91. Robert E. B. Lucas

Internal Migration and Urbanization: Recent Contributions and New Evidence

November 1998

SUMMARY OF POINTS

Patterns of Internal Migration
Changing place of residence at least once is quite common in a wide range of countries. Cross country comparisons of migration are made difficult by lack of comparable data. However, four recent LSMS enable comparisons of place to place movement of adults since birth. The fraction of adults who had moved were 61% in Russia (1995), 53 % in Ghana (1988), 35 % in Pakistan (1991) and 23 % in Vietnam (1993). Census data show 8 percent of Thailand's population moved in the five years from 1985-90 and 7 percent of India's population moved from 1976-1981.

By no means all internal migrants move from rural to urban areas. In some countries, (e.g. India and Ethiopia) intra-rural movement is the most common mode, while in others (Republic of Korea and Peru) inter-urban movement dominates.

Mobility of rural populations varies considerably. In Ghana and Russia, less than half of all rural resident adults were born in the same place that they now live. In contrast, more than 83 percent of Vietnam's and 68 percent of Pakistan's village adults report being born in the village where they now live.

Seasonal and temporary migration within the rural sector adds significantly to the relative importance of rural-rural movements in some contexts (such as northern Thailand).

Migrants are much more likely to move shorter distances. In India, much of the internal migration is therefore within a district. Even rural-urban migrants often move residence over a short distance suggesting that commuting is not always an attractive or viable alternative.

The fraction of population living in urban areas is significantly higher among countries with higher incomes, though the rise in degree of urbanization tapers off at higher incomes. Even given the level of GDP per capita, the percent of population in urban areas rose significantly over time at the rate of just over 3.5 percentage points per decade from 1960 to 1990.

Among the 26 developing countries for which data exist, migration plus reclassification of rural areas as urban areas contributed about 39 percent of urban population growth during the 1980s.

The 'residual' method applied to intercensal data is incapable of distinguishing the role of net urban migration from that of urban reclassification in total urban population growth. More detailed data for India suggest that migration alone contributed about 19 percent of urban population growth from 1960-1980 while reclassification contributed a further 28 percent (20 percent from new towns alone). Similarly, in China where migration plus reclassification has played an important role in total urban population growth, the number of towns more than quadrupled from 1982 to 1990.

However, a substantial fraction of adults residing in urban areas report having been born in rural areas -- 30 percent in Russia, 29 percent in Vietnam, 20 percent in Pakistan and 18 percent in Ghana. A smaller fraction of the adults living in larger urban areas report being born in a rural area than do the adult populations of the smaller towns

The role of migration plus reclassification in urban population growth varies considerably across countries (with a coefficient of variation of 0.4 in our sample of 26 developing countries during the 1980s).

In a broader sample of 46 developing countries from 1960 to 1990, the average contribution of migration plus reclassification to urban population growth is highest in East and South East Asia, followed by South Asia and Sub-Saharan Africa. North Africa and the Middle East, South and Central America are quite similar and have the lowest average contributions.

Countries in which a large fraction of the population already live in urban areas exhibit a significantly smaller role for migration and reclassification in contributing to urban population growth. A diminishing role for rural-urban migration is, of course, to be expected as the rural population base dwindles.

The role of migration plus reclassification in urban population growth is also lower among the higher income developing countries.

However the negative association with income levels is largely associated with the larger extent of urbanization in the higher income countries. Once the degree of urbanization is controlled for, the pattern with respect to income levels alters. At lower income levels the role of migration and reclassification in urban population growth rises with income, but beyond a GDP per capita of about 4000 (1985) US dollars this association reverses.

Even controlling for the degree of urbanization and income levels, considerable regional variation exists in the contribution of migration and reclassification to urbanization. This role is highest in East Asia and South America, followed by North Africa and the Middle East, then Central America and the Carribean, Sub-Saharan Africa and finally South Asia with the lowest contribution. These regional differences may reflect many components (including geography, fertility, and development strategies).

Despite the controls which have existed on migration in China, Russia and Vietnam rural-urban migration has been substantial. At least in China, this has been particularly true as market reforms have limited the efficacy of some of the migration controls.

Our observations suggest that most of the lifetime urban-urban migrants living in large urban areas were born in other large urban areas, rather than small towns.

In most countries, immigrants tend to concentrate in urban areas and the fraction of the adult population who have immigrated is typically higher in large towns than in small towns.



The Causes of Internal Migration

Earnings and employment opportunities
It is well established that the greater the gap in earnings between origin and destination the more likely are working age adults to move.

It is far less clear how the significant gap in earnings is maintained between rural and urban, formal employment for observationally equivalent workers. There is some recent evidence consistent with efficiency wage explanations of high urban wages and, in some contexts, urban collective bargaining seems to play a significant role. High urban public sector pay has also been a factor in attracting urban migrants, though in some regions this premium has diminished.

Many migrants to urban areas initially enter the "informal" sector. For some (though by no means all) this is a transitory phase prior to finding more formal employment. However, statistical studies of these patterns are plagued by the lack of precision in defining the informal sector and the evidence is (consequently) mixed as to whether the formal or informal urban sector offers higher pay to observationally equivalent workers.

Migrants to town initially earn less than observationally equivalent natives, but the evidence indicates that this gap disappears within a few years and may even reverse.

Findings on whether differences in unemployment rates between locations promote migration are mixed. Limited evidence suggests that migrants often identify their urban job before migrating, but other migrants do appear to search for work after moving, either while in temporary employment or while openly unemployed. However, at least one study maintains that off-farm migration in developing countries will cease only when the earnings gap is entirely closed and hence that uncertainty about employment is irrelevant. It has also been argued (with some supporting evidence) that unemployed workers may have at least as high a chance of re-employment in their home setting where information and contacts are more readily available.

The location of newly created employment opportunities depends in part upon the development strategy adopted. The hypothesis that import substitution leads to employment concentration in large cities lacks systematic testing, though a case study of India suggests that liberalization has been a factor in promoting the emergence of new towns. However, many other factors also affect the spatial distribution of industrial employment, including the land intensity of production, availability of appropriate infrastructure, agglomeration externalities, tax incentives tied to location, as well as the local cost of labor.

It has been hypothesized that large towns offer a greater diversity of employment and hence a better chance of re-employment in the event of a lay-off. This might render large towns more attractive to migrating workers. At least one study shows that unemployment rates are lower in larger urban areas, though more systematic evidence on this issue does not seem to exist for the developing economies.

A case study of Korea suggests an interesting shift in the internal location of industry as development proceeds. As city center land prices rise, land-intensive, low skill manufacturing shifts out of the cities. Simultaneously, at least three factors may contribute to the further concentration of skilled and professional employment in the metropolitan areas: a preference for city living among wealthier, professional classes; externalities to agglomeration of corporate headquarters; and externalities generated by the presence of other skilled and professional employees.

Family and networks
Possessing a network of family and friends in town may encourage migration into town. Conversely, a well-developed network at home may discourage departure. However, empirical examination of these propositions is hampered by difficulties in eliminating spurious effects, in discerning what advantages the network offers and in defining the scope of a network.

Migration at the time of marriage, to join or accompany a spouse, does seem common (though not well documented). A few studies also suggest that parents may have the welfare of their off-spring in mind when making their own migration decision.

Urban migrants often initially settle in ethnically similar neighborhoods, which suggests that networks lower the effective cost of moving in some manner (though this may include provision of security from ethnic conflict). Evidence from some cities even shows that subsequent moves within the city are to new neighborhoods with the same ethnic composition. Moreover, a case study of Cairo is consistent with kinship, regional and common occupational networks affecting labor market outcomes.

Distance
Migration over short distances is much more common than migration to remote locations. Whether this reflects the greater cost of moving further, lack of information about more remote alternatives or less alienation in a nearby setting remains undetermined.

However, fewer out-migration opportunities in remote areas tend to generate longer residence which in turn raises the sense of alienation in contemplating moves. The result is pockets of poverty in remote places.

Wealth and capital markets
Incomplete or imperfect local capital markets may encourage out-migration either directly through restrictions on the ability of families to borrow, or indirectly through effects on employment creation. However the solitary test of this potentially interesting proposition proves inconclusive.

The opportunity cost of financing costly migrations is probably lower for wealthier families. This has two important implications. First, other things equal, migration may be more common from richer families and this in turn may exacerbate the inequality in incomes. Second, as a region becomes wealthier out-migration may actually increase (up to a point) as the financial constraint is reduced.

Empirical evidence on these two implications is mixed and controversial. Only a few cross-family studies examine the wealth effect and the results are too mixed to reach any conclusion. Some historical studies do show rising emigration as GDP increases but this is probably largely a reflection of the demographic transition and altered patterns of employment rather than an alleviation of a financing constraint.

Family strategies to contain risks
One way that families may insure themselves is by having members migrate to locations where times of economic adversity do not normally coincide with those at home. Remittances between the home base and migrant then enable consumption smoothing.

There is some evidence consistent with the remittance portion of this scenario. However no direct test of whether migration is greater from communities with higher economic risk seems to exist.

Relative economic standing in the community.
Migrants may not only care about any absolute gain in earnings through migration but also about their relative economic standing in the communities of origin and destination. One study of migration from rural Mexico found that emigration to the US (but not internal migration) was more likely among individuals with low incomes relative to others in their village. However it would clearly be premature to generalize from this evidence.

Availability and quality of amenities.
Improved amenities in a location may attract industry or permit agricultural expansion. To the extent that this results in employment expansion or higher wages, out-migration may be discouraged and in-migration encouraged.

Improved local amenities may also have a direct effect upon migrants' decisions, simply by making life in this setting more attractive.

On the other hand, some forms of improved local amenities could exacerbate net out-migration. It is feasible that improved rural transport could act in this fashion, by affecting local production patterns and hence perhaps reducing the local demand for labor, and perhaps by making departure and return visits easier (though commuting also becomes an easier alternative to migration).

Unfortunately no evidence appears to exist on the effects of amenities on migration outcomes in the developing countries. This is a major lacuna in our information. The limited US evidence suggests that amenities indeed affect migration decisions, but the relevant form of amenity differs by population group.

Incidence of violence, disease or disasters.
It is obvious that episodes of violence and natural disasters result in mass migrations either of internally displaced persons or of international refugees. What is far less well documented is the extent to which on-going violence, political repression and recurrent risk from disasters swell the stream of migrants.

Studies in a couple of South America countries have shown that higher local assassination rates result in significantly greater out-migration, apparently in addition to any effects that violence has on regional wage differentials. Indeed the study in Guatemala finds no evidence that the incidence of violence alters migrants' responses to given earnings differentials. However, a cross-country analysis of sub-Saharan Africa does find that restrictions on civil liberties reduce migrants' responsiveness to economic incentives to migrate.

Migration controls and incentives
A few countries have attempted to restrict (or even to force) internal migrations. Unless the state is prepared to take Draconian measures such controls are usually ineffective. In a number of contexts it has been found that expelled migrants soon return. In some of the socialist states, access to jobs, housing, food rations and other state benefits have been tied to a specific location, effectively preventing migration by removing the incentive to move. However, at least in China, the emergence of a more market oriented system has eliminated the efficacy of these controls and migration has duly expanded.

The Economic Consequences of Migration

A mobile labor force can be an important ingredient in enabling more efficient production in an economy. Where high wages signal high productivity, migration for wage gains generally enhances the efficiency of production.

Migration may, however, prove either excessive or insufficient to achieve maximum efficiency when market prices fail to convey appropriate signals. These failures include various possibilities for inappropriate wage setting either as a result of market failures or as a result of policies which affect labor costs. Incomplete and imperfect capital markets can also result in misleading wage signals, as can trade or other policies which distort commodity prices. In addition, socially sub-optimal rates of migration prevail when the costs of migration are not entirely born by the person or family making the migration decision., which is particularly true with respect to overcrowding costs.

A substantial literature exists on optimal policy packages to deal with specific combinations of distortions. However, implementation of most of these recommendations requires considerable information to achieve an optimal result and, in some cases, even to determine whether a tax or subsidy is required.

There are too few studies of the total contribution of internal migration to productive efficiency to generalize. A few CGE models permit such analysis but usually omit many of the market imperfections which render the question interesting. A recent evaluation suggests that internal migration contributed about 1% to 1.5% of GDP in Peru in 1979, but this relatively low estimate omits both general equilibrium effects (such as linkages) and dynamic effects (such as accumulation) resulting from the migration.

In the face of urban un(der)employment it is tempting to recommend a policy of urban job creation. The well-known Harris-Todaro model points out that such a strategy may induce so much additional rural-urban migration as to leave more people unemployed and to diminish total production. These effects may be mitigated, to some extent, when local linkages result in expansion of the urban informal sector induced by an expansion in the formal sector, and by any tendency for new jobs to be taken by initial urban residents on a preferential basis. However, little evidence exists on either feature for the developing countries.

Much of the early analysis of rural-urban migration in the developing countries assumed urban formal sector wages to be rigidly set by institutional forces. In contrast, there is a growing body of evidence that wages do respond to downward pressures in the presence of unemployment. For example, one study concludes that a doubling of unemployment in urban Cote d'Ivoire caused wages to fall by about 12%. Time series evidence for other countries support the notion of a wage-curve, in which earnings respond to unemployment albeit with a lag. This raises serious doubts about policy recommendations emerging from the literature on rigid wages in which unemployment equilibrates migration flows. Nonetheless some (temporary) corrective action may be warranted, to improve efficiency, where speed of wage adjustment is excessively slow.

Migration may also impact the rate of savings and accumulation in an economy and hence, perhaps, growth. In particular, it is commonly held that temporary migrants save a larger fraction of their earnings because risk averse migrants save for their return to a lower and less certain income and because the marginal utility derived from consumption while away from the family is low. However, supporting evidence in the context of temporary internal migrants is lacking. Moreover, temporary migration may only raise the propensity to save temporarily.

Migration may not only change the efficiency of production but profoundly alter the distribution of income through a number of channels.

Migrants presumably gain from migration unless they make errors in judgement, or a gamble with respect to migration fails to pay off, or migration is not of the migrant's own free will.

Nonetheless the extent of social mobility associated with migration may vary. The stereo-type of rural-urban migration in Latin America depicts three concentric bands of urban settlement. Bridge head, single migrants initially settle in rented accommodation in the city center where access to work is easier. As the migrant accumulates more wealth and marries, he or she moves to self-help housing on the periphery of the city. Further accumulation permits upgrading this accommodation. The city limits ultimately engulf the intermediate settlement site and the migrant is left in low-income housing between zones of rental housing and shanty towns. However, a study of Quito finds a quite uniform distribution of migrants across the city, which is attributed to policy restrictions on self-help housing and to rent control. Evidence from India suggests that a tiny group of urban migrant households fare extremely poorly, but the average migrant household enjoys a higher living standard than non-migrants, particularly after some time in town. A recent case study of frontier migration in the Tarai of Nepal also reports upward social mobility among migrants but, in this context, the extent of upward mobility is reported to be tiny.

Migration also affects the incomes of people, both at origin and destination, who do not move. One way that this happens is by altering the pattern of earnings among non-migrants as the supply of migrant labor shifts. Whether wages at origin rise and those at destination decline is, however, not obvious. The skill mix of the migrants in relation to the non-migrants, the nature of substitution between skill categories of labor in production, induced shifts in the pattern of production, and the influence of scale economies each play a role in this outcome. As a result, the evidence on whether migration results in wage convergence is very mixed.

In the longer run, the departure of skilled migrants can raise the returns to education and training of those left behind, resulting in greater investments in human capital and higher income. Countering this are at least two forces. First, there is some evidence of agglomeration economies driven by a pool of well educated workers. This can imply that departure of skilled personnel actually lowers the returns to education. Second, the education of children left behind by migrating parents faces two opposing forces: migration may provide the resources to finance more and better education; but lack of parental presence may lower commitment to schooling.

The other major route through which migration may impact incomes of non-migrants is through remittances. The extent to which poor and rich rural families benefit from this is a matter of some dispute. Early village studies in India suggest that rural-urban migration is rare among the very poorest of rural households, more common among agricultural laboring families, declines again among somewhat better-off village households, but that the educated children of the rural elite commonly move to town. Combined with village study observations that net remittances from town to village are small and that the children of the wealthy are more likely to retain their rural ties and to remit, this implies that remittances may largely benefit relatively affluent rural families. However evidence from a household survey in Mexico presents a different picture in which the dominant form of migration from a village (either internal migration or departure to the US) results in remittances which reduce village income inequality.

Family Risk Strategies

Individual migrants may remain in more or less active contact with other family members who stay at home, through visits, sending remittances and perhaps ultimately returning home. To the extent that migrants essentially remain active members of the home group, a geographically extended family exists, perhaps straddling separate settlements within the rural sector, straddling the rural-urban divide, or even international boundaries.

A possible advantage of this strategy is that a family may be able to self-insure, to some extent, by placing members in alternative settings where times of economic crisis do not normally coincide.

If many families actively straddle the rural-urban divide then perceiving the urban sector and its development as separate from that of the rural sector can be quite misleading.

There is evidence consistent with consumption smoothing through income sharing within villages and even within wider ethnic communities but direct tests on families, abilities to smooth consumption through migration have not been conducted.

However studies in Botswana, India and Kenya have each found remittance patterns consistent with family arrangements to offer mutual insurance through migration. In contrast, one study in semi-arid India found that temporary local wage employment was a more common vehicle for insurance in that context. Where commuting for temporary wage employment during a bad state of the local economy is cost-effective, migration and remittance for insurance may be less necessary, but this remains untested.

Ordering Information: