices of forward sales and spot sales of private housing in Hong Kong over the past ten years (from July 1991 to March 2001), respectively, in Hong Kong dollars. They are registered in the Government Land Registry and were obtained from a third-party value-added data provider. There were 3,062 pairs8 of repeated pre-sales transactions and about 270,000 pairs of repeated spot transactions. The ex post real rate of interest, which is the difference between the nominal interest rate and the inflation rate, is used as a proxy for y^sub t^. Inflation rates are derived from the Consumer Price Index series A for non-luxury commodities, whereas nominal interest rates are derived from the Hong Kong inter-bank offer rate for three-month deposits.

Exhibit 3 shows the estimated coefficient f in Equation (2) in the form of a price index. The findings indicate that the forward price index for which the term to maturity is zero closely tracks the ex post spot price index, as well as the government's index in Exhibit 1a.

In order to test the change in the lead-lag relationship between the two markets under different scenarios of information flow, the VRs are segregated into two sub-groups: low and high forward sales. When the VR of a certain period is greater (smaller) than the median VR, it is grouped under the high (low) forward sales subgroup. This method of segregation is in line with that used in Abhyankar's (1995) study. Over the study period, the median VR is 20.59%, and the minimum and maximum VRs are 4.08% and 47.43%, respectively. The standard deviation of the VRs is 10.39%.

The Granger causality results are presented in Exhibit 5. Panels A and B show the test results during the periods of low volume ratios (fewer presales) and high volume ratios (more pre-sales), respectively. From Panel A, the first null hypothesis that SR^sub t^ does not Granger-cause FR^sub t^ is rejected at either 1, 2, or 4 lags at the 10% level. Yet, the null hypothesis that FR^sub t^ does not Grangercause SR^sub t^ cannot be rejected at either the 1, 2, or 4 lags at the 10% level. From Panel B, both the null hypotheses of no Granger causality are rejected at the 10% level.

The results support the contention here that the lead-lag relationship depends on the flow of price information, which is measured by the volume ratio. When the volume ratio is low (i.e., there are fewer forward sales relative to spot sales), the information on prices mainly flows from the spot market to the forward market. Therefore, that the spot returns are found to lead the forward returns in the real estate market during periods of low volume ratios, but not vice versa. This spot-leadsfutures finding is opposed to the futures-lead-spot results from other studies of financial markets. On the other hand, when the volume ratio is high (i.e., there is more forward sales relative to spot sales), information on prices flows more freely between the two markets. Therefore, feedback effects are found in their rates of return during periods of high volume ratios. The bi-directional results concur with the general findings in many financial studies.

A Granger causality test is also conducted on the sample by dividing it into two groups. The groupings are based on the three windows. Exhibit 6 shows the results of these Granger causality tests. Panels A, B and C show the test results during the following: (1) window 2 where VRs are smaller than the median; (2) windows 1 and 3 where VRs are greater than the median; and (3) the entire sample period, respectively. The results bear a strong resemblance to the previous results, where SR^sub t^ was found to be statistically significant to Granger-cause FR^sub t^, but not the other way around during the window 2. Similar to the previous results, FR^sub t^ and SR^sub t^ show strong feedback effects during windows 1 and 3 and the entire sample period.

This confirms that changes in the flow of pricing information among the markets affect the causality relationship between them.

Conclusion

Based on the actual transaction prices in the spot and the forward contracts markets of residential properties in Hong Kong, lead-lag relationship between the two markets was tested. The findings indicate that during a period of few transactions in the forward market, the spot return leads the forward return but not vice versa. On the contrary, when the volume ratio is higher (more forward sales), there are feedback relationships between the two markets. The hypothesis was further tested by dividing the sample into three windows. The findings reveal that the spot returns led the forward returns but not vice versa during the period of low VR. There were bi-directional relationships of causality between the two returns during the period of high VR. The results imply that when the forward market does not provide sufficient price information to the spot market, the forward market can be led by spot market.

This paper suggests that liquidity in the futures market determines the lead-lag relationship between financial futures and the spot market. Investors employing real estate derivatives such as forward contracts to hedge investment risk have to be cautious of the market frictions in the derivative markets.

Yet, this study is limited by the efficiency of the price discovery mechanism in real estate markets. It is recognized that the volume of transactions (price information) in real estate markets is relatively small in comparison with those in financial futures and the spot markets. Applying financial theories to real estate markets remains a fundamental problem.