A Comparative Analysis of Inflation-Hedging Characteristics of Property and Non-Property Assets in Singapore

Transcription

1 A Comparative Analysis of Inflation-Hedging Characteristics of Property and Non-Property Assets in Singapore SING, Tien Foo * & LOW, Swee Hiang Yvonne Department of Real Estate School of Design and Environment National University of Singapore 4 Architecture Drive Singapore Abstract: The study empirically tests the inflation hedging characteristics of the property and nonproperty assets in Singapore. The results support that property provides better hedge against inflations than non-property assets: stock, securitized property and Treasury bills. Industrial property is found to be the most effective hedge against both expected and unexpected inflation, whereas shop offers only significant hedge against the expected inflation. The returns of the two assets establish more than one-to-one correspondence relationships with the inflations. Therefore, industrial property and shop, when included in an investor s portfolio, provide not only hedges against their own inflation risks, the returns also protect inflation risks of other assets in the portfolio. In the sub-period analyses involving four different 5- yearly periods, it was found that more assets established significant hedges against both expected and unexpected inflation in the most recent sub-period from 1993 to When the tests are extended to examine the inflation hedging characteristics of assets in the high and low inflation environments, residential property is found to be a good hedge against unexpected inflation in the low inflation period, whereas industrial property shows a better hedge against inflations during the high inflation period. In conclusion, our empirical tests do not find sufficient evidence to reject the hypothesis that property assets in Singapore are generally better hedges against inflation than the non-property assets. Keywords: Property and non-property assets, inflation-hedging, expected and unexpected inflations Please do not quote and circulate the paper without permission! * Correspondence please forward to the first author by mail or at rststf@nus.edu.sg. Comments are welcome. The authors wish to thank Dr Thang, C.L.D for her kind comments on the earlier draft of the paper.

2 A Comparative Analysis of Inflation-Hedging Characteristics of Property and Non-Property Assets in Singapore 1. Introduction With economic condition on the rise, albeit some hiccups along the way such as the 1997 Asian s financial crisis, there has been a marked increase in the value of investment assets such as stock, bond, property and global funds, in Singapore over the last ten years. In November 1993, the Singapore government privatized its Singapore Telecom (now known as SingTel), one of the leading telecommunication operators in Asia offering a wide range of telecommunication services, via the initial public offers (IPOs) on the Singapore Stock Exchange (SES). The SingTel IPOs, which has a current market capitalization of S$36.16 billion (as in June 2000), have been overwhelmingly received. Every man in the street in Singapore was given the opportunity to share the upside potential and the growth of SingTel through the government s share-ownership scheme. Under the scheme, Singaporeans were allowed to purchase designated SingTel IPO shares at discount using their savings in the Central Provident Fund (CPF) accounts. 1 Besides stocks, property constitutes one of the most valuable assets of an individual portfolio. The high price of property in Singapore does not deter individual to own more than one property. More often than not, one property is bought for the owner s occupation purpose, while other properties are let out for rental income, i.e., for investment purposes. These have come about because of the growing affluence of Singaporeans as well as easier access to funds such as their Central Provident Fund savings. The inflation rate in Singapore as indicated by the Consumer Price Index (CPI) has been relatively stable over the last ten years. Its rate of increase has been hovering around 2% per year and is one of the lowest in the world. For investors who hold assets for long-term returns, inflation risk is always of great concern to them because high inflation erodes the real returns accruable to them in the future. On this account, assets that are inflation-hedges appear more attractive than those that are not. 1 The Central Provident Fund (CPF) is a form of compulsory saving for retirement purposes. Both employers and employees contributed to the savings via an automatic deduction from the employees monthly salary. The CPF savers are allowed to withdraw the money from the ordinary account of the CPF to purchase public houses or private properties approved under the residential property investment scheme, and also to invest in approved stocks and unit trusts. 1

3 Traditionally, real estate has been regarded as one of the best inflation hedges, especially in a land scarce country like Singapore. Due to the inelastic and limited supply of land resources, Singaporeans believe that prices of real estate in Singapore would rise indefinitely in the long term. As such, real estate has been accepted as an attractive alternative to the traditional investments such as stocks, government and corporate bonds, and treasury bills. Compared to the overseas studies on the same topic, there is limited empirical evidence published on the inflation hedging characteristics of the property returns in Singapore. This study therefore aims to fill the empirical gap by examining the extent and effectiveness of the property returns in hedging the inflation in Singapore. Distinguished from other unpublished studies in Singapore, there are three major differences in our empirical tests. First, the empirical tests on the inflation-hedging features in Singapore are comprehensive, both in terms of the length and depth of the samples. The empirical samples cover a wide range of property (all-property, shop, office, residential and industrial properties) and non-property assets (all-share, allproperty share, Treasury bills) for a 21-year period from the first quarter of 1978 to fourth quarter of Second, the Fama and Schwert (1977) framework though is adopted in this study to test the assets hedges against three types of inflation: observed, expected and unexpected inflations, our proposed regression models take specific diagnostic measures on the serial correlation of the models. Unlike some previous studies, the low Durbin-Watson statistics obtained in the regression are corrected by using k-lags autoregressive error terms. Third, the sub-period analyses whereby the entire sample period is sub-divided into four 5-yearly sub-periods and also sorted into high and low inflation regimes, are not found in the previous inflation-hedging studies on Singapore markets. The 5-yearly sub-period analysis allows us to examine the inter-temporal changes of the inflation hedging characteristics of assets. The other test provides evidence on different responses of assets returns to inflation during high and low inflation regimes. It is hoped that the research findings would help investors to better understand the inflation-hedging characteristics of different assets when making the investment decision. This paper is organized into six sections. Section 2 reviews the past empirical studies on inflation-hedges of property and non-property assets. Section 3 gives an overview 2

4 of the inflation experience of Singapore in the past twenty years. The theoretical framework and empirical methodology proposed for the study are discussed in Section 4. Different components of the empirical model for inflation-hedging tests and the data sources used are also discussed. The empirical analyses using data for the past twenty-one years are carried out and the findings are presented in Section 5. Section 6 concludes the study with a summary of the key findings. 2. Literature Review The empirical study by Fama and Schwert (1977) provides a robust framework for testing the Fischer s hypothesis of inflation. This classical methodology has then been widely adopted in testing the inflation ability of different assets in different countries. The results obtained from various empirical tests were varied even though the same methodology has been adopted. By classifying the studies according to the asset types, the findings of selected studies are reviewed in this section. Real estate has long been regarded as a good hedge 2 against inflation. Empirical evidences obtained from independent country and cross-country studies have been supportive of the hypothesis that the property returns move in one-to-one correspondence with the inflation rates. For the real estate market in the US, the evidence of positive hedges against inflation in particular the expected inflation has been consistent (Fama and Schwert, 1977; Brueggeman, Chen and Thibodeau, 1984; Hartzell, Hekman and Miles, 1987; Coleman, Hudson-Wilson and Webb, 1994; Liu, Hartzell and Hoesli, 1997; Wurtzebach, Mueller and Machi, 1991). The findings on the unexpected inflation are, however, less conclusive. Wurtzebach, Muller and Machi (1991) found that office and industrial property returns showed no significant hedge against unexpected inflation. It is believed that the low inflation experienced in 1980s was a factor contributing to the insignificant regression results. In the UK, real estate also provides good protection against inflation (Limmack and Ward, 1988; Liu, Hartzell and Hoesli, 1997; Hoesli, MacGregor, Matysiak and Nanthakumaran, 1997; Miles, 1996; Matysiak, Hoesli, MacGregor and Nanthakumaran, 1996). On the unexpected inflation, Limmack and Ward (1988) found that office and shop offer no 2 Assets that have the ability to protect investors from the effects of inflation are generally labeled inflation hedges. 3

5 significant hedge, whereas Hoesli, MacGregor, Matysiak and Nanthakumaran (1997) showed that the unexpected inflation was not hedged by the income, but it was significantly hedged by the capital returns of the real estate. The results are not dissimilar for studies in countries like Switzerland (Hoesli, 1994; Liu, Hartzell and Hoesli, 1997), Canada (Newell, 1995), New Zealand (Newell and Boyd, 1995), Australia (Newell, 1996) and Hong Kong (Ganesan and Chiang, 1998). On stocks, the empirical results in the US studies in general support the hypothesis that stock offers no significant hedge against inflation (Fama and Schwert, 1977; Hartzell, Hekman and Miles, 1987; Rubens, Bond and Webb, 1989; Coleman, Hudson-Wilson and Webb, 1994; Gultekin, 1983). The same findings were obtained by Liu, Hartzell and Hoesli (1997) for the UK stocks. In another independent study by Hoesli, MacGregor, Matysiak and Nanthakumaran (1997), it was also shown that the capital returns of stock did not provide a hedge, but the income of the stock did significantly hedge against the inflation. The findings by Hoesli (1994) and Liu, Hartzell and Hoesli (1997) also rejected the hypothesis that the Swiss stock is a good inflation hedge. The findings by Newell (1996) and Liu, Hartzell and Hoesli (1997) on the Australia stock markets were contradicting. For New Zealand (Newell and Boyd, 1995) and Hong Kong (Ganesan and Chiang, 1998) stock markets, there were perverse hedges against unexpected inflation, whereas in Canada, the stock return was insignificantly related to the expected inflation (Newell, 1995). For the inflation hedging behaviors of the property stocks or the equivalent of the Real Estate Investment Trusts (REITs), the findings in countries like the US (Liu, Hartzell and Hoesli, 1997), the UK (Hoesli, MacGregor, Matysiak and Nanthakumaran, 1997; Liu, Hartzell and Hoesli, 1997), Switzerland (Hoesli, 1994; Liu, Hartzell and Hoesli, 1997) and Australia (Newell, 1996; Liu, Hartzell and Hoesli, 1997), showed no significant hedge against inflation. In Hong Kong, the empirical results of Ganesan and Chiang (1998) showed that stock has a partial hedge against expected inflation, but a perverse hedge against uexpected inflation. For the bond markets, it was also generally accepted that bond offers no significant hedge against inflation in countries like the US (Hartzell, Hekman and Miles, 1987; Coleman, Hudson-Wilson and Webb, 1994), the UK (Hoesli, MacGregor, Matysiak 4

6 and Nanthakumaran, 1997), Canada (Newell, 1995), New Zealand (Newell and Boyd, 1995) and Australia (Newell, 1996). Based on the literature above, it was clear that the subject of inflation hedging of assets has been widely studied in other countries particularly in the US and the UK. The evidence may not be the same for all the studies, but we could generalize the findings for the four main asset types. Real estate offers significant hedge against inflation, in particular the expected inflation in most of the countries. The hedge against unexpected inflation is not significantly shown by the real estate returns. For stocks, property stocks and bonds, we found no hedge against inflation in most of the countries, though minor variations were observed in some tests. Therefore, our general perception that real estate provides good protection against the erosion of purchasing power is not empirically unfounded. In Singapore, there is no published empirical finding thus far on the inflation-hedging characteristics of various investment assets, though there are few unpublished works investigating the inflation hedging effectiveness of property and property stocks. The present study aims to expand upon existing local literature in this subject. 3. Inflation Rate Movement in Singapore from 1978 to 1998 Singapore has been very successful in its efforts to contain the inflation. Between 1978 and 1998, the average inflation rate, i.e. changes in CPI, was as low as 0.62% per quarter, or an equivalent of 2.48% on an annualized basis. The historical movements of inflation rates are graphically shown in Figure 1. The late 1970s and the early 1980s were the periods experiencing high inflation due mainly to the fluctuating oil prices. The second oil crisis hit in 1979 had caused the inflation rate soaring to 3.64%, or 14.56% on an annualized basis, in the third quarter of The uncertainty in the oil prices and the pressure of wage increases in the early 1980 had kept the inflation rate high at around 2.32 % per quarter between the 3 rd quarter of 1979 and the 3 rd quarter of The inflationary pressure only began to ease in the 1982, which saw the largest decrease in inflation by 1.15% in the 3 rd quarter of The contraction of the economic outputs since 1982 as a result of the slowing down in the world demand had brought Singapore s economy into the first post-independence 5

7 recession in 1985 with a negative GDP growth of 1.6%. A string of cost and taxcutting measures had been implemented to restore the international competitiveness of Singapore s economy and to boost exports. The CPI s rates of change have been subdued at an average quarterly rate of 0.12% between the 1st 1985 and the 1 st quarter 1989 as a result of the cost cutting measures. The 1990s was a period characterized by stable inflation and rapid output growth. The inflation rates have been kept within a narrow quarterly band of 1.44% (3Q1995: 0.06% and 4Q1990: 1.50%) before the major Asian s financial crisis in July The financial crisis has not spared the Singapore economy, and it suffered a 4.12% quarterly contraction in the outputs in the 1 st quarter of The inflation has also declined by 0.67% in the same quarter of The contraction effects triggered the concerns of deflation spiral in the economy, where businesses and households may hold back their investments and consumptions in anticipation of further declines in prices. The consumer price index registered negative growth rates for three consecutive quarters in also witnessed four consecutive declines in the quarterly property prices by 8.21% to 13.01% due to the weak demand. The historical analysis of the inflation rate movement helps to shed lights on the effects of inflation on the hedging performance of different assets. 4. Empirical Methodology 4.1 Theoretical Framework According to Fischer (1930), inflation is a phenomenon whereby there is a sustained and inordinate increase in the general price level. If serious, it can erode confidence in a country s currency, inhibit the conduct of business, or even lead to the collapse of the economy. Inflation hedge, on the other hand, is a protection against the risk of losing the purchasing power as a result of the rising price of a good. An asset is an inflation hedge if and only if its real return is independent of the rate of inflation (Fischer, 1930). Fama and Schwertz (1977) provide an operational definition that is widely used in the empirical tests of inflation-hedging hypothesis. According to the definition, an asset is said to be a complete hedge against inflation, if and only if the nominal return on the asset varies in a one-to-one relationship with both expected and unexpected inflations. 6

8 Theoretically, the expected nominal rate of return of an asset is composed of three components: expected real rate of return, expected inflation and unexpected inflation, which can be written as follows, where ~ E R ~ ~ φ, ) = E( ~ ι φ ) + E( φ ) + γ [ E( φ )] (1) ( jt t 1 t jt t 1 t t 1 j t t t 1 ~ E( R jt φ t 1, t ) = Expected nominal rate ofreturn given the information φ at t-1 E( ~ ι jt φt 1) = Expected real rate of return given the information φ at t-1 ~ E( t φ t 1 ) = Expected inflation rate given the information φ at t-1 ~ [ t E( t φ t 1 )] = Unexpected inflation rate which is given as the difference between the actual and expected inflation at time t. = Actual or observed inflation rate. t The tilde signs (~) denote the random variables. The expected nominal rate of return model in equation (1) could be tested empirically using the following regression model: ~ R jt = α j + β j ) ~ E φ + γ j [ )] ( t t 1 t E( t φ t 1 + ε jt (2) Where α j, β j, and γ j are the regression coefficients, and ε j is the random error term. The hypothesis of the inflation-hedging tests could be set such that if the null hypothesis: β j = γ j = 1.0 could not be rejected, the regression coefficients, β j and γ j are statistically indistinguishable from 1.0. The asset is then said to be a complete hedge against inflation. In other words, the expected nominal return of the asset varies in one-to-one correspondence with both the expected and unexpected inflation. The signs of the regression coefficient indicate whether an asset is a positive hedge or a negative hedge against inflations. An asset offers a partial hedge against the respective inflation if the coefficient is found to be less than 1.0, but statistically distinguishable from 0. For asset that has a coefficient that is statistically higher than 1.0, the asset hedges against not only the inflations on a one-to-one basis, it also offers additional hedge against the inflation risks of other assets in the individual portfolio. 7

9 4.2. Data Sources The data employed in this study comprise the Urban Redevelopment Authority s (URA) all-property price index and the related sub-indices: the residential property price index, the office property price index, the shop property price index and the industrial property price index, the Stock Exchange of Singapore s (SES) all-share index, the SES all-property share index, the consumer price index (CPI) and the 3- month Treasury bill rates. The 1978 to 1987 s Treasury bill rates data are available in the Monetary of Authority s Quarterly Statistical Bulletin, whereas all other data are obtained from the TRENDS database of the Statistic Department of Singapore. To be consistent with other property indices, the CPI, the Treasury bill rates and the stock indices are converted from monthly to quarterly series. The URA property price indices are generated from actual transactions lodged as caveats with the Registry of Titles and Deeds. The data are not subject to smoothing problems inherent in the appraisal-based data used in many overseas studies. The SES all-share index represents the price movement of all listed Singapore-incorporated and Singapore-currency traded companies on the Stock Exchange of Singapore. The SES has also compiled six sub-indices categorized by industrial. The SES all-property price index is the closer proxy for benchmarking the performance of the securitized property assets. The nominal asset returns (R j ) are calculated as the first difference of the log-prices of the assets, i.e. R jt = ln (P jt / P jt-1 ) Measures Of Inflation Rates Two most commonly used measures of inflation are the consumer price index (CPI), or the retail price index in some countries, and the gross domestic product (GDP) deflator. The CPI is a Laspeye-based index where the average inter-temporal change in prices is measured based on a fixed basket of goods and services consumed by households as surveyed in the based year. 3 The GDP deflator, on the other hand, is a relative measure of the aggregate net output or the value-added of the economy between the current year and the base year. The GDP deflator measures a wider range 3 The basket of goods and services comprises food, clothing, housing, transport and communication and miscellaneous, and it is kept fixed so that changes in the index reflect only price changes and not changes in the type of goods and services purchased. 8

10 of goods and services, and the base of the outputs may change when there is a structural change to the economy. In our study, the CPI is used as a proxy of inflation rate. The actual rate of inflation ( t ) is computed as a simple difference in the price level over the price in the previous period, which is given as, t = (CPI t CPI t-1 ) / CPI t-1 (3) Expected inflation, as the name implies, is a rate that is formulated based on an individual expectation and also the information that are available in the previous period (t-1). It captures the change in the purchasing power expected at the beginning of the period (Hamelink, Hoesli and McGregor, 1997). Unexpected inflation, on the other hand, reflects the random errors observed between the actual and the expected inflations. The error terms are caused by the inefficiency of the market where not all the past information is processed by the market. There are different approaches proposed to represent the expected and unexpected inflations (Fama and Gibbons, 1982; Gultekin, 1983; Gibbons, 1984; Hartzell, Hekman and Mile, 1987; Limmack and Ward, 1988; Harvey, 1989; inter-alia), and all have their own merits and demerits. A simple way of representing the expected inflation proposed by Fama and Schwert (1977) is to use the lagged period 3-month Treasury bill as a direct proxy. For the unexpected inflation (U jt ), we follow the approach of Fama and Schwert (1977) by determining the unexpected inflation as the difference between the actual ~ inflation and expected inflation, i.e. U jt = [ t E( t φ t 1 )], where t is the actual ~ inflation and E φ ) is the 3-month Treasury bill rate fixed in the previous period. ( t t Descriptive Statistics For Asset Returns & Inflation Rate Table 1 shows the historical statistics of the asset returns and inflation rate for the sample period from 1978 to The results show that property assets did better than both stocks and treasury bills in terms of their mean quarterly returns. Residential property performed the best with the highest mean quarterly return of 2.48%. The 9

11 other three property assets performed below the average mean property market return (all-property) of 2.24%. However, the nominal rates of return: office (2.08%), industrial (1.94%) and shop (0.87%), exceed the quarterly mean rate of inflation of 0.62%, which suggest a positive real return for all property assets. In comparison, stocks and treasury bills seemed to have poor or negative hedging against inflation as their rates of return are all less than the inflation rate. The quarterly mean returns of the SES all-share and all-property share are 0.07% and 0.23% respectively. The 3- month Treasury bill has a negative mean quarterly return of 0.92%. The standard deviation for treasury bills was the highest at 35.08%, followed by the standard deviations of stocks at 17.91% and property at 6.76%. Based on the descriptive statistics, it appears that property assets with the lowest volatility and the highest mean rate of returns are the better investments for hedging inflation compared to stocks. The Pearson rank correlation coefficients between the asset returns and the inflation rate are summarized in the last column of Table 1. The coefficients are all positive and significant at a 5% level with the exception of the SES all-share and the 3-month T-bills. Property in general moves closely with the changes in inflation. The URA allproperty index has the highest correlation coefficient of with the inflation, and it follows in a descending order by industrial property (0.453), residential property (0.444), office (0.406) and shop (0.222). The correlation between the securitized property and inflation is also significant with a coefficient of In comparison, the Pearson rank correlation coefficients for the SES all-share and the 3-month T-bills are insignificant at a 5% level Empirical Model Specifications In the Fischer s hypothesis, the expected nominal rate of return is the sum of the expected real rate of return and the expected inflation. If the market is efficient, the expected real return should be independent of the expected inflation. Based on this theoretical framework, Fama and Schwert (1977) develop the following testable empirical model specification as given in equation (2), which also includes an unexpected inflation component, 10

12 ~ R jt = α j + β j ) ~ E φ + γ j [ )] ( t t 1 t E( t φ t 1 + ε jt (2 ) The unexpected inflation component is included to take account of the unanticipated shocks when the market could not respond efficient enough to new market information. As the actual inflation at time t is the sum of the expected and ~ unexpected inflations at time t, i.e. t = E φ ) + U t, we also test the regression of ( t t 1 the nominal returns on the actual inflation without decomposing it into the expected and unexpected elements as follows, R jt = α j + δ j t + ε jt (4) The two regression models specified above are susceptible to the violation of the independent and uncorrelated error assumption, i.e. the error term is auto-correlated. This problem is particularly severe when time-series data, which are non-stationary in levels, are employed in the regression estimation. Our preliminary empirical estimations have shown significant auto-correlation in the models as indicated by relatively low Durbin-Watson (DW) statistics. 4 In our model, the autocorrelations in the error term are corrected by including k-lagged error terms in an autoregressive process in order for the error term to become a white-noise process. The autoregressive error adjustment process is carried out in the Eview program as follows, k it = ai + i= 1 ε b ε + η i t i i (5) The Fama and Schwert (1977) test (equation 2 ) is repeated for different sub-periods and also for periods of high and low inflations. The results of the analyses are summarized in next section. 4 The results of the preliminary analyses are not included in this paper, and they would be made available upon request. 11

13 5. Analysis of Results This section presents the empirical findings of the inflation hedging tests for different asset classes, which include four different property types (residential, office, shop and industrial), securitized property, stocks and short-term government treasury bills. First, the results of the multiple regression tests covering the entire sample period from 1978 to 1998 are summarized. The tests examine inflation hedging ability of the sample assets independently against the actual inflation (equation 4) and also against both the expected and unexpected inflations. Next, the sample period is divided into four sub-periods of 5 years each. The tests given by equation (2) are repeated for each sub-period and the results are discussed. Lastly, we proceed on to examine the differences in inflation hedging ability of assets between the high and the low inflation periods Hedging Against Actual Inflation The regressions given by equation (4) test the hedging ability of assets against actual inflation and the results are summarized in Table 2. The regression coefficients of the actual inflation variable, δ j, are all positive, but only two of the regression coefficients: shop and industrial property, are statistically significant at 5% level. Residential property return has the lowest δ j coefficient of while the SES allproperty share return has the highest coefficient of Although the inflation coefficients of the stock market models: all-share and all-property share, are higher compared to other asset models, they are not statistically significant. In other words, the securitized assets are not good hedges against actual inflation. The regression results reject the inflation-hedging hypothesis for all the assets except for the shop and industrial property. Shop (δ j = 2.863) and industrial property (δ j = 2.813) returns provide more than perfect hedge against actual inflation, and their returns significantly hedge more than one-to-one corresponding variation in the actual inflation Hedging Against Expected & Unexpected Inflation The tests of inflation hedging against expected and unexpected inflations are conducted by running the empirical model given by equation (2 ). The results of the 12

14 regressions as summarized in Table 3 indicate that the coefficients for the expected inflation (β j ) and the unexpected inflation (γ j ) variables range from to and to respectively. Residential property return (β j = and γ j = 0.635) has the lowest coefficients while the Treasury bill return (β j = and γ j = 7.431) has the highest coefficients, but they are not statistically significant at 5% level. Overall, the results are consistent with those found in the tests against actual inflation. The securitized assets, despite the high coefficients, fail to reject the null hypothesis statistically. Thus, they offer no hedge against both expected and unexpected inflations. Shop and industrial property are two assets that provide significantly strong hedges against expected inflation (β PSHOP = and β PIND = 2.938). Industrial property, besides hedging against the expected inflation, also provides more than oneto-one corresponding hedge against the variation in the unexpected inflation yearly Sub-period Analysis In this sub-period analysis, the entire sample period of 21 years is divided into four sub-periods consisting of a 5-year interval for the first three sub-periods and a 6-year interval for the last sub-period. This sub-period analysis is undertaken to uncover any hidden trends and also to further breakdown the inflation hedging characteristics of the property into smaller sub-intervals. For the ease of tabulation, we denote the period between 1978 and 1982 as I ; the period between 1983 and 1987 as II ; the period between 1988 and 1992 as III, and the period between 1993 and 1998 as IV. The results of the R 2 and the Durbin-Watson (DW) statistics are given in the upper panel of Table 4. The goodness of fit of the regression models as indicated by the R 2 values vary widely from (Property) to (Office). The Durbin- Watson statistics show no serial correlations since they have been fully eliminated through the auto-regressive process. The number of lag used for the autoregressive error terms is about 3. The sub-period analysis as shown by the lower panel of Table 4 reveals interesting inflation-hedging characteristics of different assets. The results show that industrial property is the only asset that offers consistent positive hedges against both expected and unexpected inflations. All other assets have mixed results with positive and negative coefficients recorded for different sub-periods. In the sub-period comparison, 13

15 the period IV ( ) appears to be the most significant period for hedging where two property (all-property and residential) and two non-property (all-stock and securitized property) assets are found to have significant positive hedges against expected and unexpected inflations. Period III is the worst period for inflation hedging as none of the coefficients are statistically significant at 5% level. In the first sub-period of 1978 to 1982, industrial property (β j =4.499 and γ j = 3.557) is the best choice for a portfolio that requires positive hedges against expected and unexpected inflation. Office (β j = 3.232) is also a good asset for hedging expected inflation during period I. The positive hedging ability of office increases in period II with a higher coefficient of at 5% significance level. For the hedge against expected inflation, three assets: shop (β j =-8.709), all-share (β j = ) and unsecuritized property (β j = ), show significant but negative coefficients at 5% level. The period II experienced the economic recession with an output growth of 1.6% in 1985 and also the black Monday crisis that hit the stock market as evidenced by a drastic 53.49% decline in the all-share indices in the 4 th quarter of These are believed to be the contributing causes for the perverse hedges against the expected inflation that are reflected by the coefficients of the assets. The Sub-period III from 1988 to 1992 witnessed the recovery of the economy, but it seems like the rates of recovery in the asset prices were not able to catch up with the inflation rate changes. None of the assets shows significant results in the inflationhedging tests during this period. In the last sub-period from 1993 to 1998, property stock returns provide the best inflation hedges against both expected (β j = ) and unexpected (γ j = ) inflations. A strong inflation hedging performance was also indicated by the all-stock returns during this period. The impressive results of the allstock and all-property stock indices, which peaked in the 4 th quarter of 1993 with quarter-to-quarter returns of 15.95% and 28.88% respectively, coupled with the stabilization of the inflation rates of the period, are the main factors for the high coefficients obtained for the two non-property assets. 5 The all-property share index has also dropped by 19.48% in the fourth quarter of

16 5.4. High & Low Inflation Periods Analysis In this section, the periods of high and low inflation were divided at the median inflation rate from 1978 to Sample periods with inflations above the median are classified into the high inflation group while those samples with inflations below the median are classified into the low inflation group. For the entire study period, 42 periods are in the high inflation group while 41 periods are in the low inflation group. We then apply the same regression model specification given by equation (2) to analyze the inflation hedging characteristics of the assets in these two sub-periods. The regression results as summarized in Table 5 show that assets in general performed better in the low inflation period by looking at the magnitudes of β j and γ j coefficients. However, the results with the exception of office and residential returns are all statistically insignificant at 10% level. Office sector indicates strong and significant, but perverse relationships with the expected (β j = ) and unexpected inflations (γ j = ). Whereas, residential property shows a strong hedge against the unexpected inflation in the low inflation period, whereby a 1% increase in the unexpected inflation leads to 14.84% increase in the residential property price. During the high inflation periods, industrial property with significant and positive regression estimates is the best choice for investor. The industrial property returns offer hedges of more than 6-time for expected inflation and more than 8-time for unexpected inflation. For the non-property assets, the results are insignificant at 10% level and thus the hypothesis of zero inflation hedging characteristics cannot be rejected. However, for the risk-free Treasury bills, the coefficients are higher in the high inflation period compared with those in the low inflation period. If the coefficients have been significant, it would have implied that the 3-moth Treasury bill is a safe and effective option for hedging inflation, when all other assets under-perform the inflation during the high inflation period. 15

17 6. Conclusion Our study provides empirical evidence to verify the traditional believe that property is a good hedge against inflation. Using the empirical data of Singapore markets, our results did not reject the hypothesis that property is a good hedge against inflation. The inflation hedging performance of property in general is found to be better than the non-property assets: stock, securitized property and Treasury bills. Industrial property is found to be the most effective hedge against both expected and unexpected inflation, whereas shop offers significant hedge only against the expected inflation. The returns of the two assets establish not only a perfect one-to-one correspondence relationship with the inflation rate they also increase at a faster rate than the increase in the inflation rate. Therefore, industrial property and shop, when included in an investor s portfolio, would provide sufficient hedge against its own inflation risk as well as the inflation risks of other assets in the portfolio. The results of the 5-yearly sub-period analysis show that except for industrial property that offers consistent positive hedges against both expected and unexpected inflations. All other assets have a mixture of both positive and negative coefficients for different sub-periods. In other words, while an asset is a good hedge against inflation in one period, it can also turn into a negative inflation-hedge in another period. It thus implies that the points of entry or exit of investment are important in determining the effectiveness of inflation-hedging characteristics of the assets. In the sub-period analysis, the period IV ( ) appears to be the most significant period for hedging inflation where two property (all-property and residential) and two nonproperty (all-stock and securitized property) assets are found to have significant positive hedges against expected and unexpected inflations. Period III has the worst inflation hedging record, as none of the coefficients are statistically significant at 5% level. When the level of inflation is controlled in the analysis, i.e. by dividing the sample period into high and low inflation periods, it was found that more assets perform better in the low inflation period than in the high inflation period. Residential property shows a strong hedge against unexpected inflation in the low inflation period. Therefore, it supports the hypothesis that asset returns offer better protection or 16

18 hedging against inflation in the low inflation periods. This hypothesis, however, does not apply to office sector as the results indicate that office returns are significantly but negatively correlated with both the expected and unexpected inflation. The only asset that offers strong hedge against both expected and unexpected inflations during the high inflation regime is industrial property. Its returns increase more than one-to-one increases in the inflations. For the non-property assets, namely stocks, securitized property and Treasury bills, our findings are consistent with the general perception that they offer insignificant hedges against inflation with the exception of the sub-period of 1993 to The SES all-share and all-property share indices have produced excellent inflationhedging records in the sub-period of 1993 to 1998, as a results of the stock market boom and optimisms, coupled with the stabilization of the inflation rates during the periods. Therefore, in conclusion, we could not reject the hypothesis that property assets in Singapore are generally better hedges against inflation compared to nonproperty assets. 17

19 Reference: Brueggeman, W.B., A.H. Chen and T.G. Thibodeau Real Estate Investment Funds: Performance and Portfolio Considerations. AREUEA Journal 12(3): Coleman, M., S. Hudson-Wilson and J. R.Webb Real Estate in the Multi-Asset Portfolio. S.Hudson-Wilson and C. H. Wurtzebach, editors. Managing Real Estate Portfolios. Burr Ridge, IL and New York: Richard D. Irwin, 3: ,. Fama, E. F., and G. W. Schwert Asset Returns and Inflation. Journal of Financial Economics 5(2): Fisher, I The Theory of Interest. MacMillan, New York. Ganesan, S. and Y H. Chiang The Inflation-Hedging Characteristics of Real and Financial Assets in Hong Kong. Journal of Real Estate Portfolio Management 4(1): Gultekin, N.B Stock Market Returns and Inflation: Evidence from Other Countries. Journal of Finance 38(1): Hamelink, F., M. Hoesli and B. MacGregor Inflation Hedging Versus Inflation Protection in the U.S. and the U.K. Real Estate Fiannce Hartzell, D., J. Hekman, and M. Miles Real Estate Returns and Inflation, Journal of the American Real Estate and Urban Economics Association 15(1): Hoesli, M Real Estate as Hedge Against Inflation: Learning from the Swiss Case. Journal of Property Valuation and Investment 12(3): Hoesli, M., B. D. Macgregor, G. Matysiak and N. Nanthakumaran The Short- Term Inflation Hedging Characteristics of U.K Real Estate. Journal of Real Estate Finance & Economics 15(1): Limmack, R.J. and C.W.R. Ward Property Returns and Inflation. Land Development Studies 5(3): Liu C. H., D. J.Hartzell and M. E. Hoesli International Evidence on Real Estate Securities as an Inflation Hedge. Real Estate Economics 25(2): Matysiak, G., M.Hoesli, B. MacGregor and N. Nanthakumaran The Long-term Inflation-Hedging Characteristics of UK Commercial Property. Journal of Property Finance 7(1): Miles, D Property and Inflation. Journal of Property Finance 7(1): Newell, G Is Canadian Real Estate a Hedge Against Inflation? The Canadian Appraiser Newell, G The Inflation-Hedging Characteristics of Australian Commercial Property: Journal of Property Finance 7(1): Newell, G. and T. Boyd Inflation-Hedging Attributes of New Zealand Commercial Property. New Zealand Valuers' Journal Ruben, J., M. Bond and J. Webb The Inflation-hedging Effectiveness of Real Estate. Journal of Real Estate Research 4: Wurtzebach, C.H., G.R. Nueller and D. Machi The Impact of Inflation and Vacancy on Real Estate Returns. Journal of Real Estate Research 6(2):

20 Table 1: Quarterly Asset Returns & Inflation Rate ( ) VARIABLE Mean Median Max. Min. Standard Deviation Correlation# All-Property R t (PALL) *** Return Residential R t (PRES) *** Property Return Office Property R t (POFF) *** Return Shop Return R t (PSHOP) ** Industrial Property Return SES All-Share Return SES All- Property Return Treasury Bills Return Actual Rate of inflation R t (PIND) *** R t (SALL) R t (SPRO) ** R t (TREA) t # Results of the Pearson correlation coefficient *** Indicates correlation is significant at the 1% level (2-tailed) ** Indicates correlation is significant at the 5% level (2-tailed) 19

21 Table 2: Regression Results for Actual Inflation Dependent Variable α j δ j R 2 DW Statistics R t (PALL) (0.709) (1.020) R t (PRES) (0.611) (0.979) R t (POFF) (0.438) (0.936) R t (PSHOP) ** (-0.778) (2.346) R t (PIND) ** (0.122) (2.367) R t (SALL) (-0.910) (1.232) R t (SPRO) (-1.016) (1.606) R t (TREA) (-0.456) (0.337) t-statistics are given in the parentheses. ** Significant at 5% level. 20

22 Table 3: Regression Results for Expected & Unexpected Inflations Dependent Variable α j β j γ j R 2 DW statistics R t (PALL) (1.034) (1.212) (1.461) R t (PRES) (1.089) (0.797) (0.988) R t (POFF) (-0.445) (1.471) (1.240) R t (PSHOP) * (-1.335) (1.857) (1.530) R t (PIND) ** 2.842** (-0.256) (2.533) (2.350) R t (SALL) (-0.668) (1.231) (1.105) R t (SPRO) (-0.554) (1.568) (1.481) R t (TREA) (1.509) (1.346) (1.599) **Significant at 5% level t-values are given in parentheses. * Significant at 10% level 21

23 Table 4: 5-Yearly Sub-Period Regression Results Dependent Durbin-Watson (DW) Statistics Variable I II III IV I II III IV R t (PALL) R 2 R t (PRES) R t (POFF) R t (PSHOP) R t (PIND) R t (SALL) R t (SPRO) R t (TREA) Dependent β j γ j Variable I II III IV I II III IV R t (PALL) (0.808) (-0.597) (0.469) 4.452** (2.062) (0.758) (0.960) (0.571) 4.148* (1.912) R t (PRES) (0.636) (-0.366) (0.239) 4.882** (2.026) (0.584) (1.195) (0.428) 4.626* (1.915) R t (POFF) 3.232* (1.887) (1.343) (-0.034) (-1.415) (0.723) (2.157) (-0.057) (-1.142) R t (PSHOP) (0.769) ** (-2.676) (1.590) (0.664) (0.152) (-1.104) (0.826) (0.429) R t (PIND) (2.418)* (0.103) (0.346) (0.935) 3.557* (1.912) (1.386) (0.486) (0.820) R t (SALL) (-0.034) ** (-3.399) (-0.480) (3.233)* (-0.382) (-1.458) (-0.515) ** (3.761) R t (SPRO) (0.192) ** (-2.828) (0.258) ** (5.040) (-0.319) (-1.160) (0.426) ** (6.331) R t (TREA) (0.035) (-1.022) (-0.285) (-0.401) (0.277) (-0.165) (-0.112) (0.065) ** Significant at 5%level * significant at 10% level t-values in parenthesis Estimates of α j are not included in the table. 22

24 Table 5: Regression Results for High & Low Inflation Periods Analysis Dependent High Inflation Low Inflation Variable β j γ j R 2 DW β j γ j R 2 DW R t (PALL) (1.282) (1.381) (1.420) (1.764) R t (PRES) * (0.790) (0.924) (1.500) (1.871) R t (POFF) ** ** (0.702) (0.769) (-1.970) (-2.013) R t (PSHOP) (1.222) (1.189) (0.523) (0.798) R t (PIND) 6.196** 8.131** (2.561) (3.385) (1.200) (1.123) R t (SALL) (-0553) (-0.647) (0.597) (0.826) R t (SPRO) (-0.086) (-0.132) (0.581) (0.940) R t (TREA) (1.443) (1.549) (1.598) (1.762) ** Significant at 5%level * significant at 10% level t-values in parenthesis Estimates of α j are not included in the table. 23

25 4% Figure 1: Inflation Rate of Singapore for the period 1978 to % Inflation Rate(%) 2% 1% 0% Q -1% Q Q Q Q Q Q Q Q Q Q Year -Qtr -2% 24