Rural-urban mobility in Thailand: A decision-making ... - Springer Link

DEMOGRAPHY©

Volume 22, Number 4

November 1985

RURAL-URBAN MOBILITY IN THAILAND: A DECISION-MAKING APPROACH Theodore D. Fuller Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061 Paul Lightfoot University of Hull, Hull, England Peerasit Kamnuansilpa National Institute of Development Administration, Bangkok 10240, Thailand

INTRODUCTION

This paper develops an individual, behavioral model of the decision-making process for rural-urban mobility, and applies this model to data from a Northeast Thai context dominated by circular rural-urban movement. Ultimately, the model attempts to explain the movement patterns of villagers. A vast literature attempts to identify the phenomena that regulate the mobility process. The importance of economic determinants of migration patterns is well documented (Shaw, 1975; Arnold and Cochrane, 1980; Todaro, 1969, 1976). Numerous studies, however, lead to the inescapable conclusion that simple economic models do not provide adequate explanations of migration patterns (Chamratrithirong, 1976; Roberts, 1981; Rhoda, 1983). The literature has identified a number of noneconomic variables that influence the migration process as well, including transportation networks (Prasartkul, 1977), educational opportunities (Rhoda, 1983), the "bright lights" phenomenon (Vichit-Vadakan, 1983), community norms (Hugo, 1981), and household-level dynamics (Findley, 1980; Harbison, 1981; Lightfoot et al., 1983). While we acknowledge the importance of macro-, community-, and householdlevel phenomena as regulators of the mobility process, our model focuses on individual-level factors in an effort to refine our understanding of rural-urban mobility. Our focus on the individual level permits us to examine more closely why-given an overall objective context-some people move while others do not. The model is developed in a context dominated by circular rural-urban movement. I Various forms of circulation are now important in Southeast Asia and are quantitatively far more significant than the types of permanent migration recorded by national censuses or traditional surveys (Goldstein, 1978; Hugo, 1982). Like other southeast Asian countries, Thailand has substantial circular rural-urban movement. This movement in Thailand has been recognized for at least two decades (Textor, 1961; deYoung, 1963) and recently has been quantitatively detailed in village studies in Central, North, and Northeast Thailand (Lauro, 1979; Singhanetra-Renard, 1981; Lightfoot et al., 1983). MODEL

We use five explanatory variables in our theoretical model of mobility behavior: past mobility experience, urban social contacts, information about urban areas, evaluations of alternative destinations, and mobility plans. Operational definitions for each variable are provided in a later section. But first, hypothesized relationships among the five conceptual variables are discussed. As has been recognized repeatedly in the migration literature, past mobility 565

566

DEMOGRAPHY, volume 22, number 4, November 1985

patterns are likely to influence present and future levels and directions of movement. At least two components of this pattern can be distinguished conceptually. First, objective conditions which created a mobility stream in the first place (e.g. differential wage rates) are likely to continue, thereby maintaining the migration stream. Second, the very fact that migrants have gone from point A to point B creates conditions which facilitate the attraction of new migrants between these two points (e.g. the flow of information). This implies considerable inertia in the spatial patterns of migration. Distinguishing the second aspect of this inertia from the first enables us to recognize that current mobility streams will not necessarily correspond exactly to current objective realities. Social contacts have a well-recognized impact on migration (Schwarzweller et aI., 1971; Findley, 1977). In the Thai context, Rabibhadana (1975) has noted that migrants in Bangkok often find employment and lodging near or with relatives. By providing channels for information transmission and mechanisms for sponsorship, social contacts at a destination facilitate movement to that place. Social contacts may also have a small effect on the "objective conditions" in the cities of destination. Much urban employment occurs in the traditional rather than the modern sector; within the traditional urban sector, a degree of social obligation exists between the workers and owners of economic enterprises (McGee, 1968). Consequently, new opportunities may be created as a matter of social obligation rather than economic need. While this effect may be small in relation to the total urban economy, it may be crucial for the individual migrant. Migrants who are moving within well-established migration streams will have a better chance offinding employment in the city than migrants from places without a tradition of movement to that destination. 2 Information is a key intervening variable, partly because it can be expected to have considerable explanatory power in an analysis of existing patterns of movement, and partly because it is one of the very few variables associated with movement which can be manipulated as a means of altering patterns of movement. Because they possess only limited information about opportunities, migrants typically consider very few, if any, alternative destinations before moving (Goodman, 1981). Consequently, information has an important influence on the selection of a destination (Findley, 1977; Rhoda, 1983). Ritchey (1976:392), echoing a point made earlier by Speare (1971), notes that "relatives and friends are the migrant's major source of information about the receiver area prior to migration." Information may also affect propensity to move by affecting the villager's perception of his welfare in the village (Byerlee, 1974) and thereby influencing evaluations. Considerable attention has been given to community satisfaction as a determinant of movement, at least since Wolpert's (1965) introduction of the notion of "place utility." Using Wolpert's work as a springboard, Speare (1974) proposed a model of mobility decision-making with two stages, one involving a decision about whether to move, and another involving a decision about where to move. Various empirical studies explore the notion of subjective measures as predictors of movement and, in particular, the notion of a two-stage model. Using individual-level United States data, Speare (1974) documented the importance of subjective evaluative measures in a study of residential mobility. Speare's model and data suggest that background variables (e.g. age, proportion of friends and relatives nearby) affect one's level of satisfaction but have no direct effect on plans to move or subsequent mobility. Bach and Smith (1977) examine community satisfaction, mobility plans, and subsequent mobility in the context of intercounty migration; their results generally support the utility of a two-stage model of mobility. Extending these ideas beyond the work of Speare and of Bach and Smith, we apply the notion of evaluation of places in a

567

Rural-Urban Mobility in Thailand

context of circulation, rather than one of residential mobility or "permanent" migration. 3 At some point in the decision-making process there must be a final calculus of the desirability and practicality of moving. The villager evaluates the situation in terms of what he knows about potential destinations and what he recognizes as his own aspirations and obligations. Various contingencies may intervene between his stated intentions concerning moving to town and each of the independent variables which help determine those plans, or between these stated intentions and actual mobility behavior. It is reasonable to suppose, however, that people's stated intentions represent fairly realistic analyses of opportunities, aspirations, and constraints for each respondent. Speare (1974) and Bach and Smith (1977) find plans to move are significantly related to subsequent mobility. In sum, the model focuses attention on a number of social and cognitive factorsmobility experience, social contacts, information, evaluations, and plans-all of which can be operationalized at the individual level, and these are hypothesized to constitute a chain offactors that collectively influence an individual's propensity to move to a particular destination. Specifically, each independent variable is hypothesized to influence each subsequent variable in the model. Using a longitudinal design, we will investigate the model separately for Bangkok and Northeast Thai towns so that comparisons can be made regarding mobility to these two destinations. There are several reasons for this. Many villagers in Northeast Thailand have more contact with Bangkok than with towns in their own region, both in terms of movement and in terms of urban social contacts. Since the villagers apparently make a distinction between Bangkok and Northeast towns, we probably should too. Furthermore, given the Thai government's interest in slowing down growth of Bangkok while encouraging the growth of regional urban centers (Fuller et aI., 1983, 1985), it is important to understand the similarities and differences between migration to Bangkok and the Northeast towns. 4 THE SITE

The data presented were collected in six villages in Northeast Thailand during two waves of interviews: July-September 1978 and July-August 1979. The Northeast Region of Thailand has been characterized by substantial out-migration for decades, much of it directed toward Bangkok. Within the Northeast, Roi-et Province was selected because of its high rate of migration to Bangkok and low income. Within Roi-et, Atsamat District was selected because its population growth rate during the 1970s was similar to that for the entire province and because its level of accessibility and level of development were typical of much of the province. The six villageswhich range in population from just under 500 persons 'to just over I,OOo-are relatively poor, they are neither immediately adjacent to an urban area nor excessively remote, and they are broadly representative of many areas supplying migrants to Bangkok and other urban centers. 5 Approximately 50 percent of the households in each village were selected, for a total of 356 households. A simple random sample was obtained, with the proviso that selected households should have in residence at the time of the initial interview at least one person within the demographic groups from which the great majority 'of Thai rural-urban migrants are drawn: males 15-39 and never-married females 15-39. Two interviews were conducted with persons from each household: the household head (or representative) and one person in the above-mentioned demographic groups. A household head age 39 or younger was eligible to be interviewed with the second questionnaire; where more than one eligible person was available for the

568


second interview, random methods were used to select one respondent. Both questionnaires were administered in each household in both years. If the household member age 15-39 who was interviewed in the first wave was not present at the time of the second wave, mobility data for that person were obtained from another member of the household (usually the head or spouse of head). Most of the data discussed here derive from the second questionnaire. Villages in Thailand are organized administratively into tambon ("communes") consisting of about ten villages. The six villages were drawn from two communes (Nong Muun Than and Phon Muang) which appeared to be similar in terms of size, accessibility, and basic social and economic characteristics. Approximately equal numbers of households were selected from each commune. During the course of the study we discovered differences between the two communes that are quite relevant to our current interests. Villagers from Nong Muun Than have a higher level of contact with Northeast towns and with Bangkok than do villagers from Phon Muang. This is true not only in terms of the rate of movement to these destinations, but also in terms of the levels of social contacts, information, evaluations, and plans to move to these urban centers. The largest differences are in terms of rates of movement to Northeast towns. In the year prior to the initial survey, about 11 percent of villagers age 15-39 from Nong Muun Than went to a Northeast town, compared to about 6 percent for villagers from Phon Muang; the corresponding figures for movement to Bangkok are 26 percent for Nong Muun Than and 19 percent for Phon Muang. Although differences exist between the communes, Bangkok is clearly the more common urban destination for villagers from each commune. The study villages were the site of an experimental information program designed to alter patterns of rural out-migration. The design and impact of the information program are discussed in detail in Fuller et at. (1985). Briefly, "the information program consisted of efforts to stimulate the exchange of relevant information already possessed by individual villagers and to introduce credible employment information into the village. The latter activity involved posting job opportunities, sponsoring job searches, and creating a committee to oversee these activities" (Fuller et aI., 1985:604). "Comparison of rural-urban movement patterns before and after initiation of the information program indicates that, within one year, the program significantly increased migration from three experimental villages to urban places in the same region, as intended. This pattern of change was replicated in each experimental village, and was most pronounced among recent nonmigrants. There was little change in movement from the experimental villages to Bangkok, suggesting that the effect of the program may have been primarily to encourage movement to Northeast towns rather than to divert Bangkok-bound migrants. During the same time period, there was little change in the patterns' of out-movement from three control villages" (Fuller et aI., 1985:617). Nong Muun Than was the experimental ("treatment") commune, while Phon Muang was the control commune. Although we will argue that the experimental intervention had little or no effect on the results of the decision-making model, the initial differences between the communes and the existence of the program indicate the desirability of maintaining the distinction between the two communes. VARIABLES

Mobility Patterns of Villagers (MOVE] and MOVE2)

Because of the importance of circular mobility in Thailand, in this paper mobility is defined as including all absences from the home village of more than one week.


569

Detailed mobility histories were obtained for each respondent during the second interview. The record of mobility histories includes the approximate amount of time that each villager spent in Bangkok or in Northeast towns during each of four seasons, with five categories ranging from "none" of the time to "all" of the time. Continuous variables are formed by aggregating the seasonal data into annual records for each villager. MOVEI indicates the amount of time that each villager spent at the specified location during the year prior to the first survey; MOVE2 indicates the amount of time that each villager spent at the specified location during the year between the two surveys. The duration of most moves is more than thirty days. Urban Social Contacts (CONTACTS)

Respondents were asked what friends, relatives, or other personal contacts they had in any Thai town. Sixty-eight percent of respondents had at least one contact in Bangkok, compared with 29 percent who had at least one contact in a Northeast town. The total number of contacts in Bangkok was nearly three times that for Northeast towns, with mean values of 1.1 and 0.4 contacts per respondent, respectively. Information about Urban Areas (INFO)

Information about urban areas was assessed by asking villagers to report their perceptions of the wages paid for various urban occupations, the cost of urban housing, and the location of specific training opportunities in urban areas. For present purposes, what matters is whether or not a villager was willing and able to give a response, rather than the accuracy of the response. The series of questions was repeated for Bangkok versus Northeast towns. An index (INFO) was constructed by allotting each villager one point for each item about which the villager offered some concrete response. Summing this information for trade schools, wages for three occupations, and rental costs gave each villager a score ranging from zero to five, with a higher score indicating a higher level of information. Mean scores were 3.6 for Bangkok, and 3.2 for Northeast towns. Evaluations of Rural and Urban Areas (EVAL)

The villagers were asked to indicate their perception of the relative position of (a) the village, (b) Northeast towns, and (c) Bangkok on a dozen dimensions. The questions required the villagers to rank each of the three locations on each dimension. For example: "Do you think your standard of living would be better in the village, Northeast towns, or--Bangkok? Of the remaining two, which is better?"6 The villagers overwhelmingly feel that the village is superior to both Northeast towns and Bangkok in terms of such things as the standard of living, the friendliness of the people, working conditions, and as a place to raise children. In spite of the frequency with which the Northeast is plagued by either drought or flooding (both occurred during the short period of this study), most villagers perceive that life in the cities is more risky. A plurality of villagers rate the village as best in terms of health facilities and education for the children. Only in terms of learning a trade do the villagers perceive the village to be clearly inferior to urban areas. 7 Each villager provided a rank ordering of the three locations for each of the twelve dimensions. These twelve rank orderings were aggregated by summing for each location the rank that a villager assigned that location for each of the twelve dimensions. Since villagers ranked each location" 1", "2", or "3", the raw score for EVAL ranged from 12 to 36. EYAL was rescaled to range from 1 to 25, with a higher score indicating a more favorable evaluation of the destination in question. A

570


villager failing to rank a particular location for one or more dimensions does not have an INFO score for that location. Migration Plans (PLANS) A series of questions was asked to obtain from the villagers subjective estimates of the probability that they would move to Bangkok or a Northeast town. Each villager was placed into one of seven categories, ranging from "definitely would go to Bangkok within the next year" or "definitely would not go to Bangkok within the next five years." A similar series of questions sorted villagers into the same seven categories with respect to Northeast towns. Villagers were more likely to anticipate moves to Bangkok than to expect that they would go to a Northeast town in the foreseeable future. Over a quarter (29 percent) of villagers felt they definitely or probably would go to Bangkok within one year, while 39 percent indicated they definitely or probably would not move to Bangkok within the next five years. In contrast, only 15 percent of villagers indicated they definitely or probably would move to a Northeast town within the next year, while 48 percent indicated they definitely or probably would not move to a Northest town at any time within the next five years. Thus, the mobility plans and expectations of the villagers were consistent with their recent mobility behavior. RESULTS

The model discussed above is represented as a system of linear structural equations. For purposes of analysis, the theoretical model is represented as two models, one for Bangkok and one for Northeast urban destinations, each having two submodels-one for each commune. The four covariance matrices are reproduced in the Appendix. The two submodels for each destination are analyzed simultaneously using LISREL, which produces effect coefficients similar to regression coefficients. The system of equations has no reciprocal causation and no measurement model. Initially, unstandardized coefficients in each model were constrained to be equal for both communes. As a.result of initial analyses, certain coefficients were permitted to vary freely across groups and nonsignificant relationships were omitted from the submodels. After identifying which relationships are omitted from (a) all four submodels, (b) Bangkok submodels and (c) Northeast submodels, the overall models are evaluated and then the final submodels are discussed. Neither CONTACTS nor INFO or EVAL has a significant direct effect on MOVE2 for any group. These three variables affect subsequent moves indirectly through mobility plans. The direct links from each to subsequent mobility are omitted from the final submQdels. The variance in MOVE2 is permitted to vary freely across groups. For the Bangkok model, the link from CONTACTS to EVAL is omitted for both groups; the link from MOVEI to INFO is also omitted for the treatment submodel. In addition, the variance in INFO is permitted to vary freely across groups. For the Northeast model, the link from MOVEI to INFO is omitted for both groups; the link from MOVEI to CONTACTS is omitted for the control submodel; and the link from MOVEI to PLANS is omitted for the treatment submodel. Additionally the magnitude of the link from MOVE 1 to MOVE2 and the link from EVAL to PLANS is omitted for the control commune. As can be seen from the above, the final structural models for both communes are quite similar. Most of the differences between the two communes reflect differences in the potency of MOVEI as a predictor of endogenous variables.


571

FIGURE I. MOBILITY TO BANGKOK, TREATMENT COMMUNE SUBMODEL: STANDARDIZED COEFFICIENTS.

1. In the Northeast submodels, MOVEI affects CONTACTS for the treatment

commune but not for the control commune. 2. In the Northeast submodels, MOVEI affects PLANS for the control commune but not for the treatment commune. 3. In the Northeast submodels, while MOVEI has a significant effect on MOVE2 for both communes, the effect is stronger for the treatment commune than for the control commune. 4. MOVEI affects INFO only for the control submodel of movement to Bangkok and not for the other three submodels. One remaining distinction in the structural models for different communes does not involve MOVE 1: in the Northeast submodels,. EVAL has a significant effect on PLANS only for the treatment commune. 8 The overall chi-square for the Bangkok model is 22.70 with 22 degrees offreedom and has a probability of .419; the overall chi-square for the Northeast model is 11.19 with 22 degrees offreedom, and has a probability of .972. Unlike conventional uses of chi-square, the smaller the chi-square and the larger the probability, the better the fit. These results mean that the differences between the observed covariance matrices and the matrices estimated from our model are not statistically significant. The goodness-of-fit indices meet a minimum standard of .90; the value for the Bangkok model is .978, for Northeast towns .991. No theoretically meaningful modification significantly improves the model; among the four submodels, the

572


SI

Sp

FIGURE 2: MOBILITY TO BANGKOK, CONTROL COMMUNE SUBMODEL: STANDARDIZED COEFFICIENTS.

maximum modification index is 5.183. The maximum normalized residual is 1.626. The two models and four submodels appear to provide an excellent fit with the data. Standardized solutions for the Bangkok model are presented in Figures 1 and 2 (Figure 1 for the treatment submodel and Figure 2 for the control submodel). Standardized solutions for the Northeast town model are presented in Figures 3 and 4 (Figure 3 for the treatment submodel and Figure 4 for the control submodel). Maximum likelihood estimates and (-values from LISREL 6.1 are presented in Table 1. 9 As hypothesized, all significant effects are positive. In three of the four submodels, move histories have a significant effect on the number of urban social contacts. The amount of explained variance, however, is minimal, ranging from 2.1 percent to 3.8 percent. The number of urban social contacts, in turn, affects the amount of information that villagers have about urban places. Contrary to our hypothesis, however, history of moves has an effect on information for only one of the four submodels (movement to Bangkok from the control commune). The amount of explained variance for information varies from 3.7 percent to 7.3 percent. For all four submodels, villagers who have spent more time at an urban place and have more information about that place are likely to evaluate it more favorably. Interestingly, urban social contacts affect evaluations of Northeast towns but not the evaluation of Bangkok. This difference may be due to greater familiarity with Bangkok; a multiplicity of sources of information are available for Bangkok, so personal contacts may exert a relatively weak influence on the formation of


573

~I

C p

FIGURE 3. MOBILITY TO NORTHEAST THAI TOWNS, TREATMENT COMMUNE SUBMODEL STANDARDIZED COEFFICIENTS.

evaluations. The amount of explained variance for evaluations is somewhat higher than for urban social contacts and information; it ranges from 7.5 percent to 12.8 percent. In all four submodels, plans for movement are significantly affected by both urban social contacts and information. In three of the four submodels, evaluation and history of moves significantly affect plans; in those three submodels, history of moves has the strongest effect on plans. In two ofthese three submodels, evaluation is the second strongest variable; in the submodel where history of moves is not a significant factor (movement to Bangkok from the treatment commune), evaluation emerges as the variable with the strongest effect on plans. The amount of explained variance for plans varies from 12.1 percent to 24.6 percent, and exceeds 20 percent for three of the four submodels. There is substantial continuity in mobility. Although five variables were hypothesized to affect MOVE2, only MOVEI and PLANS have significant direct effects on MOVE2. As noted above, the other three independent variables exert their influence indirectly through PLANS. Nonetheless, the two variables have substantial predictive power for MOVE2. The relatively high accuracy of plans as a predictor of subsequent moves can be illustrated as follows. Among the villagers who indicated they "definitely" or "probably" would move to Bangkok within one year, 52 percent actually did move to Bangkok during the next year; among those who stated that they definitely or probably would not move to Bangkok within five years, only 9 percent went to Bangkok during the next year. The corresponding percentages for

574


SI

Sp

FIGURE 4. MOBILITY TO NORTHEAST THAI TOWNS, CONTROL COMMUNE SUBMODEL: STANDARDIZED. COEFFICIENTS.

Northeast towns are 38 and 10, respectively. In multivariate analysis, however, the effect of MOVEl on MOVE2 overshadows that of PLANS in each submodel. The amount of explained variance for MOVE2 ranges from 21.1 percent to 67.6 percent, and exceeds 40 percent for three of the four submodels. 1O DISCUSSION

We have argued that information is a key intervening variable in the migration decision-making process, not least because it, more than other relevant variables, is subject to intervention. The present study supports this argument. While information has no direct effect on subsequent moves, it does affect evaluations of both alternative destinations and mobility plans. Clearly, not all the information villagers receive about urban centers will be positive, and information will therefore not necessarily have a positive effect on evaluations of urban areas and plans to move to them. In the study area, however, villagers who were better informed about Bangkok or Northeast towns in terms of the relatively factual dimensions tapped by our questions were more likely to have favorable evaluations of these places and be planning to move to them. Furthermore, as noted above, research presented elsewhere has shown that it is possible to introduce objective information which will encourage movement to a specified location. The remainder of the discussion addresses a single methodological issue: whether or not the experimental information program described above exerted any substantial impact on the relationships among the variables. One might suppose, since the information program affected the overall level of mobility from the experimental

Rural·Urban Mobility in Thailand

575

Table I.-Maximum likelihood estimation for each mobility decision-making submodel: Unstandardized solution a Independent variable CONTACTS

Dependent variable

.218 (2.036)

.445 (3.539) .730 (2.557) .379 (2.221)

.502 (2.716) .194 (1.967)

.502 {2.716} .194 (1.967)

.429 (3.471) .219 (2.944)

.429 (3.471) .219 (2.944)

PLANS

.114 (3.989)

.114 (3.989)

.227 (5.233)

MOVES

.023 (2.025)

.023 (2.025)

.031 {2.719}

CONTACTS

.197 (2.628)

.197 {2.628} .255 (1.978) 1.203 (4.353) .979 (6.636) .688 (20.524)

.216 {2.598}

.237 (4.069)

PLANS

.218 (2.036)

EVAL PLANS

EVAL PLANS MOVE1

Northeast town Treatment Control

.445 (3.539) .730 (2.557) .379 (2.221)

.237 (4.069)

INFO EVAL

INFO

Bangkok Treatment Control

INFO EVAL PLANS MOVE2

1. 203 (4.353) .979 (6.636) .688 (20.524)

1.313 (3.968) .714 (10.421)

.031 (2.719)

1.313 (3.968) 1.122 (3.541) .310 (5.200)

at-statistic in parentheses.

commune to Northeast towns, that the program may have also affected relationships among the variables in the model. We will argue, however, that the experimental program did not, in fact, affect the decision-making process itself. Thus, the outcome was affected by altering the inputs, not by altering the process. The issue is simplified somewhat by recognizing that MOVEl, CONTACTS, INFO, EVAL, and PLANS all refer to events that occurred prior to implementation of the information program. Thus, the program cannot have affected relationships among these five variables, and we can focus our attention on the equation in which MOVE2 is the dependent variable. While there are a number of differences among the four submodels, there is great consistency in terms of the variables which have a significant effect on MOVE2. In all four submodels, only MOVEI and PLANS have significant effects on MOVE2, and in each submodel MOVE} has a stronger effect than does PLANS. The most striking difference between the two Northeast submodels is that the effect of MOVE} on MOVE2, which is a measure of stability of mobility, is substantially greater in the treatment commune than in the control commune (the unstandardized coefficients are .714 and .310). If stability were lower in the treatment commune, it would be easy to argue that the stability was reduced because of the information program. However, since stability is lower in the control


576

Appendix table.-Covariance matrices for the four submodels CONTACTS

INFO

EVAL

PLANS

MOVE2

MOVEI

Mobility to Bangkok from the treatment commune (N=176) CONTACTS INFO EVAL PLAN MOVE2 MOVEI

1. 395 0.259 0.930 0.869 0.185 0.202

1. 288 0.533 0.426 0.037 0.074

17.453 3.338 0.889 0.837

6.234 0.794 0.915

0.570 0.528

0.708

Mobility to Bangkok from the control commune (N=152) CONTACTS INFO VAL PLANS MOVE2 MOVEI

1. 212 0.417 0.346 0.205 0.013 0.064

1. 904 1. 422 0.827 0.036 0.193

20.310 3.383 0.397 0.992

5.782 0.625 0.767

0.593 0.459

0.694

Mobility to Northeast towns from treatment commune (N=173) CONTP.CTS INFO EVAL PLANS MOVE2 MOVEI

0.444 0.174 0.700 0.421 0.070 0.080

2.309 1.606 0.912 0.074 0.077

13.836 3.762 0.583 0.484

5.346 0.337 0.177

0.502 0.271

0.371

Mobility to Northeast towns from the control commune (N=150) CONTACTS INFO EVAL PLANS MOVE2 MOVEI

0.456 0.230 0.179 0.136 -0.006 -0.020

2.429 0.756 0.640 0.005 -0.006

11.179 1.133 0.081 0.440

4.270 0.201 0.289

0.167 0.091

0.265

commune, another explanation must be sought. Since the overall level of mobility to Northeast towns is lower from the control commune than from the treatment commune (6 percent versus 11 percent), a plausible explanation is simply that more routinized mobility patterns have developed in the treatment commune. Villagers in the treatment commune may have a pattern of going to town year after year (e.g. during agricultural slack seasons), and, while some control villagers do go to town each year, fewer people from the control village do so consistently year after year. The existence of repeat movers would be consistent with other analyses of the same data (Lightfoot et al., 1983). SUMMARY

This paper has analyzed a theoretical model of mobility decision-making. The model relies entirely on individual-level factors rather than macro-level factors as determinants of migration decision-making. The individual-level variables included in the model are: recent mobility history, urban social contacts, information about urban areas, evaluations of different locations, migration plans, and actual movements in the period subsequent to an initial interview. The empirical results indicate

577


that with some exceptions there are relatively strong links of the type suggested in the model among these variables. The model was evaluated separately for two groups of villages for movement to Bangkok and for movement to Northeast Thai towns. Thus, four submodels were estimated, providing an opportunity to observe how robust the model is with respect to varying destinations and origins. Although certain differences are found among the four submodels, the overwhelming feature is their similarity. Where differences do exist, they generally reflect differences in the effectiveness of prior mobility as a predictor of other variables in the process. Clearly, a villager's previous history of movement is a key factor affecting subsequent movement and the entire decision-making process. The primary effect of having friends and relatives in a particular urban center is to increase the amount of information a villager has about that urban center. Information has a significant effect on evaluations and plans. Except in one submodel, evaluations have a significant effect on plans; and the existence of plans-which to some extent represent a culmination of social contacts, information, and evaluations-is the only factor other than previous mobility which has a significant direct effect on subsequent movement. Thai policy makers are searching for ways to stimulate the growth of regional urban growth centers and reduce the growth of Bangkok. From the standpoint of intervention, a key variable in this process would appear to be information. Not only is information level related to evaluations of an urban area and mobility plans, but, compared to other variables in the model, it appears to be relatively amenable to modification by inputs deriving from a source external to the village itself. It appears difficult to modify evaluations or migration plans directly, though both could be indirectly influenced by informational inputs. Movement history would be difficult, if not impossible, to manipulate; while villagers could be sponsored for short trips to town, this is not likely to produce much long-range effect. The possibility of expanding the network of urban social contacts-to build "social bridges" connecting the villages to Northeast towns-has a certain theoretical appeal, but in practical terms it would appear to be more feasible to introduce appropriate information into the village to alter subsequent mobility patterns. NOTES I Different researchers have focused on different aspects of mobility, and definitions proliferate. Goldstein and Goldstein (1981 :50-66) provide an extensive review of variations in definitions of migration and circulation. For current purposes, we can say simply that circular rural-urban mobility refers to repeated movements between arural origin and an urban destination. Our specific measure of mobility is defined below. 2 It is worth noting that about 5 percent of our respondents in the survey of people age 15-39 said they knew someone in a city who could actually provide them with ajob, as opposed to merely helping in the search for a job. 3 Using data for rural-urban migrants in towns of Northeast Thailand, Fuller argues that "there is widespread satisfaction with the quality of life found at the destination" compared to the village (Fuller, 1981: 101) and furthermore "satisfaction with urban life affects the preference of migrants for remaining in town versus returning to their home villages... " (Fuller, 1980:729). While the cited study relies on data from urban destinations, the current study is based on data obtained at the rural origins. 4 Northeast Thailand is one of four officially designated regions in Thailand. In this study, "Northeast towns" refers to any of the fifteen provincial capitals in Northeast Thailand. 5 More detail about the study area and its location in the national context can be found in Fuller et aI.,

(1983).

6 The full list of dimensions of comparisons are: standard of living, riskiness of life, opportunities to be rich, learning a trade, working conditions, friendliness of people, raising children, education of children, health facilities, respect of villagers, comfort and convenience, and enjoyment.

578


7 Recall that the sample is restricted to young adults aged 15-39, the age group from which most of the out-migrants are recruited. Even this relatively young group finds much to praise in the village. S When the link from EVAL to PLANS is retained for the control commune and the magnitude of the relationship is permitted to vary freely between the two communes, the unstandardized coefficient for the control commune is .039, with t = 0.821. 9 Standardized coefficients used in this paper are calculated as the unstandardized coefficient multiplied by the ratio of the estimated standard deviation of the independent variable to the estimated standard deviation of the dependent variable, following customary procedure for regression analysis. In analyzing multiple groups, LISREL 6.1 employs a different procedure, using a weighted average of variance estimates from each group, so that the LISREL 6.1 standardized coefficients for two groups are equal whenever the corresponding unstandardized estimates are equal. A comparison of the two procedures and a justification of the procedure used here is found in Acock and Fuller (1985). 10 For the Bangkok submodels, including an effect of EVAL on MOVE2 significantly reduces chisquare if different effects are permitted for the two communes (chi-square is reduced by 6.71, with two degrees of freedom). In this case, the effect of EVAL on MOVE2 in the control commune is negative (-.018), while the effect in the treatment commune is positive (0.14). Although the apparent reversal of effect is intriguing, this relationship is not included in Figures I or 2 or in Table I for two reasons: (a) the effect in each submodel is weak and nonsignificant and (b) from a substantive point of view, it is difficult to understand why villagers who have a more favorable evaluation of Bangkok would, other things being equal, be less likely to move to Bangkok.

ACKNOWLEDGMENTS

An earlier version of this paper was presented at the Section on "Migration Strategies, Migration Intentions, and Migration Decisions," Annual Meeting of the Population Association of America, Minneapolis, Minnesota, May 3-5, 1984. The research reported here was funded by the Rockefeller and Ford Foundations' Research Program on Population and Development Policy (RF77065; Allocation 164). We are grateful to Alan C. Acock, Lee M. Wolfle, and anonymous referees for their comments. REFERENCES Acock, A. C. and T. D. Fuller. 1985. Standardized solutions using LISREL on multiple populations. Sociological Methods and Research 13:551-557. Arnold; F. and S. Cochrane. 1980. Economic Motivation Versus City Lights: Testing Hypotheses about Inter-Changwat Migration in Thailand. World Bank Staff Working Paper No. 416. Washington, D.C.: World Bank. Bach, R. L. and J. Smith. 1977. Community satisfaction, expectations of moving, and migration. Demography 14:147-167. Byerlee, D. 1974. Rural-urban migration in Africa: theory, policy and research implications. International Migration Review 8:543-566. Chamratrithirong, A. 1976. Fertility, Nuptiality, and Migration in Thailand: Multiphasic Response Theory. Ann Arbor: University Microfilms. deYoung, J. E. 1963. Village Life in Modern Thailand. Berkeley: University of California Press. Findley, S. 1971~ Planning for Internal Migration: A Review of Issues and Policies in Developing Countries. Washington, D.C.: U.S. Bureau of the Census. - - - . 1980. Migration as a Household Innovation. Paper Presented at the Annual Meeting of the Population Association of America, Denver, Colorado. Fuller, T. D. 1980. Satisfaction with urban life: the judgment of villagers transplanted to small urban centers in Thailand. Rural Sociology. 45:723-730. - - - . 1981. Migrant evaluations of the quality of urban life in Northeast Thailand. The Journal of Developing Areas 16:87-104. Fuller, T. D., P. Kamnuansilpa, P. Lightfoot, and S. Rathanamongkolmas. 1983. Migration and Development in Modern Thailand. Bangkok: Social Science Association of Thailand. Fuller, T. D., P. Lightfoot, and P. Kamnuansilpa. 1985. Toward migration management: a field experiment in Thailand. Economic Development and Cultural Change 34:601-621. Goldstein, S. 1978. Circulation in the Context of Total Mobility in Southeast Asia. Papers of the EastWest Population Institute, No. 83. Honolulu: East-West Center. Goldstein, S. and A. Goldstein. 1981. Surveys of Migration in Developing Countries: A Methodological Review. Papers of the East-West Population Institute, No. 71. Honolulu: East-West Center. Goodman, J. L. 1981. Information, uncertainty, and the microeconomic model of migration decision


579

making. Pp. 130-148 in G. F. Dejong and R. W. Gardner (eds.), Migration Decision Making: Multidisciplinary Approaches to Microlevel Studies in Developed and Developing Countries. New York: Pergamon. Harbison, S. F. 1981. Family structure and family strategy in migration decision making. Pp. 225-251 in G. F. Dejong and R. W. Gardner (eds.), Migration Decision Making: Multidisciplinary Approaches to Microlevel Studies in Developed and Developing Countries. New York: Pergamon. Hugo, G. J. 1981. Village-community ties, village norms, and ethnic and social networks: a review of evidence from the Third World. Pp. 186-224 in Dejong and Gardner (eds.), Migration Decision Making: Multidisciplinary Approaches to Microlevel Studies in Developed and Developing Countries: New York: Pergamon. - - - . 1982. Circular migration in Indonesia. Population and Development Review 8:59-84. Lauro, D. J. 1979. The Demography of a Thai Village. Unpublished PhD dissertation, Australian National University. Lightfoot, P., T. Fuller, and P. Kamnuansilpa. 1983. Circulation and Interpersonal Networks Linking Rural and Urban Areas: The Case of Roi-et, Northeast Thailand. Papers of the East-West Population Institute, No. 84. Honolulu: East-West Center. McGee, T. G. 1968. Revolutionary change and the Third World city: a theory of urban involution. Civilisations. 18. Reprinted in T. G. McGee. The Urbanization Process in the Third World. London: Bell, 1971. Prasartkul, P. 1977. Patterns and Determinants of Internal Migration in Thailand. Unpublished PhD dissertation. Ithaca, NY: Cornell University. Rabibhadana, A. 1975. Bangkok Slum: Aspects of Social Organization. Unpublished Ph.D. dissertation, Cornell University. Rhoda, R. 1983. Rural development and urban migration: can we keep them down on the farm? International Migration Review 17:34-64. Ritchey, P. N. 1976. Explanations of Migration. Pp. 363-404 in A. Inkeles, J. Coleman and N. Smelser (eds.), Annual Review of Sociology Vol. 2. Palo Alto, California: Annual Reviews. Roberts, K. 1981. Agrarian structure and labor mobility in rural Mexico. Population and Development Review 8:299-322. Schwarzweller, H. K., J. S. Brown and J. J. Mangalam. 1971. Mountain Families in Transition: A Case Study of Appalachian Migration. University Park, Pennsylvania: Pennsylvania State University Press. Shaw, R. P. 1975. Migration Theory and Fact: A Review and Bibliography of Current Literature. Bibliography Series, No.5, Philadelphia, PA: Regional Science Research Institute. Singhanetra-Renard, A. 1981. Mobility in North Thailand: A View from Within, in G. W. Jones and H. V. Richter (eds.), Population Mobility and Development: Southeast Asia and the Pacific. Development Studies Centre Monograph No. 27. Canberra: Australia National University. Speare, A., Jr. 1971. A cost-benefit model of rural to urban migration in Taiwan. Population Studies 25:117-130. - - - . 1974. Residential satisfaction as an intervening variable in residential mobility. Demography 11:173-188. Textor, R. B. 1961. From Peasant to Pedicab Driver. Southeast Asian Studies, Cultural Report Series, No.9. New Haven, Conn.: Yale University. Todaro, M. P. 1969. A model of labor migration and urban unemployment in less developed countries. The American Economic Review 59:138-148. - - - . 1976. Rural-urban migration, unemployment and job probabilities: recent theoretical and empirical research. Pp. 307-385 in A. J. Coale (ed.), Economic Factors in Population Growth. London: Macmillan. Vichit-Vadakan, J. 1983. Small towns and regional urban centers: reflections on diverting Bangkok-bound migration. Thai Journal of Development Administration 23:79-99. Wolpert, J. 1965. Behavioral aspects of the decision to migrate. Papers of the Regional Science Association 15: 159-169.