Estimating the Willingness to Pay of Industrial Firms for Japanese Industrial Parks

Abstract

This study aims to investigate the determinants of the willingness to pay (WTP) of industrial firms and the sales price of public companies for industrial parks, and the gap between the two. The sales price is estimated using ordinary least squares (OLS), while WTP and the gap are estimated using random-WTP probit models and disaggregated sales data of Japanese industrial parks between 2002 and 2007. Testing the model reveals two important differences between firms’ WTP and sales price. First, the sale price is decided in order to cover the development cost since only the sale price is influenced by the average land price of each region while WTP is not. Secondly, the current demand of the industrial sector is insufficiently represented in the sales price; this is because firms’ WTP is significantly more sensitive than suppliers’ sales prices in terms of accessibility to both large and small cities and agglomeration economies.

1. Introduction

After World War II, industrial parks established by national and local public companies played a crucial role in the rapid development of the Japanese manufacturing sector. These cheap and easily obtainable industrial parks with complete infrastructure have not only fulfilled the growing demand of manufacturing companies for industrial land, but have also supported the policy for encouraging investment in less-developed regions. However, since the mid 1980s, Japanese-yen appreciation resulting from the Plaza Accord has led to the relocation of domestic plants to growing Asian countries like China and Thailand; this development has also led to an increase in the number of unsold industrial parks. According to Akai (2005), unsold industrial parks belonging to land development corporations (public companies established by local governments) have become non-performing assets; however, financial difficulties faced by local governments and problematic accounting systems of public companies have postponed the possible settlement of debts arising from such unsold property through appropriate discounts and the redevelopment of property.¹

However, in recent years, a new demand for domestic industrial locations has emerged in an industrial tidal wave of horizontal specialisation within East Asia and integration of multiple domestic plants into headquarter plants with R&D facilities. It is important for these plants to have accessibility to urban areas and transport facilities, which, in turn, will provide them with access to human capital and business partners;² hence, there is a growing demand for new industrial parks that satisfy these needs and for a revision in the price of existing parks.

In the present study, we conduct an empirical investigation of industrial parks provided by national and public companies in Japan, focusing on both the demand and supply sides independently. Our questions are twofold. What is the most important factor for the new industrial location demand? To what attributes and to what extent does the price setting of the supply side diverge from the current demand? In other words, our focus is to ascertain the current requirements of industries and determine how the price setting and development strategy of the public sector diverges from these demands. The goal of this paper is to provide the basic implications of and information for discussion on the fundamental policy direction for the public provision of industrial parks.

Substantial existing literature that has estimated the value of industrial estate traded within the private sector has revealed some important factors determining these values: the location and facilities of industrial estates (Ambrose, 1990; Kowalski and Paraskevopoulos, 1990), amenities for workers such as shopping and education (Sivitanidou and Sivitanides, 1995) and environmental aspects such as contamination by previous owners and remediation of it (Jackson, 2002). Although most of these studies estimated the market price or rent of industrial estate using the hedonic approach and other methodologies (Lockwood and Rutherford, 1996), some studies like Blackley (1985) particularly highlighted detailed attributes of individual industrial plants and their bid-rent for industrial estate.

While the price of industrial estate traded within the private sector is constituted of the firms’ expected profit from it, the sales price of industrial parks announced by the public sector is significantly influenced by the intention of supply-side factors such as industrial policy, cost of development and the management system of public companies. Therefore, merely estimating prices themselves does not provide correct information concerning the demand of industrial firms for industrial parks. For example, Saz-Salazar and García-Menéndez (2005) and Lin and Ben (2009) investigated Spanish and Taiwanese industrial parks respectively and showed that the sales price of industrial parks is also determined by attributes similar to those of private land, such as accessibility to cities and highways; however, they also showed that publicly provided parks are cheaper than private parks. In the case of Japan, it is also suspected that the price of public industrial parks diverges from the market price because of cost-covering and inflexible price setting by public companies, which in turn is caused by their distorted accounting system.

The current study estimates the WTP of the industrial sector for industrial parks, independent of an OLS estimation for their sales prices, and estimates the gap between the WTP and the sales price. Our empirical methodology for capturing WTP is identical to a dichotomous-choice contingent valuation method (CVM) approach based on questionnaire data.³ To the best of our knowledge, the present paper is the first to apply this method to the actual sales data of industrial parks and capture the supply–demand gap. Although the random bid-rent model given by Blackley (1985) for estimating the bidding price of industrial firms is actually similar to our model, his study treated non-public industrial estate and did not mention the supply–demand gap. Moreover, he focused on firms’ attributes, while we focus on the attributes of land.

The remainder of the paper is organised in the following manner. Section 2 presents a random WTP model and its estimation methods. Section 3 provides a brief description of the data on Japanese industrial parks and independent hypotheses for sales price and WTP. Section 4 presents the results of the estimation and provides brief policy implications for industrial parks. Section 5 provides concluding remarks.

2. Model

The present section presents a decision-making model for firms purchasing industrial parks and the methods of estimating the WTP function of the firms.

Suppose that there are industrial parks with at least one vacant block, which are managed by public companies, and that potential manufacturing firms are searching for sites for their new plants. The public companies set and announce the sales price of each industrial park, which is based on the balance of payment for each individual business but which might be affected by various matters, like the regional development policy of the local government. On the other hand, the manufacturing firms have a random WTP for each industrial park, based on the expected profit from operations at the park, and they are willing to purchase a block of the industrial park if and only if the firms have a higher WTP for the park than the sales price. Hereafter, at least one block of an industrial park is sold if and only if the highest WTP of the potential firms is higher than the price. What the present study highlights is the maximum WTP of potential firms rather than that of individual firms; this assumption is similar to that employed in previous random bid-rent models.

Now we suppose that the sales price of an industrial park, j, is determined as

P_{j} = κ + γ X_{j} + ϖ_{j}

and that the maximum random WTP function of manufacturing firms to j is described by

V_{j} = α + β X_{j} + ε_{j}

where, $X_{j} = (X_{j 1}, \dots, X_{jM})$ , $β = (β_{1}, \dots, β_{M})$ and $γ = (γ_{1}, \dots, γ_{M})$ are vectors of the attributes of j and their coefficients respectively; $ω_{i}$ and $ε_{i}$ are normally distributed random variables whose standard deviations are $θ_{ϖ}$ and $θ_{ε}$ respectively.

With these assumptions, at least one block of an industrial park is purchased if and only if $V_{j} \geq P_{j}$ holds. Therefore, if an industrial park has price $P_{j}$ and if at least one block of it is sold (or unsold) in a period, the probability that these two phenomena simultaneously appear is described as follows

\begin{matrix} Pr (κ + γ X_{j} + ω_{j} = P_{j}) Pr (V_{j} < P_{j}) \\ if t_{j} = u \\ Pr (κ + γ X_{j} + ω_{j} = P_{j}) {1 - Pr (V_{j} < P_{j})} \\ if t_{j} = s \end{matrix}}

(1)

where, $t_{j} = {s, u}$ is a state of park j; $s$ means j has sold more than one block, while $u$ means nothing was sold during the period.

When we denote the density function of $ϖ_{j}$ and the accumulative distribution function of $ε_{j}$ as $f_{ω} (\cdot)$ and $F_{ε} (\cdot)$ respectively, equation (1) is rewritten as follows

\begin{matrix} f_{ω} (P_{j} - κ - γ X_{j}) F_{ε} (P_{j} - α - β X_{j}) \\ if t_{j} = u \\ f_{ω} (P_{j} - κ - γ X_{j}) {1 - F_{ε} (P_{j} - α - β X_{j})} \\ if t_{j} = s \end{matrix}}

(2)

We estimate the above probit model by maximising the following likelihood function by $α$ , $κ$ , $β$ and $γ$

\begin{matrix} L = \sum_{j = 1} [\ln f_{ω} (P_{j} - κ - γ X_{j}) + {φ_{j}}_{u} \ln F_{ε} (P_{j} - α - β X_{j}) + {φ_{j}}_{s} \ln {1 - F_{ε} (P_{j} - α - β X_{j})}] \\ where, {φ_{j}}_{t} = {\begin{matrix} 1 & if t = t_{j} \\ 0 & otherwise \end{matrix} \end{matrix}

(3)

However, since $P_{j}$ is observable in the data, the sales price of the supply side and the WTP of the demand side in equation (3) are additive separable; hence, these two objective variables are estimated independently. That is, the sales price and WTP are estimated by maximising the following independent likelihood functions, $L_{1}$ and $L_{2}$ respectively

L_{1} = \sum_{j = 1} [\ln f_{ω} (P_{j} - κ - γ X_{j})]

(4)

\begin{matrix} L_{2} = \sum_{j = 1} [{φ_{j}}_{u} \ln F_{ε} (P_{j} - α - β X_{j}) \\ + {φ_{j}}_{s} \ln {1 - F_{ε} (P_{j} - α - β X_{j})}] \end{matrix}

(5)

Since $ϖ_{j}$ is normally distributed, the estimation of (4) perfectly coincides with the OLS estimation of $P_{j}$ . In equations (4) and (5), the standard deviations of $ε_{j}$ and $ϖ_{j}$ are also endogenous. When $F (x; μ, θ)$ denotes the cumulative density function of a normally distributed variable with average $μ$ and standard deviation $θ$ , $F (x / θ; μ / θ, 1) = F (x; μ, θ)$ holds; hence, equation (5) can be transformed as follows

L_{2} = \sum_{j = 1} [{φ_{j}}_{u} \ln F (ϕ P_{j} - α' - β' X_{j}) + {φ_{j}}_{s} \ln {1 - F (ϕ P_{j} - α' - β' X_{j})}]

(6)

where, $ϕ = 1 / θ_{ε}, α' = α / θ_{ε}, β' = β / θ_{ε}$ . After estimating $α'$ , $β'$ and $φ$ in order to maximise equation (6), we can estimate parameters $α$ , $β$ by dividing $α'$ , $β'$ by $φ$ .

The present study employs this estimation and estimated the t-value for each parameter and maximum likelihood ratio for the entire model. We set a null hypothesis for equation (6) for all parameters, including $φ$ , to be 0. This means that none of the variables, including price, ever affects the determination of firms to purchase industrial parks. However, if we set the same hypothesis to equation (5), it is described as $θ_{ε} \to \infty$ ; hence, it is impossible to calculate the likelihood under the null hypothesis.

Furthermore, we estimate a gap of decision-making between price setting by the suppliers and WTP by the firms. For that purpose, we translate equation (6) as follows

\begin{matrix} L_{2} = \sum_{j = 1} [{φ_{j}}_{u} \ln F (ϕ (P_{j} - {\hat{P}}_{j}) - α'_{d} - {β'}_{d} X_{j}) \\ + {φ_{j}}_{s} \ln {1 - F (ϕ (P_{j} - {\hat{P}}_{j}) - α'_{d} - {β'}_{d} X_{j})}] \end{matrix}

(7)

${\hat{P}}_{j} = \hat{κ} + \hat{γ} X_{j}$ is the hedonic estimate of the sales price yield from estimating equation (4); hence, $P_{j} - {\hat{P}}_{j}$ is the hedonic residual. Also, $α_{d}' = (α - κ) φ$ and $β_{d}' = (β - γ) φ$ represent gaps in the parameters between the function of the sales price and that of firms’ WTP. If a coefficient $β_{dk}'$ is statistically significant, we can conclude that suppliers of land and firms have different criteria in evaluating for attribute k. These parameters representing gaps may be estimable by applying the maximum likelihood method to equation (7). We should note that the $φ$ from equations (6) and (7) coincides since these two equations are entirely the same. Also, the values of $α' = α_{d}' + κ φ$ and $β' = β_{d}' + γ φ$ , calculated by the results of equations (7) and (4), perfectly coincide with the value estimated from equation (6); thus, the maximised likelihood value of these two estimations are naturally the same.

3. Data and Hypotheses

3.1 Hypotheses on WTP

Our basic assumption on WTP is identical to the bid-rent theory of Blackley (1985) that firms’ WTP for industrial parks includes the operational profit yielded from the land. In other words, an increase in firms’ productivity and a reduction in costs will increase WTP.

Factors affecting WTP

Accessibility to the market, client companies and the firms’ own factories are all important factors for firms in deciding location, as this would enable firms to reduce freight costs to the market and communication costs to the client. As in many previous studies, such as Kowalski and Paraskevopoulos (1990) and Lockwood and Rutherford (1996), these effects were summarised by accessibility to central business districts (CBDs) and various transport facilities. The offices of the companies themselves and those of client companies are usually located in the CBDs of large cities, thereby making them easily accessible to banks and other business services.

Further, since amenities for workers have a positive influence on industrial land rent (Sivitanidou and Sivitanides, 1995), even small cities assume importance owing to aspects such as commuting accessibility and the local environment of employees and their families, which includes food and consumption facilities, accessibility to medical facilities, and cultural and educational services. Although distances to both large and small cities are important, the existence of a small city influences location demand only when there are no larger cities that are located nearer than it is, as large cities can entirely substitute small ones. Therefore, we consider the effect of small cities by employing a dummy variable, while effects of large cities are captured by distance from these cities. Furthermore, among transport facilities, the presence of highways is of particular importance since more than 90 per cent of domestic freight in Japan is transported by road; hence, we employ distance to highways as a variable.

Further, the spatial agglomeration of the industrial sector is presumed to have a positive effect on WTP for two reasons. The first is a positive externality or economies of agglomeration among firms that arises owing to the advantages firms enjoy from being located in close proximity with each other; this leads to an increase in firms’ productivity because of the thickness of input and output markets (for example, Marshall, 1890). Secondly, if agglomeration reveals the potential of investment or the number of firms seeking a new location within the area, the maximum WTP will increase with the scale of agglomeration. Although barely any literature has mentioned the effect of agglomeration on industrial land price, Lin and Ben (2009) actually indicated a positive effect in this regard.

Finally, although the effect of market land price around industrial parks on WTP is theoretically ambiguous, we employ this variable in order to compare WTP with sales price, which will be significantly affected by this variable. In accordance with our basic hypothesis that firms only consider profit from production at the locations, we assume that WTP is not influenced by land price itself. However, the market price of land might affect WTP if firms consider the resale value or asset value of the land. In order to substitute land price, we employ average land price for commercial use (LPC) of each prefecture as a proxy for the land price of the nearest city’s CBD. Since the price of land gradually decreases with increasing distance from the CBD, a high LPC implies a high land price of each location when its distance from CBD is fixed.

Variables affecting the possibility of sales

Some attributes of industrial parks directly influence their sales, rather than the profit of firms, and hence their WTP. For example, a certain amount of unsold industrial land within the same region, or excess supply, will disperse the potential customers of industrial parks and then decrease the possibility of sales of individual parks. Another example is promotional activities for sales by industrial companies and local governments. In order to capture such activities, a price discount for industrial parks within the observation period will be a good proxy. We can regard it not only as a discount of the price but also as an indicator of effort for sales of a public company; hence, we examine price discount as a dummy in order to separate it from the effect of price itself.

Although we distinguish the excess supply effect and the price discount effect from the determinant factors of WTP because they do not affect the profitability of each park, they can also be assumed to influence WTP since they have some impacts on the demand price for each park. However, the determinant factors and non-determinant factors of WTP are statistically inseparable in our probit model because these variables can be actually treated and estimated in the same manner as variables included in $X_{j}$ in equations (6) and (7).⁴

3.2 Hypotheses on Sales Price

Although the sales price is determined by public companies, we propose that sales price represents a tendency that is similar to firms’ WTP because it somewhat considers the condition of demand for industrial land. Saz-Salazar and García-Menéndez (2005) and Lin and Ben (2009) indicated that accessibility and agglomeration have positive effects on the sales price of industrial parks, which is similar to our assumption regarding WTP.

However, our specific hypothesis is that the sales price is not simply determined by demand because it is also affected by industrial policies and the management system of public companies. In Japan, the land development business of public companies is required to be self-supporting, while also beneficial for regional development; hence, the sales price tends to include land acquisition costs and, in our model, LPC is the proxy of such costs. Further, these companies have a problematic accounting system that prevents flexible repricing considering current demand. Because of that, the coefficient value of some factors might be different between WTP and sales price, although the signs are the same. However, we have no ex ante anticipation for the supply–demand gap; hence, we merely attempt to estimate the model.

3.3 Data

The present study uses Kojo Tekichi Soran (a comprehensive list of industrial land on sale) from 2002 and 2007, which is the disaggregated data of Japanese individual industrial parks assembled and published by the Japanese Ministry of Economy, Trade, and Industry (METI).⁵ With this database, detailed information is available for all Kojo tekichi (suitable land for industrial use) for sale. The present study uses samples included in both the 2002 and 2007 data in order to discriminate whether each industrial park is sold or not during the five years by differentiating the area of unsold land between the two periods. Since the databases are separated by period, we identify two samples in each database from various information such as their name, location and total area, as well as information from the websites of each local government, when we connect samples from the two periods. However, even when two samples are identified, we exclude the samples whose total areas differ by more than 10 000 square metres between the two periods and those with an unsold area that is smaller than 10 000 square metres.

The present study decides that, if the difference in the area between the two periods is more than 5000 square metres, at least one block of the industrial park has been sold.⁶ The estimation of equations (6) and (7) does not require information about how much land is sold, but merely whether one block has sold or not. In addition, for the focus of the present study, we limit the scope of the investigation to the completed industrial parks owned by the public sector.⁷ As a result, 373 samples are available for the present study.

We employ the following variables for sales price and WTP, as listed in Table 1, for explaining our hypotheses: accessibility to large and small cities (D-20 and 5-DM respectively) and highways (D-HI), industrial agglomeration effects substituted by the gross regional product of the manufacturing sector at the prefectural level (GPPM), market land price substituted by the average land price of commercial land use in each prefecture (LPC), discount of sales price (PD-DM) and total area of unsold industrial park within a prefecture (IPA). We use the data from 2002 for the explanation variables, but the data from 2007 for variables included only in the 2007 dataset.⁸ A list of the variables and their definitions are shown in Table 1. Table 2 shows a basic statistical description of the samples.

Table 1.

Description of the variables

Variables	Definition
Price	Average price of each industrial park per unit area (yen/square metre)
D-HI ^†	Distance to the nearest highway interchange (km)
D-20 ^†	Distance to the nearest city with over 0.2 million population (km)
5-DM ^†	Dummy—the value is 1 if both of the following conditions are satisfied: there is a mid-size city with a population between 0.05 million and 0.2 million within 10 km of the park; and, that is the nearest city for the park
LPC	Average commercial land price of the prefecture where each industrial park is located (yen/square metre)
GPPM	Aggregated product of the industrial sector of the prefecture where each industrial park is located (10 billion yen)
PD-DM	Dummy—sales price of 2007 is lower than that of 2002 by equal to or more than 1000 yen
IPA	Aggregated unsold area of the industrial parks of the prefecture where each industrial park is located

†

Data from 2007 are used for these variables; otherwise, 2002 data are used.

Table 2.

Descriptive statistics of the selected samples

Variables	Average	S.D.	Maximum	Minimum
Price (yen/square metre)	29 222	26 738	214 000	1 009
Total area, including sold and unsold area (1000s square metres)	582	1 428	15 329	10
Unsold area (1000s square metres)	165	239	2 234	10
D-HII (km)	11.9	15.1	108	0
D-20 (km)	42.0	36.4	294	3
Number of samples (Number sold)	373 (260)

4. Empirical Results

Table 3 shows the results of three estimations: the hedonic estimation of sales price by OLS, the estimation for WTP and a gap between the WTP and the hedonic estimations. Note that the results of the probit models represented in Table 3 are the estimated values from equations (6) and (7), or $α'$ , $β'$ , $α_{d}'$ and $β_{d}'$ , which are not the parameters of WTP by themselves; rather, the parameters of the WTP and the gaps are derived by dividing them by $φ$ .

Table 3.

Results of the estimation

	Hedonic		WTP		Gap between WTP and hedonic
	Coefficient	T-value	Coefficient	T-value	Coefficient	T-value
D-HI	−123.74	−1.72*	−35.28	−0.70	−28.98	−0.58
GPPM	4.31	0.71	14.99	1.90*	14.67	1.87*
LPC	0.20	10.60***	0.004	0.23	−0.006	−0.39
D-20	−458.00	−6.77***	−208.03	−3.84***	−185.09	−3.65***
D-20²	1.90	5.35***	0.75	2.51**	0.66	2.29**
5-DUM	3 493.49	1.57	6 386.92	2.64***	6 201.91	2.70***
PD-DUM	–	–	4 683.49	3.71***	4 792.59	3.63***
IPA	–	–	−414.76	−1.59	−415.06	−1.59
constant	14 796.51	4.08***	9 209.03	2.64***	8 444.98	2.70***
$φ$	–	–	0.05	1.36	0.05	1.36
Adjusted R²	0.48		–		–
ρ²	–		0.24		0.24

Notes: * identifies significance at 10 per cent; ** identifies significance at 5 per cent; *** identifies significance at 1 per cent.

Regarding the results of the hedonic estimation, D-20 and its square, and average land price for commercial use (LPC) are the significant variables at the 1 per cent level; that is, the sales prices of industrial parks also have tendencies similar to general land prices nearby. Further, distance from highway interchanges (D-HI) is significant at the 10 per cent level. Although we could not observe significant effects from other variables on land value, the sign of their coefficient is consistent with its hypothesis. The adjusted R² of the model is 0.48, which is a slightly smaller value than those in the previous studies. The reason is that our dataset includes all the Japanese industrial parks, whereas the previous investigations have been limited to regions within countries.⁹

The estimation of equation (6) for WTP represents similar results to those of the hedonic model except for a few variables. While D-20 and its square represent, as expected, a significant influence on WTP at the 1 per cent level, LPC regarded as the land price of the central district has no significant effect. This result implies that firms’ WTP is not affected by the standard price of nearby land, but by the profit from the resulting operations at that industrial park. In addition, access to small cities (5-DUM) and the production of the industrial sector (GPPM) have significant effects respectively at the 1 per cent and 10 per cent levels. The result of 5-DUM implies that accessibility to the cities is important even if they are small because it improves the living environment and commuting time of the employees. The positive and significant sign of GPPM also implies that the industrial firms consider the agglomeration effect in locating plants.

Furthermore, the estimation of the supply–demand gap also represents the significant effects of D-20, GPPM and 5-DUM. These results provide an important implication that firms’ WTP is more sensitive than the public companies’ price setting with regard to accessibility to the city and the agglomeration economies of industrial production. Since the Japanese domestic industry has undergone reform in shifting to a labour-intensive and highly productive production form while horizontal specialisation in East Asian countries is progressing, firms choose their location with regard to the living environment of their employees and their accessibility, along with that of their business partners, to plants and offices.

Both of these probit models also yield the expected sign for the coefficients of the price discount (PD-DUM) and the aggregated unsold area of the industrial parks (IPA). Since the coefficient of PD-DUM is statistically significant, it is believed that the discount of industrial parks attracts plants. Although the present study gives it as an exogenous variable, we must do further investigation into why the holders of the parks discount the price.

The log-likelihood ratio (Macfadden’s $ρ^{2}$ ) of both models is higher than 0.2, and both of them are significant at the 1 per cent level of significance in the results of the chi-squared test for their log-likelihood.¹⁰ Therefore, we believe that these models are reliable enough.

The results obtained with regard to sales price and WTP indicate consistency with the proposed hypotheses. Attributes regarding location are primary sources for both sales price and WTP, as confirmed in various types of studies on the value of industrial land (for example, Ambrose, 1990; Lockwood and Rutherford, 1996; Saz-Salazar and García-Menéndez, 2005); moreover, agglomeration entails some advantages, as shown in Lin and Ben (2009), despite the fact that its t-value in sales price is insignificant.

However, by comparing the sales price and WTP, the major finding of this study is supply–demand dispersion. We have indicated that sales price and WTP have different determinant factors. First, the level of LPC only influences sales price while it does not influence WTP, which implies cost-covering price-setting by industrial parks. Secondly, only WTP is influenced by industrial agglomeration and accessibility to small cities, and WTP is more sensitive to accessibility to large cities than sales price; these facts imply that demand is imperfectly represented in sales price.

The difference in key determinant factors between WTP and sales price is actually shown by the following numerical example. Here, we assume an industrial park with average explanatory variables and calculate the contribution ratio for each of the seven variables (D-HI, GPPM, LPC, D-20, D-20², 5-DUM, const) by using the estimation results presented in Table 3; that is, we examine the composition of WTP and sales price of an average industrial park. The results presented in Table 4 support the insight from Table 3. First, LPC is the primary determinant factor for the sales price (38.2 per cent), while for WTP, it is the second lowest one (2.0 per cent); hence, only sales price is dependent on land development cost. Further, the contribution of D-20 is significant for both sales price (27.4 per cent) and WTP (32.9 per cent), but it contributes more towards WTP; this is the case for 5-DM.

Table 4.

Contribution ratio of each variable on sales price and WTP (percentages)

	D-HI	GPPM	LPC	D-20	D-20²	5-DUM	Constant
Sales price	2.1	1.4	38.2	27.4	8.3	1.6	21.1
WTP	1.6	12.5	2.0	32.9	8.7	7.6	34.7

Note: Contribution ratio of variable k for WTP is defined as $| {β'}_{k} X_{kj} | / \sum_{m = 1}^{M} | {β'}_{m} X_{mj} |$ and that for sales price is $| γ_{k} X_{kj} | / \sum_{m = 1}^{M} | γ_{m} X_{mj} |$ .

Our results imply that the price setting of Japanese industrial parks has some problems: it is dependent on development cost and inflexibly represents current demand. One of the biggest sources of such inefficiency of public companies is their accounting system that causes rigidity and cost-covering bias in price setting. These systems should be promptly reformed in order to achieve an appropriate discount for overvalued industrial parks, such as those inaccessible to cities and lacking agglomeration. As the coefficient of PD-DM implies, promotion of sales with some discount will have a positive result on sales. Furthermore, the results of this study could also provide some hints on the provision strategy for new industrial parks. For example, accessibility to small cities has little effect on sales price or costs, but has a significant effect on demand. Therefore, public companies will fulfill the needs of the industrial sector without increase in cost when they focus on the attribute of location.

Finally, a further numerical application of the results of our model on WTP could be a sensitivity analysis. Our estimations on WTP and supply–demand gaps are expected to provide useful information with regard to how much each industrial park should be discounted. In order to obtain the WTP and marginal gap among variables, the coefficients of WTP and WTP–sales gaps represented in Table 3, $β_{k}'$ and $β_{dk}'$ , should be divided by $φ$ , the coefficient of price and OLS residual. However, in our result, $φ$ is statistically insignificant despite the fact that its sign is consistent with that of the hypothesis. The estimated value of $φ = 0.05$ is actually small, and a sensitivity analysis with such a value of $φ$ represents that with and without 5-DM, there is a difference of 127 738 yen/square metre in terms of WTP, which seems to be too high in comparison with the average price of industrial parks represented in Table 2. Further, the value of estimated $φ$ tend to be unstable among a different set of variables. Therefore, considering the insufficient reliability of $φ$ , caution must be exercised in using the WTP function for numerical analysis to suggest policy implications.

One substantial reason for that problem is a serious multicollinearity between the actual sales price and explanatory variables of the probit model; this is not the usual case of applying a CVM questionnaire with virtual price. However, obtaining a reliable and stable result of $φ$ for more practical use of the model is a topic for future research. One possible idea for this is applying the two-step estimation given by of McMillen et al. (1992) and McMillen (1997) which adjusts the OLS estimates of market land price by a probit model having regard to which sector uses each land type; however, we must customise the model to apply it to our non-market price of industrial parks.

5. Conclusions

This study has investigated the determinant factors of the sales price of land in publicly supplied industrial parks in Japan, firms’ WTP for them and the estimated gap between the WTP and sales price.

The most significant contribution of this study is indicating the existence of a supply–demand gap by comparing sales price and WTP, rather than estimating each of them separately. The results reveal that the sales price is covering the cost, but incompletely representing current demand. Our result implies inefficient and inflexible price setting by Japanese public companies; we suggest that their problematic accounting system must be reformed in order to facilitate an appropriate repricing of industrial parks. Further, the remaining problems of numerical application of the model for practical use with regard to the reliability of the coefficients of price must be addressed in a future study.

It must be noted that, although we examined the basic attributes of industrial parks, some important aspects of location choice and industrial policy are omitted from the analysis because of data restrictions. First, information about which company purchases each unit of land will show us what attributes of firms influence their WTP, as examined in Blackley (1985). Further, information on what kinds of firms are initially operating in each industrial park will also give us an insight into how interfirm relationship influences their WTP. Secondly, we should consider policy variables such as tax incentives and subsidies by local governments. Since the development of industrial parks is a part of overall industrial policy, discussing the management of industrial parks in the context of these policies is an important issue. However, it is quite difficult to collect and organise such kinds of detailed information for all industrial parks in Japan. Gathering comprehensive and correct data about policies is particularly hard because there are numerous types of policies in place to make locations more attractive; moreover, more than one policy may be in effect in the same industrial park. One feasible idea for investigating these detailed policies is to narrow the research focus to one specific region; this could also serve as a subject for further study.

Footnotes

Acknowledgements

This paper represents part of the research carried out by the author as a Research Fellow at the Institute of Transport Policy Studies (ITPS), Japan. The author appreciates the continual advice of Shigeru Morichi, Makoto Ito, Tadahiro Okuyama and other ITPS staff. The author would also like to thank Yoshihisa Asada, Seil Mun, Tomoya Mori and two anonymous referees of this journal for their helpful comments. Needless to say, the author is responsible for any remaining errors and policy recommendation in the present paper. The electronic data for Kojo Tekichi Soran in 2002 were kindly donated by METI.

Funding

The author wishes to acknowledge financial support from the Grant-in-Aid for Scientific Research (B23330095).

Notes

References

Akai

(2005) Issues and reality of local public companies on housing, road and land, Financial Review, 76, pp. 76–123 [in Japanese].

Ambrose

(1990) An analysis of the factors affecting light industrial property valuation, Journal of Real Estate Research, 5, pp. 355–370.

Blackley

P. R.

(1985) The demand for industrial sites in a metropolitan area: theory, empirical evidence, and policy implications, Journal of Urban Economics, 17, pp. 247–261.

Hannemann

W. M.

(1984) Welfare evaluations in CV experiments with discrete responses, American Journal of Agricultural Economics, 66, pp. 332–341.

Jackson

T. O.

(2002) Environmental contamination and industrial real estate prices, Journal of Real Estate Research, 23, pp. 179–199.

Kowalski

J. G.

Paraskevopoulos

C. C.

(1990) The impact of location on urban industrial land prices, Journal of Urban Economics, 27, pp. 16–24.

Lin

S.-W.

Ben

T.-M.

(2009) Impact of government and industrial agglomeration on industrial land prices: a Taiwanese case study, Habitat International, 33, pp. 412–418.

Lockwood

L. J.

Rutherford

(1996) Determinants of industrial property value, Real Estate Economics, 24, pp. 257–272.

Marshall

(1890) Principles of Economics. London: Macmillan.

10.

McMillen

D. P.

(1997) Multiple regime bid-rent function estimation, Journal of Urban Economics, 41, pp. 301–319.

11.

McMillen

D. P.

Jarmin

Thorsnes

(1992) Selection bias and land development in the monocentric city model, Journal of Urban Economics, 31, pp. 273–284.

12.

Saz-Salazar

S. del

García-Menéndez

(2005) Public provision versus private provision of industrial land: a hedonic approach, Land Use Policy, 22, pp. 215–213.

13.

Sivitanidou

Sivitanides

(1995) Industrial rent differentials: the case of greater Los Angeles, Environment and Planning A, 27, pp. 1133–1146.