Dynamics of private industrial space demand in Singapore

Abstract This study empirically examines the dynamics of the private industrial market in Singapore using a Vector Error Correction Model (VECM), which is derived based on the theoretical framework of an extended accelerator investment model. The GDP in manufacturing sector (LMGDP) and the composite leading indicator (LCLI) were two unrestricted long-run forcing variables included in the VECM for the industrial space demand. In the generalized forecast error variance decomposition analysis, one-standard deviation shocks to the manufacturing GDP (LMGDP) was found to account for an average 67.10% of the variances of the private industrial space demand (LPRD). It was also found that the most volatile impulse responses from the industrial demand variance were inflicted by one-standard error shocks on the ecm and the manufacturing GDP terms.In Singapore, industrial real estate demand or take up in both private and public markets has been predicted in an adaptive process by looking at the absorption rate in the previous periods. The supply of new industrial real estate stocks was adjusted in a piece-meal basis with the objective of meeting the demand projection, which would presumably not differ substantially from the previous year's take-up.1 Koh (1987) undertook a rigorous study of industrial real estate stock demand using a two-stage least squares model. He found significant relationships between the industrial space demand and economic variables like industrial output, gross domestic product, manufacturing employment and wage, interest rate and investment commitment. In Koh's model, the positive effects of industrial output on the industrial real estate demand were indirectly channeled via the employment equation. There were also other unpublished researches that used Shenkel's (1965) employment model (Lam, 1984) and multiple regression models (MRM) (Chan, 1976; and Mao, 1977)2 to determine the industrial real estate stock investment in Singapore.

There were no direct tests of the acceleration effect of the industrial output on the industrial real estate stock in Singapore. The earlier empirical studies that tested the flexible accelerator models were based primarily on the industrial real estate data in the United States and the United Kingdom (Nicholson and Tebbutt, 1979; Wheaton and Torto, 1990; Giussani and Tsolacos, 1994; and Tsolacos, 1995). They all found positive relationships between industrial real estate stock adjustments and manufacturing outputs. When the test of the acceleration effects is done on the industrial real estate market in Singapore, two factors need to be taken into consideration. First, Singapore has gone through different phases of industrialization than the U.S. and the U.K. (Zhu, 2000).3 For manufacturing firms that move up the technological ladder, more physical real estate stock would be substituted for investments in advanced productive capitals. As a result, the level of production output may increase at the expense of smaller real estate capitals. Second, the constraint of land resource in Singapore may create inelasticity in the supply of physical industrial real estate stocks. Firms may, therefore, be induced to invest and procure more physical real estate stocks than required during the weak output cycle. They could then optimize the excess real estate stocks when the output cycle rebounds.

The technology-related substitution and the land resource constraint factors may create negative acceleration effects on the industrial real estate stock adjustment process. This study thus attempts to test the dynamics of private industrial real estate stock demand in Singapore using an expanded flexible accelerator model. This article is organized as follows. First, the empirical literature on industrial real estate is reviewed, which helps sets the objectives of the study. Next, the salient features of the structure and the stock-flow process of the industrial market in Singapore is discussed. This is followed by a discussion of the theoretical accelerator model that incorporates other economic variables. Next, the empirical methodology and data are discussed. An industrial real estate demand function that comprises both the long run correction and short run fluctuations of the economic variables is formulated using a vector error correction model (VECM) with lag terms. Based on the proposed VECM structure, the post estimation analyses that include in-the-sample forecasting, impulse response and variance decomposition of the shocks to exogenous variables are discussed. The final section presents the conclusions.

Singapore's Private Industrial Real Estate Market

For a land and resource scarce country, Singapore has been transformed into one of the fastest growing economies in the world in the thirty-five years since independence. The main engine of growth has been the manufacturing sector, which accounted for 25.17% of the total gross domestic product (GDP) on average for the periods from 1985:1 to 1999:4. Currently, the land allocated for industrial use constitutes 12.2% of the total available land,4 of which approximately 76.2% (as in 1985) was made up of state lands.5 The state lands for industrial use are supplied to the market via two channels: inter-government agency alienation and the government's land sale program. Industrial lands are alienated to government agencies involved in industrial development, which include Jurong Town Corporation (JTC), Housing Development Board (HDB), Port Authority of Singapore (PSA) and Land Transport Authority (LTA). Private developers or manufacturing firms are also allocated industrial lands via the land sale by tender program administered by the Urban Redevelopment Authority (URA). Privately owned industrial lands are another source of industrial real estate supply. Private industrial lands are mainly held in 99-year, 999-year or freehold tenures, whereas the industrial lands coming through the state-channels are either via the land sale or alienation arrangement on shorter leasehold tenures of either 30- or 60-years.