Labour Migration in Less Developed Countries: Comment

Myron J. Frankman
McGill University

In the March 1969 issue of the American Economic Review , Michael Todaro formulated “an economic behavorial model of rural-urban migration” (p. 138) to serve as a framework in the analysis of the employment effects of alternative economic policies. While Todaro and John Harris have elaborated upon and expanded the model (1970), the basic model originally developed by Todaro remains essentially unaltered. It is the purpose of this communication to call attention to some of the shortcomings in Todaro’s basic model and to take issue with the policy recommendations which he has derived from it.

Central to the labour migration model formulated by Torado is the comparison by the potential migrant of the present value of expected income streams from employment in the rural and urban areas. When the present value of the expected urban real income stream exceeds that of the rural income stream (or vice versa ), migration occurs.

The focus of attention is on rural to urban migration, which is the prevalent historical and present trend. (1) However, the model is essentially reversible, although asymmetrically as a scalar proxy, a > 1, for "city lights" and other amenities is included in the calculation of net urban real income. Similarly, while no definition of the term "potential migrant" is offered, it is, in theory, applicable to all workers, whether rural or urban. All workers are potential migrants and all potential migrants are homogeneous, as may be observed from the following assumptions: all workers have identical planning horizons, have identical fixed costs of migration, and use identical discount factors, which are constant over the planning horizon. (Todaro:1969, p. 141). Workers are also homogeneous in the economic sense of being unskilled. However, these unskilled workers are assumed to be able to shift without retraining from the rural sector to either the urban "traditional" sector or the urban "modern" sector.

It would appear more realistic to assume that the planning horizon is more likely to be the same for workers within the same age bracket. Empirical evidence amply demonstrates that the average age of migrants is well below the average for the labour force as a whole. (2) Fixed costs of migration, on the other hand, are likely to vary directly with family size and with the distance from the worker's present location to the contemplated destination. The use of a constant rate for discounting future earnings is also subject to challenge. What of the migrant who attaches greater importance to the opportunities for his sons' improved economic and social situation rather than his own or the migrant who is willing to forego higher immediate earnings in rural employment in order to acquire skills for which training is available only in urban areas? A method which heavily discounts remote earnings cannot adequately explain the behaviour of these individuals.

In the calculations made by the model’s potential migrant, the expected value of urban real income, E[Y u(t)], in period t varies directly with p 1(t), the probability of having a modern sector job in that period (Todaro:1969, p.142).

where, Y l(t) represents the net urban modern sector real income in period t.

The lower the probability that the potential migrant will be absorbed in the modern sector, the less likely it is that he actually will migrate. Thus the rate of flow of migrants from rural areas and urban unemployment rates are both equilibrated by the effect of p 1(t) on the expected value of urban income. The probability variable p 1(t) is, in turn, directly related to the probability p 1(t) of having been selected from the pool of urban traditional workers in that period or any previous period. This latter probability is empirically determinable, being a ratio between the modern sector job openings in period t relative to the number of accumulated job seekers in the urban traditional sector. In fact, however, the potential migrant, not possessing perfect knowledge, is likely to overestimate the numerator and underestimate the denominator of the job selection fraction ¶ 1(t), with a consequent overestimate of the present value of urban income.

In his otherwise rigourous model, Todaro does not precisely define the urban traditional sector, but notes that it includes “the overtly unemployed, the underemployed or sporadically employed, and those who grind out a meagre existence in petty retail trades and services”. (1969, p. 139) With the exception of the overtly unemployed, all others receive some income. However, employment in the urban traditional sector is regarded in equation (1) as if it provided no income whatsoever. Excluding overt unemployment from the definition of the urban traditional sector we may rewrite equation (1) as follows:

where, p 2(t) is the probability of having an urban traditional sector job in period t, and

Y2(t) represents the net urban traditional sector real income in period t.

It is my contention that the urban traditional sector does not merely provide a positive income, but is likely to be more remunerative on the average than rural underemployment, the latter being the prospect facing many rural workers.

Insofar as modern sector employment largely requires semi-skilled and skilled workers, p 1(t) for the unskilled migrant is likely to be extremely low. (3) Indeed, there may be little chance of obtaining modern sector employment within the planning horizon of the unskilled worker. If, nonetheless, the worker not only decides to migrate, but also decides, after a period of acclimatization, to remain in the urban area, (4) then we may conclude that the present value of the expected income stream from urban traditional employment (which includes an adjustment by a>1, the scalar proxy for city lights) is at least comparable to that from rural employment. (5)
If on the average urban traditional employment is more lucrative than rural employment and if the probability of overt unemployment is relatively small, then an alternative explanation must be sought to explain why cities have not grown even more rapidly through internal migration. The answer may well rest in non-economic factors (Balán: 1969). (6)

Equation (1) leads Todaro to suggest that "one could easily conceive of a situation in which the urban-rural real income differential .... was positive, while the 'expected' differential .... was negative". (1969, P. 142). Equation (2) leads me to suggest that this occurs infrequently and hence rural-urban migration is less readily equilibrated than the Todaro model suggests. Nor can one necessarily rely on changes in the domestic terms of trade between manufacturing and agriculture to reduce the expected urban-rural income differential as increases in the price of wage goods are likely to lead to price controls and/or the free importation of foodstuffs.

While disagreeing with portions of Todaro's analysis, I concur with his general conclusion that the urban-rural earnings differential should be lowered. However, I do not agree with Todaro (1969, p. 146), Lloyd Reynolds (1965, p. 39), and others who suggest that this suppressing wage increases in the urban modern sector. What is called for are efforts to increase the articulation between the urban and rural sectors of the economy and to engage in precisely that which Todaro decries, i.e. the "premature mechanization" of agriculture (1969, p. 146). It is investment in agriculture, with its relatively short gestation period and low capital-output ratio which offers the best prospect for narrowing the urban-rural earnings differential through its effect on the output and price of wage goods. Investment in agriculture is likely to be far more effective in reducing the earnings differential and more feasible politically than the combination of an incomes policy with measures to induce more efficient use of existing capital resources in agriculture that Todaro recommends (1969, p. 146).

Premature mechanization of agriculture is criticized by Todaro because of the problems it poses for rural labour absorption. However, mechanization generates employment in such related activities as transportation, commerce, communications., finance, administration and maintenance. Werner Baer and Michel Hervé have argued that “modern industry requires a substantial service sector in order to function”. (1966, p. 104) Modern agriculture also requires a substantial service sector.

Todaro acknowledges t1int material progress has historically been accompanied by a transfer of labour from rural to urban areas, yet he envisions the slowing or reversing of this trend by "bringing 'city lights' to the countryside". (1969, p. 149) W.A. Lewis in answering the question "Is Economic Growth Desirable?", places the increase in the range of human choice at the head of his list of the advantages of growth (1955, p. 420). Is it not the increase in the range of choice that is at the heart of urbanization? One may be able in some measure to bring city lights to the countryside, but it can hardly be expected to substitute for the diversity found in urban areas. It is as unlikely that the desire for urban living can be overcome as it is that the desire for economic development itself can be overcome. Rather than recommending the reversal of historical trends, it would seem more fruitful were economists to concentrate on policies by which these trends could be turned to developmental advantage.


1. Modification of the model would be necessary to accommodate migration between urban areas or between rural areas. Urban to urban migration is of particular importance in Latin America where migration to large cities comes primarily from either secondary urban centers or towns rather than farms (Herrick:1965 and Mangin:1967, p. 69).

2. Of economically active recent migrants to Santiago, Chile in a sample studied by Bruce Herrick, 31.3 percent had come between the ages of 16-20, while 63.9 percent were between the ages of 16-30 at the time of migration (1965, p. 76).

3. The probability ¶ l(t) of having been selected from the pool of urban unskilled traditional workers is more properly a ratio between the job openings in the modern sector for unskilled workers and the size of the pool of unskilled.

4. In Latin America, where population pressure on the land is largely absent, migration back to the land is relatively small (Mangin:1967, p. 89).

5. Some potential migrants may also adjust expected rural real income b<1, a scalar proxy for "rural hardship".

6. It is possible that for some rural workers the urban and rural scalar proxies are reversed, such that a<1 and b>1.


1. Baer, W. and Hervé, M., "Employment and Industrialization in Developing Countries." Quart. Jour. Econ. , Vol. 80, Feb. 1966, pp. 88-107.

2. Balán, J., “Migrant-Native Socioeconomic Differences in Latin American Cities: a Structural Analysis,” Latin Am. Research Rev. , Vol. 4, Spring 1969, pp. 3-29.

3. Herrick B., Urban Migration and Economic Development in Chile , Cambridge, Mass., 1965.

4. Lewis, W.A., The Theory of Economic Growth , Homewood, Ill., 1955.

5. Mangin, W., “Latin American Squatter Settlements: a Problem and a Solution,” Latin Am. Research Rev. , Vol. 2, Summer 1967, pp. 65-98.

6. Reynolds, L.G., "Wages and Employment in a Labor-Surplus Economy," Am. Econ. Rev. , Vol. 55, March 1965, pp. 19-39.

7. Todaro, "A Model of Labor Migration and Urban Unemployment in Less Developed Countries," Am. Econ. Rev. , Vol. 59, March 1969, pp. 138-148.

8. Todaro, M.P. and Harris, J.R., "Migration, Unemployment and Development: a Two-Sector Analysis," Am. Econ. Rev. , Vol. 60, March 1970, pp. 126-142.