The Effect of Regional Inequalities on Migration: A Comparative Analysis of and

Boris A. Portnov*

ABSTRACT

This article posits three main questions:

- Is there a general “mechanism” through which disparities in regional development affect patterns of cross- migration?

- Which aspects of regional inequalities (climate, employment, housing availability, etc.) have the most profound effect on rates and direction of inter-area migration?

- Which planning policies and strategies are conducive to increasing the migration attractiveness of peripheral development regions?

In an attempt to answer these questions, the 1985-1995 statistical data for two relatively small and densely populated countries – Israel and Japan – are used.

A general model of the factors affecting cross-district migration is suggested, and analysis-of-variance is used to explain the factors influencing rate of cross-area migration in the two countries.

Although these countries differ substantially in respect to population size and local development, they appear to exhibit considerable similarities in general patterns of cross-district migration.

Since the late 1980s, the attractiveness of core regions in both countries has tended to decline, while that of peripheral areas appeared to grow. It is

* Center for Desert Architecture and Urban Planning, J. Blaustein Institute for Desert Research, Ben-Gurion University of the Negev, Sede-Boker Campus, Israel.

Published by Blackwell Publishers Ltd., © 1999 IOM 108 Cowley Road, Oxford OX4 1JF, UK, and International Migration Vol. 37 (3) 1999 350 Main Street, Malden, MA 02148, USA. ISSN 0020-7985 588 Portnov

argued that this resemblance of migration patterns is related to a similar balance of employment and housing availability in different geographic areas.

INTRODUCTION

Uneven distribution of population and economic activity is an intrinsic aspect of regional development world-wide. Examples of countries that have ex- tremely uneven patterns of regional development include: Norway, whose population and industries are heavily “skewed” towards the Oslo metropolitan region; China, whose population and economic development are predomi- nantly concentrated along the country’s coastal strip; Israel, in which the concentration of economic development and population in the coastal area is also clearly evident; and Japan, whose geographically small Kanto and Kinki contain nearly half the country’s population and a large share of all available jobs.

Regional disparities in the availability of housing and employment, and provision of services, profoundly affect cross-district migration. Economically advanced regions tend, in general, to attract migrants, while outward migration is an essential attribute of economically lagging, specifically peripheral, areas (Bourne, 1975; Lipshitz, 1997; Portnov and Pearlmutter, 1997; Zheng, 1997). The effect of regional inequalities on migration is not, however, straightforward. Whereas inequalities in regional development are cumulative and persistent (Richardson, 1977; Friedmann and Weaver, 1979; Armstrong and Taylor, 1993), the patterns of interregional migrations are inherently dynamic (Vining, 1982; Balchin, 1990).

While many studies have focused on social and economic disparities of regional development (Abe, 1996; Armstrong and Taylor, 1993; Friedmann and Weaver, 1979; Diamond and Spence, 1983; Balchin, 1990), the effect of these inequalities on interregional migration has not been sufficiently studied and understood. Examination of these relationships may, however, provide answers to the following questions:

1. Is there a general “mechanism” through which disparities in regional development affect patterns of cross-district migration?

2. Which aspects of regional inequalities (climate, employment, housing availability, services, etc.) have the most profound effect on the rates and directions of interregional migration?

3. Which planning policies and strategies are conducive to increasing the migration attractiveness of peripheral development areas? The effect of regional inequalities on migration: a comparative analysis 589

This article attempts to answer these questions by examining recent patterns of cross-district migration in two relatively small and densely populated countries – Israel and Japan. It begins with a brief review of factors and forces affecting long-distance migration and regional inequalities, followed by an overview of the general patterns of regional development and migration in Israel and Japan. Empirical models of “neutral migration” are suggested to determine a specific balance of employment and housing which is conducive to migration attractiveness of an area.

FACTORS INFLUENCING MIGRATION

In the general literature on migration decision-making, motivations for migration are commonly defined in terms of a “push-pull” continuum based on George’s (1970) classification of long-distance migration. This classification identifies two types of migrations: (1) those caused by necessity or obligations, and (2) moves stemming from socio-economic and cultural needs. While the first type of migration motives relates mainly to area-of-origin political and/or religion (push) factors, the second type is assumed to be determined by both economic hardships in the area of origin (push factors) and economic oppor- tunities of the area of destination (pull factors).

In this framework, an economic motive is generally considered the strongest for migration. As Fischer (1976) argues, a predominant motive for long- distance migrations appears to be economic: migrants generally seek better jobs, higher wages, better schooling, and a materially better life. Vining (1982), and LaLonde and Topel (1997) also stress the key role of this motive in determining patterns of cross-district and international migration. They argue that differences in level of technology and socio-economic development are the major cause of migration between areas.

Residential satisfaction motive is another important determinant of migration behaviour. Availability of housing in the area is traditionally considered a key factor affecting rates of inward and outward migrations (Lipshitz, 1997; Portnov and Pearlmutter, 1997). The prime cause of “housing” migrations is attributable to changes in family composition which occur as a family goes through its life-circle. “Housing” migration can also stem from dissatisfaction with residential and environmental conditions at the place of origin (De Jong and Fawcett, 1981; Portnov and Pearlmutter, 1997).

Other motives commonly cited as influencing cross-area migration include: aspirations for higher social status (Findley, 1977; Ehrlich, Ehrlich and Holdren, 1972; De Jong and Fawcett, 1981); maintain community-based social and economic ties (Uhlenberg, 1973); family and friends affiliation (Fischer, 1976; Moore and Rosenberg, 1995; Lipshitz, 1997; Portnov, 1998); 590 Portnov and attaining life-style preferences (Rapoport, 1977; Portnov and Pearlmutter, 1997; Lipshitz, 1997).

ISRAEL AND JAPAN: GENERAL PATTERNS OF URBAN DEVELOPMENT

Patterns of urbanization in Israel and recent development policies

The majority of urban settlements in Israel are concentrated along the Mediterranean coast or in close proximity to its major metropolitan centres – , and (Figure 1, page 605). The population of these urban centres and their immediate hinterlands (the Tel Aviv, Central, Jerusalem and Haifa districts) exceeds 3.5 million, or nearly 70 per cent of the country’s population (ICBS, 1997).

Concentration of Israel’s population and lack of development in outlying regions create significant problems for national security and for maintaining the future capacity for immigrant absorption. Realizing the extent of these problems, government and planning officials in Israel have favoured at various times the implementation of a policy promoting population dispersal to periph- eral regions of the country (Newman, Gradus and Levinson, 1995).

From the early 1950s to late 1960s, this policy was sustained primarily by directing new immigrants to sparsely populated areas (Fialkoff, 1992); since the early 1970s the policy has been gradually replaced by geographically selective “aid programmes”. These provide public housing to certain categories of newcomers as well as financial incentives to builders and buyers in specified development areas (Lerman and Lerman, 1992). In the mid-1980s, the policy of “direct absorption” was introduced, under which new immigrants received an amount of money that lasts for a limited period (“the basket of absorption”) and can decide for themselves where they want to settle (Lipshitz, 1997).

Current issues of regional and urban development planning in Japan

The core area of Japan’s industrial development is concentrated in three highly urbanized clusters, viz., Tokyo and Yokohama (the core cities of the Kanto region); Kobe, Osaka, and Kyoto (the core cities of the Kinki region); Nagoya (the core city of the Tokai region); Hiroshima (the core city of the Chugoku region), and Kitakyushu and Fukuoka (the core cities of the Kyushu region) (Figure 2, page 606). Population density within these urban areas exceeds 7,000-12,000 persons per square kilometre (JSB, 1997).

One reason why population density is so high is that almost all the flat land for building in Japan is on the eastern coast. The only substantial areas of available The effect of regional inequalities on migration: a comparative analysis 591 flat land are on the northern island of , which has a severe climate with harsh winters (Vining, 1982).

As Abe (1996) notes, until the early 1970s, urban development in Japan was characterized by large-scale “job-driven” migration from local regions to three metropolitan areas – the Tokyo, Osaka and Nagoya regions. From the first oil crisis in 1973 until the late 1980s, return migration from metropolitan areas to local regions declined significantly. From the end of the 1980s to the 1990s, net migration into the Tokyo metropolitan region again increased; overcrowding of population and economic activity in Tokyo remains one of the most serious regional problems in Japan.

In light of ongoing concentration of population and economic activity in a few metropolitan centres, three major regional development policies were adopted by the government: (1) construction of a national network of transportation and communication to form “national development axes” along the eastern and western coasts; (2) creation of growth poles in local regions; and (3) relocation of industries from the metropolitan regions to the peripheral areas (Abe, 1996; Markusen, 1996).

INTERREGIONAL INEQUALITIES AND MIGRATION TRENDS

As Table 1 (page 610) shows, Israel and Japan differ substantially in respect to many development indicators, including population size, land area, and overall rate of population growth. For example, Israel’s population (5,500,000 residents) is less than five per cent of Japan’s (125,500,000 residents), and the annual rate of population growth in Israel (2.1 per cent) greatly exceeds the rate in Japan (0.2 per cent). For the present analysis, it is important that regional distribution of population and economic activity is considerably uneven in both countries. The extent of these inequalities can be illustrated by using simplified Williamson’s index (Table 1).1 Although the values of this index for the two countries are not identical, the values are considerable. The extent of these inequalities and their effect on patterns of cross-district migration are discussed below.

ISRAEL

Four quantitative indicators have been used to illustrate the degree of social and economic disparities among the administrative regions of Israel: population share, employment, housing rent, and the structure of population growth (Table 2, page 611). The latter indicator – the structure of population growth – is defined as the ratio between the migration component of growth (MB = migration balance) and net natural growth (NG = birth minus death rates). This ratio may be considered an important indicator of a district’s 592 Portnov development, since an excess of migration over natural growth indicates the area’s ability to attract newcomers and retain existing residents (Portnov and Erell, 1998).

Concentration of population and economic development in core areas of Israel (the Tel Aviv, Jerusalem, Haifa and Central districts) is clearly evident in Table 2. Although the land area of these districts amounts to only 14 per cent of the country’s territory, they contain up to two-thirds of the country’ population and nearly three-quarters of all available jobs. Unsurprisingly, housing prices in these districts are considerably higher than in those in the country’s periphery. In the , for instance, the price of housing in 1995 was almost 40 per cent above the national level, while in the country’s peripheral areas (the Northern and Southern districts), it was well below the national average.

The “skewed” concentration of population and economic activity seems not to be reflected in patterns of the districts’ migration attractiveness. Indeed, as Table 2 shows, the districts which in 1985 and 1995 had the highest MB/NG ratios include both core and peripheral areas: the Central, Haifa, Northern and Southern districts. To explain this phenomenon, a number of factors must be taken into consideration.

First, a high proportion of overall population growth in Israel is due to foreign immigration. During the recent wave of mass immigration from the former Soviet Union (1990-1991), immigration accounted for as much as 80 per cent of population increase (ICBS, 1997). As Lipshitz (1997) points out, new immigrants in Israel unequivocally prefer core districts as an initial place of residence. This, in turn, may help explain the above high values of the MB/NG index found in some core regions. Trends of in-country migration and immigrant absorption graphically represented in Figure 1 are generally in line with this assumption. During the recent wave of mass immigration (1990-1991), the overwhelming majority of the NI (new immigrants) settled in the country’s central areas – the Haifa, Tel-Aviv, and Central districts. Concurrently, two out of three of the above central districts (the Tel-Aviv and Haifa) experienced in these years a substantial loss of their established population due to a negative annual migration balance.

The difference in migration behaviour of NI and in-country migrants (IM) becomes even clearer if only two aggregated geographic areas of the country – core and periphery – are considered. This aggregation makes it possible to differentiate suburban sprawl to geographic areas adjacent to the major population centres, and outward migration to the periphery. Figure 2 shows that the core districts appeared to maintain a quantitative prevalence over the peripheral areas with respect to annual number of NI. This prevalence remained virtually unchanged during the years covered by the present case study. Meantime, with respect to IM, the balance between “core” and The effect of regional inequalities on migration: a comparative analysis 593

“periphery” changed dramatically within the time-span in question: while the migration attractiveness of Israel’s core districts tended to decline, that of peripheral areas appeared to grow.

The influx of NI to a particular district of Israel and the outflow of IM from the district seem to be interrelated. For instance, the beginning of mass migration from the Tel Aviv district after 1990-1991 clearly coincides with the beginning of the above immigration wave (see Figure 2). Even in the Central district, which can be considered the suburban “buffer” of Greater Tel Aviv, the balance of internal migration fell below zero level in 1991 when the district accommo- dated some 37,000 NI (Figure 1). Though indirectly, the parallel processes of internal migration and NI’s initial distribution apparently correlate via housing prices: influx of new immigrants to a particular locality boosts housing demand and increases housing prices. Through a spiral process this may eventually cause the outflow of current residents (who cannot afford decent housing in the metropolitan centres) to areas of the country where housing is more widely available and more affordable (Portnov and Pearlmutter, 1997). Unsurprisingly, this migration stream is often associated with the internal migration of recent immigrants and involves a change from rental housing to home ownership (Lipshitz, 1997).

In explaining inter-regional patterns of population growth in Israel, the country’s relatively small land area (some 21,500 square kilometres) should also be taken into consideration. Relatively short distances between various parts of the country make it possible to commute on a daily basis across regional borders. Migrants can move into relatively cheap housing in the periphery and still maintain their jobs in the core, as is the case with many immigrants from the former Soviet Union (Lipshitz, 1997). Such commuting occurs from the Central administrative district and the Ashqelon subdistrict (which is a part of the Southern district) to the core areas of the country – the Tel Aviv and Jerusalem districts – where most jobs are available. The relatively high level of migration attractiveness in recent years of the Central and Southern districts (Table 2; Figure 1) appears to be in line with this assumption.

JAPAN

Regional development in Japan is also considerably “asymmetric” (Table 3, page 612). Three core regions of the country – the Kinki, Kanto and Tokai – cover only 22 per cent of the country’s territory (Figure 3, page 607) but contain up to 60 per cent of the country’s population and a similar percentage of available jobs. As in Israel, housing rent in these overpopulated regions substantially exceeds the national average. In the Kanto district, for instance, average housing rent in 1995 was nearly 40 per cent above the national average (Table 3). 594 Portnov

Time-related changes in the selected development indicators deserve comment. As Table 3 shows, while no substantial changes occurred between 1985 and 1995 in the spatial distribution of the country’s population and location of employment, the structure of the districts’ population growth, measured by the MB/NG index, appeared to change drastically. For example, the values of this index in the Kanto and Tokai districts declined from high positive values in 1985 (0.64 and 0.15, respectively) to an almost zero level in 1995, implying that the share of migration in these districts’ population growth declined sharply. Concurrently, peripheral regions of the country appear to have become some- what more attractive to migrants. As Table 3 shows, the values of the MB/NG index in these districts either rose from negative to positive figures (the Hokkaido, Tohoku, Chugoku, Shikoku, Kyushu, and Hokukiku districts), or became more positive (the Tosan district).

This observation is in agreement with migration trends represented in Figure 3. During the past decade, relative migration attractiveness of the core (the Kanto, Tokai and Kinki) appeared to decline, while that of peripheral areas tended to increase. The absolute rate of net migration in both core and peripheral districts of the country also tended to decrease.

A number of possible explanations of these trends can be offered. First, as Table 3 shows, the relative level of housing prices in the core districts of Japan increased rapidly between 1985 and 1995. In the Kanto district, for instance, housing rent increased from 30 per cent above the national average in 1985 to almost 40 per cent ten years thereafter (Table 3). At the same time, the relative level of housing rent in almost all the peripheral areas of the country appeared to decline. In the Hokkaido district, for instance, this indicator declined from 20 per cent above the national average in 1985 to 2 per cent below the average in 1995.

Another influencing factor is the overall slowdown of the Japanese economy after 1991 (JSB, 1997). Portnov and Erell (1998) argue that the overall rate of in-country migration tends to decrease in “bad” economic years when people prefer to refrain from “costly” long-distance moves. This explains why the overall rates of net migration by districts of Japan appeared to decline. The above decline in migration attractiveness of Japan’s metropolitan regions can also be attributed to an aggressive government policy aimed at preventing excessive accumulation of population and economic activities in the country’s core areas (Abe, 1996).

RESEARCH PHASES

To identify and measure the significance of the factors which affect patterns of cross-district migration in the two countries, an analysis was carried out in two main phases: (a) data collection, and (b) modelling. The effect of regional inequalities on migration: a comparative analysis 595

Data collection

The first phase of the analysis included collection and descriptive analysis of recent statistical data for ten administrative districts of Japan (Figure 2), and six administrative (Figure 1).2 These covered the physical parameters of the districts’ development (employment, housing construction, housing prices, population density, climate), and the districts’ attractiveness to migrants. Annual statistical data for 1985 to 1995 were used to calculate the values of the research variables using comparable and complete sets of data.3 In the case of Israel, 66 observations (eleven points in time and six administrative districts) were obtained for each of the research variables; for Japan, some 100 observations (ten points in time and ten administrative districts) were obtained.4 Data for the analysis came from the Israeli Central Bureau of Statistics: Statistical Abstract of Israel (1985-1996) and Construction in Israel (1985-1996), and Japanese Statistics Bureau: Japan. Statistical Yearbook (1985-1997). Climatic indices for the Israeli case study were drawn from the Climatic Atlas of Israel (Bitan and Rubin, 1991).

The “district” approach to studying patterns of in-country migration has certain drawbacks. For instance, knowing that a number of persons entered a particular district, or moved from one to another, does not allow one to determine actual characteristics of the settlements which were the sources and the destinations of the migrants. On the other hand, the relatively large size of the districts’ land areas made it possible to concentrate on long distance migrations rather than on short distance mobility between, for instance, urban centres and their immediate periphery.

Modelling

In the second phase of the research, analysis of variance was used to determine factors affecting the pattern of cross-district migration in both countries. The ratio between migration balance and natural growth (the MB/NG index), discussed above, was used as the dependent variable. To calculate the values of this index, the following components of a district’s population growth were analysed:

- The annual net balance of internal (in-country) migration (IC) (thousands of migrants). For Israel, these data do not include information on the initial settling of new immigrants but do include sequential changes in their addresses. For Japan, the balance includes both immigrants and in-country migrants, since disaggregated data on foreign immigration are not available.

- The annual rate of foreign immigration to a district (thousands of immigrants). For Israel, this reflects initial distribution of the new 596 Portnov

immigrants, i.e., a first place of their residence.5 The overall rate of foreign immigration to Israel has been subject to substantial annual changes. For instance, during the period 1985-1995 immigration varied from 9,500 in 1986 to nearly 200,000 in 1990. In order to reduce the impact of this fluctuation on the outcome of the interregional analysis, the original immigration data for each district were normalized. The average annual immigration for the entire time-span in question (69,000 new immigrants) was assumed as the conditional baseline in calculating the normalized immigration rates (IMN).

- The rate of natural growth (NG) is the differential between annual fertility and mortality rates in a district (thousands of persons).

Given the above components of population growth, the values of the MB/NG index were calculated. using the following formulas:

MB/NG = IC / NG (Japan); MB/NG = (IC + IMN) / NG (Israel).

The following explanatory variables were included in the analysis.

1. Climate: as a proxy for climatic harshness of the area, two different indicators were used. In Israel, overheated conditions are commonly perceived as the main source of thermal discomfort, so the mean annual number of days with heat stress (Bitan and Rubin, 1991) was used as an index of a district’s climatic harshness. In Japan, the mean annual number of snowy and rainy days (JSB, 1997) was used as a proxy for climatic harshness of an area.

2. Density: population density per square kilometre of a district’s land area.

3. Economy: annual change in per capita GDP was included in the analysis as an explanatory variable. As suggested in Subsection 4.1, the rate of cross-district migration can be affected by the overall economic performance of a country as a whole.

4. Employment: overall number of employees in all the establishments and institutions located in a district (thousands).

5. Employment Change: annual change in the number of employees in a district (per cent).

6. Housing Construction: the area of gross building of housing in a given year (thousands of m2).. The effect of regional inequalities on migration: a comparative analysis 597

7. Housing Prices: since fully comparable data for Israel and Japan were unavailable, two proxies for housing prices were used. In Israel, housing prices in a district were approximated by the average market price of a standard housing unit, while for Japan average housing rent in the private sector was used. To narrow the differences caused by these two different approaches, the absolute values for a district were normalized using the national average in a respective year as the conditional baseline.

The list of variables included in the analysis covers all the main factors which a hypothetical immigrant or an in-country migrant could expectedly employ in the process of migration decision-making: housing construction, housing prices, employment, level of urbanization (population density), overall economic situation, and climatic harshness of the area. A “standard” list of possible motivations (see, for example, Ehrlich, Ehrlich and Holdren, 1972; De Jong and Fawcett, 1981; Moore and Rosenberg, 1995; Portnov, 1998) was supplemented with indicators of climatic harshness, and with data on housing construction in a district. At the same time, some “traditional” components of the motivation list (level of social and health services, earning differentials) were excluded due to restrictions on data availability and comparability.

The mean values and standard deviations of all variables included in the analysis are given in Appendix 1 (page 613).

Although regression analysis is a commonly used technique in a number of empirical migration studies (Moore and Rosenberg, 1995; Portnov and Pearlmutter, 1997; Portnov, 1998), the analysis of variance (ANOVA) was used in the present study. First, the goal of the analysis was to test the significance of explanatory variables rather than build a proper prediction model. Analysis of variance was therefore sufficient. Second, this technique allowed inclusion in the analysis of interrelated factors and also opportunity to study the “net” effect of each of these factors; the differences in the other factors are then controlled for. It was kept in mind that collinearity of explana- tory variables is often a problem when regression analysis is used. On the other hand, the use of ANOVA required testing that explanatory variables, measured on a categorical scale, are homoscedastic. In order to ensure the homoscedasticity of errors, a logarithmic transformation was applied to original explanatory variables. The Hartley’s F-Max test for violations of the similarity-of-variance assumption confirmed that variances are indeed homoscedastic. Metric values of logarithmically transformed variables were then treated as categorical values.

The effect of some of the factors described above is normally time-lagged, since it is the perception of reality, rather the actual conditions, which affects the decision making processes of individuals (Portnov and Erell, 1998). This is 598 Portnov particularly so if relevant information is not easily available, for example, from first hand knowledge. Potential migrants often become aware of changes in conditions at their prospective destination long after the changes have actually occurred. The values of some of the explanatory variables (housing construction, housing prices, employment, and employment change) were therefore lagged in order to reflect this process.

INFLUENCING FACTORS

Table 4 (page 613) shows the results of the analysis of variance. First, it should be noted that in both models the combined effect of the explanatory variables appears to be highly statistically significant (F>3.3, P<0.01 for Israel; F>5.0, P <0.001 for Japan). This implies that the explanatory variables included in the analysis appear to explain the variation of the dependent variable (the MB/NG index) reasonably well. The table also shows that in both countries the dominant factors affecting the indicator migration attractiveness exhibit considerable similarities.

In the case of Japan, the list of these factors includes employment, housing prices, and housing construction. For Israel, housing construction and employ- ment change appear to be statistically significant (F>4.0; P<0.05). All of these factors thus fall within two aggregated groups: employment and housing availability. None of these factors is totally unexpected. In particular, the effect of employment and housing on migration attractiveness of an area is emphasized by numerous migration studies carried out elsewhere. The similarity of migration determinants in two countries which differ considerably in respect to their population size and other development parameters is, however, somewhat remarkable, and this makes it possible to suggest a general model that explains the effect of regional inequalities on patterns of cross-district migration using an “employment-housing” paradigm.

MIGRATION PARADIGM

A simplified illustration of the way in which interregional inequalities in the availability of employment and housing affect the patterns of interregional (in-country) migration is represented in Figure 4 (page 608). Suppose that labour is required in an area. This may be obtained either by employing current residents of the area (either currently unemployed, from other industries in a settlement, or commuters from other localities), or by attracting migrants from elsewhere (either residents of other districts or foreign immigrants). The arrival of newcomers from other areas boosts housing demand. The market response to this demand may be time-lagged or inadequate (Figure 4) if local resources are insufficient; for instance, if land for development is not readily available in The effect of regional inequalities on migration: a comparative analysis 599 overpopulated core areas. Shortage of land for development may, in turn, increase housing prices, and this can cause the outflow of residents who cannot afford decent housing to areas where housing is more available and affordable (Portnov and Pearlmutter, 1997). High housing prices may also hamper the influx of migrants despite availability of employment.

The chart incorporated in the model (Figure 4) provides a simplified illustration of this “employment-housing” inter-link and its effect on migration. Suppose that a region has an established balance of employment and housing availability (point 1). Assume that new employment opportunities became available in the area (dE) and this employment growth is accompanied by an increase in housing availability (dH). Under these conditions, an influx of migrants to the area can be expected (point 2).

Although response of the housing market may be lagging or insufficient, the decrease of housing availability may not affect the existing level of a district’s migration attractiveness. Potential migrants and current residents of a district may still tolerate high housing prices in the area as a necessary “charge” for better employment opportunities, compared with other regions (point 3).

If, however, due to, for instance, land shortage, supply lags or a high rate of immigration, the availability of housing in the area drops beyond this “tolerability” level (point 4), out-migration from the area may become increasingly likely. It is thus suggested that this hypothetical threshold of “tolerability,” on which points 1 and 3 are located, can be termed the “neutral” migration curve (see the bold line in the chart). The area, which lies above this curve, presumably corresponds to a negative migration balance of a region (out-migration), while the zone which lies beneath it is likely to match the state of predominant in-migration to the area.

MIGRATION NEUTRALITY

The above curves of “neutral migration” may be constructed using statistical data available for the two countries under consideration. A subset of original district data was formed to include only the cases whose values of the MB/NG index are close to a zero level (±0.25). All these cases have a considerably low net balance of migration (MB) which can conditionally be assumed a “neutral migration” level.6

For the selected subset of cases, the overall annual rates of housing construction were compared with the number of jobs available in the area. To simplify the cross-country comparison, the original values of the research variables (housing construction and employment) were normalized, using the overall population size of the country as the conditional baseline. The results of the analysis in 600 Portnov

Figure 5 (page 609) show only the cases that match the above condition of migration neutrality.

The indicators selected for diagramming migration neutrality of a region – housing construction and availability of employment, – represent the respect- ive groups of factors which were found statistically significant in both countries. In Japan, the effect of another housing-related variable – housing prices – was also found significant. There was, however, insufficient evidence to conclude that this factor was significant in Israel. The latter was an important consideration in selecting the pair of determinant factors for a cross-country comparison.

The models appear to be a good fit and are relatively easy to interpret. In the case of Japan, for instance, the following “employment-housing” balance – 12,500,000 jobs (0.1*125,000,000 residents) vs. 31,250,000m2 of new housing completions per year (0.25*125,000,000 residents) – appears to correspond to the “neutral” migration balance in the area (Figure 5). Concurrently, lower rates of housing construction in the region are thus likely to cause out- migration. At the same time, for Israel, preconditions for the above migration “neutrality” are as follows: If a district has, for instance, 220,000 jobs (0.04*5,500,000 residents), 1,650,000m2 of new housing (0.3*5,500,000) should annually be completed in the area in order to maintain its migration “neutrality”. The higher rates of construction may lead to an increase of migration attractiveness of the area.

CONCLUSION AND POLICY IMPLICATIONS

According to general migration theory, the predominant motive of interregional migration is job-related: long distance migrants move to areas where more jobs are available and wages are higher (Fischer, 1976; LaLonde and Topel, 1997). The present comparative analysis of interregional migration in Israel and Japan makes it possible to partially reconsider this notion. The analysis shows that in both Israel and Japan internal migration does not appear to reflect distinctively regional disparities in employment availability. While the core regions of these countries continue to concentrate the bulk of available jobs, their attractiveness to in-country migrants appeared in recent years to decline substantially.

Analysis of the 1985-1995 statistical data suggests that this phenomenon can be explained in terms of the following “employment-housing” paradigm: availability of employment in the core areas attracts in-migrants and this boosts housing demand. The response of the local market to this demand is, however, often inadequate due primarily to a lack of land for development. This deficiency of the housing market increases housing prices and causes the The effect of regional inequalities on migration: a comparative analysis 601 outflow of current residents to areas where housing is more available and affordable. Expensive housing also hampers a further influx of migrants to the region despite the availability of employment. Due to these interrelated trends, net migration balance of the core areas of the two countries tended to decline, while that of peripheral regions appeared to grow.

On the basis of the above employment-housing paradigm, the notion of “neutral migration” is suggested. This state of migration corresponds to such a balance of housing and employment availability which causes neither a substantial influx of newcomers nor an outflow of migrants from the area.

The analysis of recent empirical data made it possible to develop “neutral” migration models (NMM) for the two countries. Although further studies in other countries are required to establish the typical form of these models, and to determine a country’s settings that influence their parameters, it is suggested that these models represent a useful tool for urban and regional development planning. The areas of their potential use are:

1. Allocation of public housing construction. NMM developed for a country as a whole and continuously updated may help local decision makers determine the amount of public construction to be allocated in priority development areas in order to encourage inward migration.

2. Restricting further population growth in overpopulated, specifically core area. NMM may assist in formulating regional development policies aimed at stimulating out-migration from areas that have reached the threshold of their “carrying capacity”. Knowing, for instance, country- specific preconditions for out-migration may lead to restrictive regional policies aimed at preventing further concentration of industrial develop- ment and housing construction in “overdeveloped” areas. These specific policies may include higher rates of taxation, stricter land-use regulation, restrictions on industrial expansion, and other policy mechanisms.

3. Targeting economic activity towards economically distressed areas. NMM may help to identify “better-off” regions and regions with a high risk of out-migration. These data may inform public policies on more effective and precise distribution of geographically specific development priorities, favoured regions that are currently in distress and potentially “problematic” areas. 602 Portnov

NOTES

1. The indicator in question is calculated as the ratio between the standard deviation of a selected development datum (i.e., population density, employment) by geographic districts of a country, and the mean value of this datum for the country as a whole. A high value of this ratio (0.2-0.3 and greater) implies that existing patterns of regional development are considerably uneven, while smaller values indicate that regions of a country are more or less homogenous in respect to the development datum in question. 2. Three statistical districts of Israel – , Samaria, and the Gaza area – were excluded from the analysis due to an assumption that migrations to and from these areas are affected by political and ideological considerations rather than by “routine” factors of in-country migrations. 3. The requirement of data comparability also played its role in selecting administrative districts as geographic units for the present comparative study. It was also taken into consideration that information on subdistricts (), which is available in Japan, is unavailable for corresponding geographic units in Israel. With due regard for this, the original data for prefectures in Japan were aggregated in district sets using the SPSS software. 4. The 1993 data for districts of Japan were not available. 5. An additional consideration was taken into account in determining the time frame for the present case study. In the mid-1980s, the policy of “direct absorption” of new immigrants was introduced in Israel. Under this policy, new immigrants can decide for themselves where they wish to settle. This policy is in sharp contrast to previous policies on immigrant absorption, through which the government intervened directly in new immigrants’ spatial distribution (Lipshitz, 1997). 6. Using the MB/NG index instead of analysing absolute migration rates allowed us to counter the differences in natural growth rates across various geographic areas, which appeared to vary substantially in the countries in question.

REFERENCES

Abe, H. 1996 “New directions for regional development planning in Japan,” in J. Aden and P. Boland (Eds), Regional Development Strategies: A European Perspective, Jessica Kingsley Publishers, London and Bristol: 273-295. Armstrong, H., and J. Taylor 1993 Regional Economics and Policy, Harvester, New York. Balchin, P.N. 1990 Regional Policy in Britain: The North-South Divide, Paul Chapman Publishing Ltd, London. , Z. 1980 “Urbanization processes in the Central Coastal Plain and their effect on a national economic planning and on policy of population dispersal”, in Israel Builds 1977, Ministry of Construction and Housing, Jerusalem: 8-13. The effect of regional inequalities on migration: a comparative analysis 603

Bitan, A., and S. Rubin 1991 Climatic Atlas of Israel for Physical and Environmental Planning and Design, Ramot Publishing Co., Tel Aviv. Bourne, L.S. 1975 Urban Systems: Strategies for Regulation. A Comparison of Policies in Britain, Sweden, Australia and Canada, Clarendon Press, Oxford. CIA 1997 The 1996 World Fact Book, Internet Edition. De Jong, G.F., and J.T. Fawcett 1981 “Motivation for migration: an assessment and a value-expectancy research model”, in G.F. De-Jong and R.W. Gardner (Eds), Migration Decision-Making. Multidisciplinary Approaches to Microlevel Studies in Developed and Developing Countries, Pergamon Press, New York: 13-53. Diamond, D.R., and N.A. Spence 1983 Regional Policy Evaluation, A Methodological Review and the Scottish Example, Gover, Aldershot. Ehrlich, P.R., A.H. Ehrlich, and J.P. Holdren 1972 Ecoscience: Population, Resources, Environment, W.H. Freeman and Company, San Francisco. Fialkoff, C. 1992 “Israel’s housing policy during a period of massive immigration”, in Y. Golani, S. Eldor and M. Garon (Eds), Planning and Housing in Israel in the Wake of Rapid Changes, Ministries of the Interior and of Construction and Housing, Jerusalem: 169-177. Findley, S. 1977 Planning for Internal Migration: A Review of Issues and Policies in Developing Countries, United States Government Printing Office, Washington, DC. Fischer, C.S. 1976 The Urban Experience, Harcourt Brace Jovanovich, Inc., New York. Friedmann, J., and C. Weaver 1979 Territory and Function, Edward Arnold, London. George, P. 1970 “Types of migration of the population according to the professional and social composition of migrants”, in A. Clifford and J. Jansen (Eds), Reading in the Sociology of Migration, Pergamon Press, Oxford: 39-47. ICBS 1997 Statistical Abstract of Israel 1996, Israeli Central Bureau of Statistic, Jerusalem. JSB 1997 Japan. Statistical Yearbook, Japanese Statistics Bureau, Management and Coordination Agency, Tokyo. LaLonde, R.J., and R.H. Topel 1997 “Economic impact of international migration and the economic performance of migrants”, in M.R. Rosenberg and O. Stark (Eds), Handbook of Population and Family Economics, Elsevier, Amsterdam: 800-850. 604 Portnov

Lerman, R., and E. Lerman 1992 “A comprehensive national outline plan for construction, development and absorption of immigrants”, in Y. Golani, S. Eldor and M. Garon (Eds), Planning and Housing in Israel in the Wake of Rapid Changes, Ministries of the Interior and of Construction and Housing, Jerusalem: 29-47. Lipshitz, G. 1997 “Immigrants from the former Soviet Union in the Israeli housing market: spatial aspects of supply and demand”, Urban Studies, 34(3): 471-488. Markusen, A. 1996 “Interaction between regional and industrial policies: evidence from four countries”, International Regional Science Review, 19(1): 49-77. Moore, E.G., and M.W. Rosenberg 1995 “Modeling migration flows of immigrant group in Canada”, Environment and Planning A., 27: 699-714. Newman, D., Y. Gradus, and E. Levinson 1995 “The impact of mass immigration on urban settlements in the Negev 1989-1991”, working paper no. 3, The Negev Center for Regional Develop- ment, Be’er Sheva. Portnov, B. 1998 “The effect of housing on migration in Israel”, Journal of Population Economics, 11(2): 379-394. Portnov, B., and E. Erell 1998 “Long-term development peculiarities of peripheral desert settlements: the case of Israel”, International Journal of Urban and Regional Research, 22(2): 216-232. Portnov, B., and D. Pearlmutter 1997 “Sustainability of population growth: a case study of urban settlement in Israel”, Review of Urban and Regional Development Studies, 9(2): 129-145. Rapoport, A. 1977 Human Aspects of Urban Form: Towards a Man-Environment Approach to Urban Form and Design, Pergamon Press, Oxford and New York. Richardson, H.W. 1977 Regional Growth Theory, MacMillan, London. Uhlenberg, P. 1973 “Non-economic determinants of non-migration: sociological consideration for migration theory”, Rural Sociology, 38(3): 297-311. Vining, D.R., Jr. 1982 “Migration between the core and the periphery”, Scientific American, 247(6): 37-45. Zheng, X.P. 1997 “China’s regional inequalities and related long-term policies”, Review of Urban and Regional Development Studies, 9(2): 115-128. The effect of regional inequalities on migration: a comparative analysis 605

FIGURE 1 FOREIGN IMMIGRATION AND IN-COUNTRY MIGRATION BY DISTRICT OF ISRAEL 1985 - 1995 (thousands)

Immigration – number of new immigrants initially settled in a district in a particular year; In-migration – net annual balance of in-country migrations in a particular district. 606 Portnov

FIGURE 2 DISTRIBUTION OF IN-COUNTRY MIGRANTS (A) AND NEW IMMIGRANTS (B) BETWEEN CORE AND PERIPHERAL DISTRICTS OF ISRAEL, 1985 - 1995 (thousands) 20 17.3

15 9.2 10 8.3 5.8 3.5 4.0 5 3.0 2.4 0.4 -0.5 -0.8 0

-5 -1.9 -2.3 -4.5 -5.9 -10 -8.2 -10 -10.1 -9.6 -11 -15

-16.3 -20

-25 1985-24.2 1995 1985 1995 A

100 92.6 90.8 88.1 90 85.1

80

70

60

50 45.5 44.1 40.5 41.1 41.6 40 37.2 33.5 35.2

30

20 13.5 9.8 10 2.4 1.3 2.2 2.9 2.4 3.9 3.3 3.9 0

1985 1995 1985 1995 B Core districts* Peripheral districts**

* Includes the Jerusalem, the Central and the Tel Aviv districts; ** Includes the Northern, the Haifa and the Southern districts. The effect of regional inequalities on migration: a comparative analysis 607

FIGURE 3 MIGRATION BALANCE BY DISTRICT OF JAPAN 1985 - 1995 (thousands) 608 Portnov

FIGURE 44 A GENERALIZEDA GENERALIZED MODEL MODEL OF THE OF FACTORS THE FACTORS AFFECTING AFFECTINGINTERREGIONAL INTERREGIONAL MIGRATION MIGRATION

Availability of employment in a district Residents of other districts Commuters and current Migrants residents of the area Foreign immigrants

Demand for housing and services

Supply of housing

Time-lags

Prices

Availability of local resources The effect of regional inequalities on migration: a comparative analysis 609

FIGURE 5 NEUTRAL MIGRATION CURVES FOR ISRAEL AND JAPAN

0.20

Japan (data points) Linear fit (R**2=.984) 0.16 Israel (data points) Polynomial fit (R**2=.405)

0.12

Employment 0.08

0.04

0.00

0.00 0.10 0.20 0.30 0.40 Housing construction

Note: For the sake of comparison, the actual values of housing construction (m2) and employment (number of jobs available in a district) are normalized using the overall population size of respective countries – Israel = 5,500,000; Japan = 125,000,000 – as the conditional baseline. 610 Portnov

TABLE 1 SELECTED SOCIO-ECONOMIC INDICATORS IN ISRAEL AND JAPAN

Indicator Israel Japan

Population size (1,000) 5,500 125,500 Land area (sq. km) 21,946 374,744 Population density (per sq. km) 246 335 Overall population growth rate (per cent) 2.11 0.21 GDP (US$ per capita) 15,500 21,300 Unemployment rate (as of 1995, per cent) 6.3 3.1 Telephones (per 1,000 residents) 470 510 Indicators of interregional inequalities:* - population density 1.545 1.002 - availability of employment 0.360 0.914 - housing construction 0.645 0.901

* Indices are calculated using 1995 district statistics (see text for definition). Source: CIA, 1997; ICBS, 1997; JSB, 1997. The effect of regional inequalities on migration: a comparative analysis 611 MB/NG ratio Housing (Israel = 100) = (Israel per centper Employment, TABLE 2 Israeli Central Bureau of Statistics, Jerusalem. of Statistics, Bureau Central Israeli per centper Population share, 1985 1995 1985 1995 1985 1995 1985 1995 Statistical Abstract of Israel, per centper Land area, SELECTED SOCIO-ECONOMIC INDICATORS OF ADMINISTRATIVE DISTRICTS IN ISRAEL, INDICATORS IN OF ADMINISTRATIVE SOCIO-ECONOMIC 1985-1995 DISTRICTS SELECTED District (see Figure 1) JerusalemNorthernHaifaCentralTel AvivSouthern 2.9 20.9Sources: 1985-1997, ICBS, 11.9 16.6 4.0 5.8 12.1 65.6 .8 17.3 14.1 20.9 10.9 12.1 24.4 14.6 13.5 22.0 10.3 13.8 21.3 15.3 13.5 23.6 129 11.4 26.0 13.4 72 23.9 131 12.8 24.3 69 102 86 0.52 130 81 108 0.58 87 .04 138 67 0.42 0.94 0.19 -0.05 0.60 1.92 1.79 -0.50 1.50 612 Portnov MB/NG ratio Housing (Japan = 100) = (Japan per centper TABLE 3 Employment, Japanese Statistics Bureau, Management and Coordination Agency, Tokyo. Agency, Coordination and Management Bureau, Statistics Japanese per centper Population share, 1985 1995 1985 1995 1985 1995 1985 1995 Japan. Statistical Yearbook, per centper Land area, SELECTED SOCIO-ECONOMIC INDICATORS OF ADMINISTRATIVE DISTRICTS IN JAPAN, INDICATORS DISTRICTS IN OF ADMINISTRATIVE SOCIO-ECONOMIC 1985-1995 SELECTED District (see Figure 3) HokkaidoTohoku 22.1KinkiChugoku 19.5Kanto 4.7 7.2 10.1 8.4TokaiShikoku 4.5 9.6 16.6Kyushu 9.8 6.4 4.8Tosan 5.0 4.5 31.1 16.4Hokukiku 10.4 11.2 6.2 9.7Sources: JSB, 1986-96. 3.5 32.2 15.8 4.3 6.0 3.0 11.9 9.9 6.5 9.9 120 31.4 15.9 3.3 3.4 2.6 11.7 96 6.2 10.2 101 33.0 98 3.5 11.2 3.4 2.6 84 95 10.4 130 117 -0.89 10.9 3.2 3.7 2.8 -0.71 99 89 139 -0.17 0.08 99 86 3.6 0.21 2.6 110 -0.49 -0.48 0.64 91 85 85 90 0.12 0.15 0.04 -1.09 -0.69 98 90 0.01 0.25 0.89 -0.33 0.04 0.18 0.38 The effect of regional inequalities on migration: a comparative analysis 613

TABLE 4 FACTORS AFFECTING THE MB/NG RATIO ACROSS ADMINISTRATIVE DISTRICTS OF ISRAEL AND JAPAN (analysis of variance)

Factor* Israel Japan F F sign F F sign

Climate 3.143 .082 .008 .927 Density .094 .761 .210 .811 Economy .025 .876 .577 .450 Employment 1.033 .363 10.232 .002 Employment change 4.364 .041 1.956 .127 Housing construction 5.962 .018 4.258 .042 Housing prices 3.433 .069 4.715 .033 Combined effect 3.357 .003 5.083 .000 Mean square 3.853 1.532 (explained) Mean square (residual) 1.148 .301

Number of observations 66 100

* Initial values of the factors were logarithmically transformed in order to comply with the homogeneity-of-variance assumption.

APPENDIX 1 STATISTICAL PARAMETERS OF THE RESEARCH VARIABLES

Variable Japan Israel Mean Std Dev Mean Std Dev

MB/NG index -.35 .65 .73 .68 Climate 140.00 54.96 1,050.00 283.52 Density 511.66 485.47 1,498.35 2,197.70 Economy 3.34 1.92 2.29 1.61 Employment 5,909.34 4,939.19 243.08 88.56 Employment change 89.96 270.16 9.04 12.37 Housing construction 11,878.01 11,116.82 722.03 513.16 Housing prices 97.30 13.90 98.15 27.71 614 Portnov

L’EFFET DES INÉGALITÉS RÉGIONALES SUR LA MIGRATION: UNE ANALYSE COMPARATIVE D’ISRAËL ET DU JAPON

Cet article pose trois grandes questions:

- Y-a-t-il un mécanisme général qui fait que les disparités de développement régional se répercutent sur les schémas migratoires transrégionaux?

- Quels aspects des inégalités régionales (climat, emplois, logements, etc.) ont l’effet le plus marqué sur le rythme et la direction des migrations transrégionales?

- Quelles politiques et stratégies de planification sont de nature à faire paraître plus attrayante la migration vers les régions à développement périphérique?

Pour tenter de répondre à ces questions, l’auteur utilise les données statistiques recueillies entre 1985 et 1995 sur deux pays relativement petits et à forte densité de population: Israël et le Japon.

Un modèle général de facteurs influant sur la migration transrégionale est proposé, et une analyse des variances est utilisée pour expliquer les facteurs influant sur le rythme des migrations transrégionales dans les deux pays.

Même si ces deux pays diffèrent considérablement quant à la taille de leur population et à leur développement local, ils semblent présenter des similitudes certaines quant à leurs schémas de migration transrégionale.

Depuis la fin des années 80, le caractère attractif des régions centrales dans les deux pays a plutôt diminué, tandis que celui des régions périphériques s’est accru. L’auteur avance l’argument selon lequel cette similitude de schémas est due à un équilibre similaire de l’emploi et de la disponibilité de logements dans différentes zones géographiques.

EFECTO DE LAS DESIGUALDADES REGIONALES EN LA MIGRACIÓN: UN ANÁLISIS COMPARATIVO DE ISRAEL Y EL JAPÓN

Este artículo plantea tres importantes preguntas:

- ¿Existe un “mecanismo” general a través del cual las disparidades en el desarrollo regional afectan los patrones de la migración entre distritos? The effect of regional inequalities on migration: a comparative analysis 615

- ¿Qué aspectos de las desigualdades regionales (clima, empleo, disponibilidad de viviendas, etc.) han afectado más profundamente los porcentajes y direcciones de la migración interregional?

- ¿Qué planificación de políticas y estrategias conduce a acrecentar la atracción de la migración a regiones de desarrollo periféricas?

En un intento por responder a estas preguntas, se utilizan los datos estadísticos acopiados entre 1985 y 1995 para dos países relativamente pequeños y densamente poblados, a saber, Israel y el Japón.

Se propone un modelo general de factores que afectan la migración entre distritos, y también se utiliza un análisis de variación para explicar los factores que inciden en el porcentaje de migraciones interregionales en los dos países.

Si bien estos países difieren considerablemente en el tamaño de su población y en el desarrollo local, tienen considerables similitudes en los patrones generales de migración entre distritos.

Desde finales de 1980, la atracción de las regiones centrales en ambos países tiende a disminuir, mientras que la de las regiones periféricas parece acrecentarse. Por ello, se arguye que esta similitud de patrones de migración tiene atinencia a un equilibrio similar de empleo y disponibilidad de viviendas en las distintas regiones geográficas.