Printer Friendly

Migration and its determinants: a study of two communities in Colombia.

This study attempts to explain the high incidence of migration from a region in southwestern Colombia to other countries, particularly the USA. The findings can shed light on the migratory patterns of people from Colombia as a whole. This study also tests whether the theory of relative deprivation is a fruitful way to model migration from Colombia. In the spirit of the new economics of migration, this study assumes that individuals migrate not only to better themselves but also to increase their families' relative economic standing in the community. Migration decisions are made jointly by family members within households as they seek to take advantage of diverse labor markets, no matter how distant, in order to maximize income given the labor resources at their disposal.

An advantage of the relative deprivation theory is that migration decisions are directly affected by socioeconomic conditions in the community. Massey (1990, p. 5) convincingly argues migration is far more dynamic and self-reinforcing than has been generally realized. That migration decisions are not only made within the family unit but that they are influenced by the socioeconomic realities of their community. These realities are, in turn, affected by evolving political, social and economic structures at the national and international levels. These relationships are then all connected to one another over time. The cross-sectional nature of the data and the fact that it was collected at one point in time prevents one from making the study as dynamic as Massey suggests. The resulting snapshot in time does however offer insights into the decision to migrate when placed against the realities of the Colombian experience.

The Survey-Generated Data

Any true test of the theory of relative deprivation requires that at least two communities of different socioeconomic status be studied. To this end, a survey questionnaire was used on randomly selected families in two neighborhoods in the city of Palmira, located in El Valle del Cauca (a departamento in southwestern Colombia). Questions were asked about each family member of legal working age (13 or older) and about the family as a whole.

The neighborhoods chosen, though next to each other, belong to markedly different socioeconomic strata. From El Prado, a community of approximately 1,286 households, a random sample of 300 families (about 25%) was selected. El Prado is an urban, blue-collar neighborhood and most of its people are either clerical employees or have small retail businesses, more than likely, in the informal sector. The other community studied is Santa Ana, where two hundred families in this better-to-do community were randomly selected, giving a family close to a 50% probability of being in the sample. Santa Ana is also urban but its residents, by most measures, would fall squarely in Colombia's middle class. As a whole, a total of 500 households were sampled and information on 1646 individuals was collected.

Table 1 shows how these two neighborhoods differ. In terms of income, education, investment and wealth, the residents of Santa Ana rank higher than residents of El Prado. The percentage of home ownership is roughly the same in both neighborhoods. Business ownership is higher for families in El Prado but their businesses are not as profitable. The higher income and wealth of families in Santa Ana allow them to hire more maids. Fifty one percent of them had maid services as opposed to only 10% in El Prado. The added expenditure on domestic services by Santa Ana residents is used later to adjust for an imputed measure of family income.

There are also differences in some key social measures. Families in El Prado are larger on average, have more children and more dependents. Given these generalities, it is safe to assume that the neighborhoods chosen belong to different socioeconomic strata and thus are appropriate selections for the purposes of this study.

The Importance of Measuring Family Income Appropriately

To measure family income in Latin America contributions from all members of the family have to be considered. In doing so, another important consideration is the prevalence of do-it-yourself work in developing nations. Measures of this type of income are often not available. Only in survey data is it possible to get an idea of its magnitude. Do-it-yourself work, such as home repairs and sewing, may represent a larger percentage of income for low income families because of lower opportunity costs afforded to them by their lower market wages or their higher experience with unemployment. Table 2 shows how the two communities differ with respect to do-it-yourself work.

As expected, a greater percentage of families engage in do-it-yourself work in the community of El Prado than in the wealthier community of Santa Ana.

My family income measure also includes the value of rent for those that did not pay rent but owned their home. (1) It includes income from renting to others (either for housing or other rental property), income from business profits, the imputed income from do-it-yourself work, and the contributions to income from all family members including next of kin.

Since relative differences in income are more important in this study than absolute differences, monthly payments for maid services were subtracted from the income of families that had such costs. This of course raised relative income for families that did not pay for domestic services.

Adjustments to the Data

Migrant remittances could not be included as they would bias the family income measure. Measures of relative deprivation at the time migration decisions were made obviously did not include such remittances. To correct for this self-selection bias, contributions to family income that would have taken place in the absence of migration had to be estimated. Hickman's selection bias estimating model was used to estimate these unknown contributions. (2)

The Theoretical Basis for the Migration Decision Model

The model proposed in this paper is asymmetric--pull and push reasons for migrating are not given equal weight. In fact, no explicit use of generalized pull factors is considered. Instead, the model assumes that people that migrate do so because they expect their life will improve in some way. The reasons for why they think so, however, can not be generalized across all families, but are particular to the profile of the family deciding to migrate.

The attractiveness of the receiving region may be due to differences in potential earnings, more employment opportunities, better schools, specialized health care, proximity to friends and relatives, greater safety, and anything else that improves quality of life.

The advantage push factors have is that they tie directly to the person and the family making the migration decision. A person with knowledge of a foreign language may be more inclined to go abroad. Families with relatives or friends abroad experience push factors that other families may not have. An already migrant family experiences push factors due to perceived differences between place of origin and place of residence. The point is that a person's choice of destination identifies which of the personal and familial characteristics they possess are the ones that matter most.

Formally, the probability that individual T migrates to some destination 'j' is a function of three variables as follows:

P([Y.sub.ij] = 1) = f(EPIG, ECFRD, [PSTD.sub.j]) (1)

Where:
EPIG                 expected personal income gain
ECFRD                expected change in family's relative deprivation
[PSTD.sub.J]         probability of successful travel to destination j.
EPIG                 is in turn assumed to be a function of personal
                     attributes that determine a person's productivity
                     in the marketplace such as the person's age,
                     education, etc. In other words:
EPIG                 [f.sub.1]](the person's age, gender, education,
                     etc.)
We now define ECFRD  where measures the sensitivity of the family's
as: ECFRD            relative  deprivation (RD) to expected changes in
                     its income. And finally [PSTD.sub.J] is defined
                     as:
[PSTD.sub.J]         [f.sub.2] (social/economic assets that make travel
                     to destination 'j' a viable alternative)


A latent variable is now defined as:

[Y.sup.*.sub.j] = XB + [[epsilon].sub.j] (2)

where X is an [N.sub.j] x K matrix and B a K x 1 vector of coefficients for the K variables that determine the main component variables EPIG, [ECFRD.sub.j] and [PSTD.sub.j]. [Y.sup.*.sub.j] is latent because it is made up of variables with very different interpretations and different levels of measurement so it has no discernable meaning. Its importance lies in that it takes values distinctly different for individuals that have different propensities to migrate to destination 'j.'

This latent variable approach, of course, assumes that there exist critical values of [Y.sup.*.sub.j] beyond which or below which probabilities of migration change significantly. Changes effected on [Y.sup.*.sub.j] through marginal changes in an explanatory variable can be sufficient to make migration a serious consideration when previously it was not contemplated. The data combined with the decisions made by the individuals surveyed will identify the values for [Y.sup.*.sub.j] around which migration to destination 'j' is more or less probable. (3)

Relative Deprivation and Its Measurement

The idea of relative deprivation is a fairly old one, appearing at least as early as in the work of Stouffer et al. (1949). It has been applied in various fields of study concerning human behavior. Its application in economics is owed primarily to the contributions by Runciman (1966), Yitzhaki (1982), Stark (1984, 1991), Stark and Yitzhaki (1988). The economic implications of relative deprivation, when applied to the study of migration, were thoroughly explored by Stark Oded in his book "The Migration of Labor" (1991). In Stark's judgment, families should feel relatively more deprived if any one or both of the two following possibilities present themselves:

1. The average income of the families earning more income than the family increases

2. The number of families earning more income than the family increases.

In that case, relative deprivation experienced by family 'i,' RD([y.sub.i]), should be measured as

RD([y.sub.i]) = E(y/y > [y.sub.i])(1 - F([y.sub.i])]) (3)

where [y.sub.i] represents family i's income and (1-F([y.sub.i])) the percentage of families in their community with incomes higher than their own earnings. Equation 3 was computed for every one of the families in the two communities sampled--a family's deprivation being of course relative to the families in their own community.

On the Determinants of Expected Personal Income Gain (EPIG)

While searching for variables significant in determining a person's EPIG, the individual's gender, age, age-squared, and the person's degree of knowledge of the English language always showed to be significant determinants of migration. Neither their highest level of education nor their total years of schooling proved to be significant.

As for speaking English, the individuals sampled were asked whether they spoke "none at all", "a little", "very well", or "fluently". Mindful of the possibility that those who had learned the language to some degree could have learned it post a migration experience, care was taken to ensure the respondent (usually the head of the household) was asked about the migrating person's knowledge of English prior to migrating.

Speaking English, even a little, proved to be significant and positively correlated with the probability of migrating. The variable "spkengl" more than likely is a proxy for the individual's sense of awareness of the world beyond his immediate experience and in some way represents his/her desire to learn and explore more about that world.

On the Determinants of the Expected Change in Family's Relative Deprivation (ECFRD)

The new economics of migration thinks of the family as a second level influence on an individual's decision to migrate. Some theorists think that sending family members to different destinations is simply the family's attempt to minimize the risk inherent in living in a situation of uncertain streams of income. This precarious existence is not unlike what many families experience living in developing nations. If various streams of income in the family are more uncorrelated, then their economic survival becomes more certain. Thus, migration offers families a way to stabilize income in the absence of access to capital markets, as is usually the case for most families in developing nations.

Relative deprivation theory relies also on the tacit agreement that migrating members will remit income to the family once successful migration takes place. Statistical evidence on the volume of remittances leads us to conclude, in the Latin American case, the closeness of the family is congruent with the tacit agreement that migrants make with family members left behind.

On the Dynamics Between Family Income and Relative Deprivation

In this study, the simultaneous use of the variables relative deprivation (rd) and annual family income (finc) either led to insignificant coefficients or to contradictory results. Stark and Taylor (1991, p. 1175) successfully used these variables separately to account for the absolute levels of income needed to finance migration (finc) and to account for the incentive to migrate that relative deprivation represents (rd).

When used separately in this study, the coefficients for both relative deprivation and family income were negative. These results are contradictory in that one argues that higher levels of relative deprivation reduce the probability of migrating while the other argues that less income increases the probability of migrating. I felt, however, that a variable created by multiplying relative deprivation and family income (rdxfinc) could potentially capture the dynamics of its two component variables as family income changes.

Figure 1 shows that "rdxfinc" rises at first in both neighborhoods. This happens because, at low levels of income, increases in family income represent large percentage increases. Increases in income at this stage are however insufficient to reduce relative deprivation in any significant way since relative deprivation is at high levels to begin with. The outcome of such interplay of course is that the product of the two variables rises rather than falls.

The "rdxfinc" curve becomes flat at the point where the percentage increase in income equals the percentage fall in relative deprivation. Increases in income beyond these values accelerate the reduction in relative deprivation in proportions larger than the increases in income taking place thus causing "rdxfinc" to fall. The decline in "rdxfinc" will eventually slow down simply because continued increases in income lose their ability to further reduce relative deprivation as families reach the upper levels of income-earners in their respective communities. The slope of the "rdxfinc" curve will again approach zero as percentage increases in income fall and get closer to the falling percentage decline in relative deprivation. The speed at which the "rdxfinc" curve rises and falls is, of course, determined by the actual distribution of income in the community under study.

[FIGURE 1 OMITTED]

If this interpretation is correct, migrants should map on the falling portion of the "rdxfinc" curve, especially where the curve shows its steepest decline, as these points represent families with the greatest gains in reducing relative deprivation if they can successfully migrate. Few if any migrants are expected to map on the rising portion or on the tail end of the "rdxfinc" curve. In these regions families cannot reduce relative deprivation (rising portion) or the expected benefits of migration are very small (flat tail end).

Empirically, this paper proposes that changes in the variable "rdxfinc" represent an appropriate way to model the marginal impact that migration (and thus expected increases in family income) has on relative deprivation. Changes in the product of family income and relative deprivation are then an appropriate measure of ECFRD or expected change in the family's level of relative deprivation as a result of migration. If correct, the coefficient fur the variable "rdxfinc" should be negative, indicating that as its value falls (rises), the probability of migrating rises (falls). This negative coefficient arises from the expected change that family income (through remittances) has on its ability to reduce relative deprivation.

On the Probability. of Successful Travel to a Chosen Destination (PSTD)

Stark (1991, p. 160), Stark & Taylor (1991, p. 1176) found that relative deprivation was significant in explaining migration from Mexico to the USA. This study tests the idea that relative deprivation measures a willingness to migrate, but that such willingness is not enough without having the capacity to migrate. This is much like demand in which the willingness to buy does not translate into effective demand without the ability to buy.

The ability to migrate, of course, is not purely measured in monetary terms. It most often requires such extensive social assets as family and friends connections at the proposed migration destination. Beyond reducing risk and uncertainty, social and economic assets are often important prerequisites for individuals to obtain visas for travel abroad.

Whether international migration is intended to be legal or illegal, most Colombian migrants attempt to get a visa first. This would usually not be the case if the person could just simply walk across a country's border illegally. But even then, financial as well as social assets are needed for successful migration to take place.

Most illegal immigrants in the USA from Colombia entered the USA with a tourist visa. They became illegal simply by staying beyond the time allotted by their entry permits. For Latin American citizens this has made obtaining a tourist visa, especially to the USA, an increasingly difficult thing to get.

Given that family income contributes towards making migration viable, higher income families should, all else equal, have higher probabilities of migration. This means that the variable "rdxfinc" is important not only because its negative slope conveys the potential marginal reduction in relative deprivation but also because its decline in value signals a rise in income--and according to this study, it means more resources for making any proposed migration a more viable consideration, all else being equal. (4)

In immigration matters, US policy continues to give priority to family members of legal residents in the USA. Having such social assets thus increases the probability of successful migration (PSTD) and this is true regardless of whether they are migrating within Colombia, to the USA or to some other country.

Testing for the Migrating Networks Effect

The variables "state" and "famex" were constructed to test for the existence of migrating networks domestically and internationally. Both of these variables showed to be significant in explaining a person's decision to migrate. "State" took the value 0 if the individual was born in the state where the study was carried out (Valle del Cauca) or the value 1 if born elsewhere in Colombia. The intent in using this variable was to test whether or not members of families with roots elsewhere in Colombia were more likely to migrate than members of families with only local roots. (5) The binary variable "famex" took on the value 1 if the family had members already living abroad or the value 0 otherwise.

Modeling the Four-Alternative Dependent Variable "Choice"

The dependent variable "Choice" offered the individuals sampled four possible alternatives to choose from: to not migrate (S), to migrate within Colombia (C), to migrate to the USA (U), and to migrate to a country other than the USA (O). Table 3 shows the results of the best-fitting model. (6)

What the Numbers Say

Interpretation is made somewhat difficult by the nonlinear relationship between the probability of migrating and the independent variables in the estimating equation. The marginal impact of an independent variable on the probability to migrate is a function of the values assumed by all other independent variables in the equation. An alternative to this problem is to compute the factor changes in the odds-ratio of migrating to one destination as opposed to another location. The advantage of computing odds ratios is they are independent of the remaining explanatory variables. An exception was the variable "age" because changes in the person's age would naturally change age-squared (agesq). An expression was derived to deal with this case. (7)

Table 4 shows the factor changes in the odds ratios as well as the marginal effects on migration probabilities when assuming the remaining independent variables are held at their sample means.

Summary of the Results

When compared to females, males have a threefold higher odds of migrating within Colombia, a 2.04 higher odds of migrating to the USA and a 1.88 higher odds of migrating to a country other than the USA. These results are at odds with results from other studies. Fields (1982, p. 553) states that "A general characterization of migration in Latin America is that females have higher average and marginal propensities to migrate than do males." However, in another study Fields (1979, p. 251) concludes, "Evidently sex-selectivity is not as important a feature of Colombian migration at least at the departamento level, as it appears elsewhere in Latin America." Fields' work on migration in Colombia was only on region-to-region migration and international migration was not considered.

The results on the effects of gender in this study are some of the strongest and most statistically significant results. My micro survey data may, however, be an exception to the more macro level data fields analyzed. Mexican migration studies by Stark (1991), Quinn (2006) have also found that men have higher propensities to migrate than women.

An improvement speaking English, even a modest one, increases the odds of migrating internationally and domestically. Compared to not migrating, it almost triples the odds of migrating to the USA, it doubles the odds of migrating to countries other than the USA and it almost doubles the odds of migrating within Colombia. Massey (1987) and Donato et al. (1992) found significant returns to improvements in English language ability when coupled with USA migrating experience.

Having family living abroad increases the odds of migrating to the USA compared to staying home and compared to migrating within Colombia. It also increases the odds of migrating to countries other than the USA as opposed to both staying home and to migrating within Colombia. The importance of migration networks abroad is a result well confirmed in the literature (Massey 1990, p. 7-8).

Having been born out of the state of one's current residence increases the odds of migrating within Colombia as opposed to not migrating at all and as opposed to migrating to the USA--the latter effect is very small, however. The variable "state," in my estimation, is an instrument variable for connections in other states (departamentos) in Colombia. It simply shows that migrant families have a higher propensity to migrate again and, more than likely, back to their place of origin.

Having less family income and feeling relatively more deprived (a one standard deviation increase in the variable "Rdxfinc") reduces the odds of migrating within Colombia and the odds of migrating to countries other than the USA. Apparently being poorer and feeling relatively more deprived, even though it also seems to reduce the probability of migrating to the USA, the effect is not statistically different from zero.

Conclusions

Some of the results found in this study were not unexpected and validates the soundness of the survey data. Others were not expected and they become perhaps the major contributions of this study. Those that were not surprising include:

1. A person's age has a positive impact on the probability of migrating to most places even though this impact wanes as the person passes middle age.

2. Persons with family members already living abroad have a higher probability of migrating abroad.

The list of somewhat unexpected results includes:

1. Males have higher propensities to migrate everywhere than females. This result was significant even when tested for different age cohorts.

2. A person's level of education as measured by their number of years of schooling or their highest degree earned was not significant in explaining migratory behavior. Taylor (1987), Massey and Espinosa (1997) as well as Quinn (2006) have found the same result when studying Mexico-US migration.

3. Having been born out of one's current state of residence increases the person's probability of migrating within Colombia as opposed to anywhere else.

The list of unexpected results includes:

1. Speaking English raises a person's probability of migrating everywhere: domestically and internationally.

2. The probability of migrating to the United States is not significantly reduced by advanced age.

3. Being part of a family that is relatively poor and experiences high levels of relative deprivation reduces the probability of migrating to most places. The only time when this result was not significant was in migrating to the USA. Speaking English, having relatives living abroad and being older mattered most when migrating to the USA. Certainly this list of requirements in no way contradicts the immigration preference policies of the USA.

As for this study's approach, a useful insight was gained in recognizing that the dynamic relationship between relative deprivation and family income was best modeled as the product of the two. Figure 2 shows that on average, the individuals who migrated belonged to families that stood to gain significant reductions in relative deprivation. The mean value of "rdxfinc" for the 172 individuals from El Prado that migrated out of 1,088 total considered could have plotted anywhere on the curve, but it fell on some of the steepest portions of the curve. The same can be said for the 108 individuals from Santa Ana who migrated out of 558 individuals considered in that community.

Policy Implications

As for the policy implications of this study, it generally reinforces the idea that no quick fixes exist for Colombian authorities to reduce migration flows, especially abroad. The propensity of individuals and families with marketable human capital, such as speaking English, with established migrating networks abroad will continue to flow out of the country. Nations on the receiving end of this flow such as the United States have more power to reduce this migration than has Colombia. I say this because the profile identified in this study of the individual most likely to migrate to the USA conforms very closely to the immigration policies of the United States.

Final Note

No study of migration within or from Colombia can be complete without addressing Colombia's chronic internal social conflict. The city of Palmira, where the studied neighborhoods are located, has been, for the most part, spared from guerrilla attacks that have plagued primarily rural areas in Colombia. In this cross-section study, the impact of Colombia's civil unrest would most definitely increase average propensities to migrate across the board. It is not difficult to imagine that the positive correlation between income and migration that resulted in this study may be also due to the ability of higher income (lower relative deprivation) to buy more of some normal goods we all value very highly: peace and safety.

[FIGURE 2 OMITTED]

Acknowledgements The author wishes to acknowledge the cooperation of the economics department at Universidad Javeriana in Cali, Colombia, for their support in the survey phase of this study. The author also wishes to thank Dr. James Hughes and Dr. Alan Levy, both of Slippery Rock University, for their editorial suggestions. The usual caveats remain regarding errors.

Published online: 12 March 2008

References

Donato, K. M., Durand, D., & Massey, D. (1992). Stemming the tide? Assessing the deterrent effects of the immigration reform and control act. Demography, 29, 139-157.

Fields, G. S. (1979). Lifetime migration in Colombia: Tests of the expected income hypothesis. Population and Development Review, 5(2), 245-259.

Fields. G. S. (1982). Place-to-place migration in Colombia. Economic Development and Cultural Change, 30, 539-558.

Massey, D. (1987). The Ethnosurvey in theory and practice. International Migration Review, 21(1), 498-522.

Massey, D. (1990). Social structure, household strategies, and the cumulative causation of migration. Population Index, 56(1), 3-26.

Massey, D., & Espinosa, K. E. (1997). What's driving Mexico U.S. migration? A theoretical, empirical and policy analysis. American Journal of Sociology, 102, 939-399.

Quinn, M. A. (2006). Relative deprivation, wage differentials and Mexican migration. Review of Development Economics, 10(1), 135-153.

Runciman, W. G. (1966). Relative deprivation and social justice: A study of attitudes to social inequality in twentieth-century England. Berkely, CA: University of California Press.

Stark, O. (1984). Rural-to-urban migration in LDC's: A relative deprivation approach. Economic Development and Cultural Change, 32, 475-486.

Stark, O. (1991). The migration of labor pp. 102-166. Cambridge, MA: Basil Blackwell.

Stark, O., & Taylor, J. E. (1991). Migration incentives, migration types: The role of relative deprivation. Economic Journal, 101, 1163-1178.

Stark, O., & Yitzhaki, S. (1988). Labor migration as a response to relative deprivation. Journal of Population Economics, 1, 57-70.

Stouffer, S. A., Suchman, E. A., DeVinney, L. C., Star, S. A., & Williams, R. M. (1949). The American soldier: Adjustment during army life. Princeton, NJ: Princenton University Press.

Taylor, J. E. (1987). Undocumented Mexico U.S. Migration and the Returns to Households in Rural Mexico. American Journal of Agricultural Economics, 69, 626-638.

Yitzhaki, S. (1982). Relative deprivation and economic welfare. European Economic Review 17, 99-113.

J. M. Valencia ([e-mail])

Slippery Rock University, Slippery Rock, PA, USA

e-mail: tejista@yahoo.com

(1) Most families living in El Prado owned their home since this was a government-sponsored housing project in 1962 that gave families that qualified a chance to own their own home.

(2) Since the best-fitting estimates of the Hickman model revealed that the p-value of the null hypothesis that (rho) [rho]=0 was only 11.88%, I decided to use the Hickman adjusted estimates of the individual's contribution for those that had migrated anyway and thus be more cautious. The Hickman estimates eliminated some high contributing individuals. For those that did not migrate (1,366 individuals), the actual monthly mean contribution to family income was a penny away from their mean contribution as estimated by the Hickman selection bias model. Those with negative estimated contributions were of course set to zero.

(3) Assuming the more tractable logit distribution for the error term and denoting as L its cumulative density function leads to the classical definition of the probability of migrating defined as:

(4) Stark and Taylor (1991, p. 1175, footnote 6 also noted in their work that the costs of migration could represent an obstacle to migration especially for the poorest families in the community. They thus reasoned that relative deprivation may not have a positive effect on the propensity to migrate but only for these families.

(5) For the purposes of this study, a migrant was someone that had migrated within the previous 18 months. Migrants within Colombia were only those that migrated to another State (Departamento).

(6) A likelihood ratio test was used to test the significance of each coefficient in explaining migration as a whole. The coefficients for "age" and "agesq" were tested together as a set. All tests rejected the null hypothesis of no significance at better than 99.9% probability. The variable "state" did so at better than 98.5% and the variable "rdxfinc" at better than 99%.

(7) Given that the factor change in the odds ratio as a result of an [delta] change in the variable [x.sub.k] is equal to

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The change in the factor odds as both age and agesq change simultaneously should be calculated as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The resulting factor change in the odds is a function of the person's age, as it should be.
Table 1 Some socioeconomic comparisons of the two communities

Socioeconomic Measure                         El Prado   Santa Ana

Family size                                       5.93        3.99
Children                                          1.59        0.89
Dependents                                        3.21        1.96
Percentage that own their home                   78.6        77.8
Value of home (millions)                         17.2        29.7
Percentage that own other property               18.3        42.0
Value of other property (millions)                3.04       14.2
Family wealth (millions)                         22.73       47.24
Years of study                                    8.88       12.75
Annual family income (thou.)                  9,149.82   16,339.3
Income per capita (thou.)                     1,947.12    4,662.61
Percentage that own a business                   29.6        19.7
Monthly business profits (thou.)                 51.86       70.02
Percentage that have a maid                      10.0        51.0
Monthly cost of maid services (thou.)             9.54       69.9

All currencies are in 1996 Colombian pesos.

Table 2 Survey results of do-it-yourself work

Activity                                  El Prado   Santa Ana

% Who fix their home                         57.2        33.2
Value of do-yourself work (thou. pesos        2.83        2.88
  per month)
% of families who sew                        40.3        18.8
Value of sewing (thou. pesos per month)       1.53        1.27

Table 3 Best-fitting model for the multinomial variable "choice"

Multinomial Logistic Regression      Number of obs=1646
                                     LR [chi square](21) = 425.98
                                     Prob > [chi square] = 0.0000
Log likelihood = -834.54236          Pseudo [R.sup.2] = 0.2033

            Choice   Coef.        SE          Z              P >
                                                       [absolute
                                                      value of z]

Colombia      Male    1.096225    0.2047499    5.35         0.000
               Age    0.1842426   0.0385769    4.78         0.000
             Agesq   -0.002003    0.0004454   -4.50         0.000
           Spkengl    0.8211127   0.1585671    5.18         0.000
             Famex   -0.0740122   0.2264112   -0.33         0.744
             State    0.7345518   0.2345793    3.13         0.002
           Rdxfinc   -1.95e-08    6.52e-09    -2.99         0.003
             _cons   -7.563153    0.8247659   -9.17         0.000
Notusa        Male    0.6327358   0.269599     2.35         0.019
               Age    0.3058873   0.0730849    4.19         0.000
             Agesq   -0.0035662   0.0009058   -3.94         0.000
           Spkengl    1.4743      0.1744314    8.45         0.000
             Famex    1.224571    0.2681236    4.57         0.000
             State    0.3031564   0.3654492    0.83         0.407
           Rdxfinc   -2.32e-08    8.89e-09    -2.61         0.009
             _cons   -11.4718     1.465624    -7.83         0.000
USA           Male    0.7161854   0.2723805    2.63         0.009
               Age    0.0787211   0.048943     1.61         0.108
             Agesq   -0.0007213   0.0005769   -1.25         0.211
           spkengl    1.995216    0.1768879   11.28         0.000
             Famex    1.703264    0.2758146    6.18         0.000
             State   -0.2163905   0.4080956   -0.53         0.596
            Rdxfmc   -5.77e-09    7.65e-09    -0.75         0.451
              cons   -8.830751    1.034802    -8.53         0.000

Choice = STAYHOME is the base outcome

Table 4 Marginal changes in probability and odds ratios

                                    Male    Spkengl
                                    (0-l)   (+l)

Avg. change in probs.   Colombia    +6.6%    +4.3%
                             USA    +1.25%   +4.20%
                           Other    +1.13%   +2.97%
                        Stay home   -9.0%   -11.5%
Factor change in odds        C/S     2.99     2.27
ratios                       U/S     2.04     7.35
                             O/S     1.88     4.37
                             C/U
                             U/C              3.23
                             O/C              1.92
                             U/O              1.68

                        Famex       State   Rdxfinc +I
                        (0-I)       (0-.l)  SD

Avg. change in probs.     -0.9%     +5.4%    -1.85%
                          +5.10%    -0.5%    -1.3%
                          +3.2%     +0.5%    -0.75%
                          -7.4%     -5.4%    +2.73%
Factor change in odds                2.08     0.722
ratios                     5.49
                           3.40               0.679
                                     2.58
                           5.91
                           3.66

Empty cells for the odds ratios indicate not significant at 5% or less
C Colombia, U USA, 0 other than USA, S stay home
COPYRIGHT 2008 Atlantic Economic Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2008 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Valencia, Jesus M.
Publication:Atlantic Economic Journal
Article Type:Report
Geographic Code:1USA
Date:Jun 1, 2008
Words:5816
Previous Article:Revenue effects of tax facilities for pension savings.
Next Article:FTPL-perspective on tradable deficit permits in the EMU.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |