The Long-Term Effect of Childhood Residential Mobility on Educational Attainment

Abstract

Moving during childhood has typically been found to have negative effects on educational attainment. The loss of important social relationships is likely the cause. Many studies that examine the effect of residential mobility on education take a relatively short-term view by only looking at the effect for those no older than their late teens or early 20s. However, because many family moves are beneficial to children, in terms of improving their neighborhood or home environment, it is possible that the positive effects of family mobility do not appear until much later. I examine the effect of childhood residential mobility on educational attainment using the 1986 Canadian General Social Survey for individuals aged 25 and over. Results suggest that over the long run, residential mobility in childhood is beneficial for later educational attainment in that those who move between birth and age 15 are more likely to eventually graduate from high school than those who remain in the same community.

ACKNOWLEDGMENTS

I want to thank Doug Downey, Paula Arriagada, and Roderic Beaujot for their assistance regarding conceptual and methodological issues. In addition, I also want to thank Peter Kivisto and the anonymous reviewers for their useful and supportive observations during the final phases of this research.

NOTES

Notes

¹ An earlier review of the literature did not come to the conclusion that residential mobility was always necessarily harmful (see Shumaker and Stokols 1982).

² The study by Werkman et al. (1981) included children of parents who moved overseas, Goebel's (1981) study included high school students from two schools in a midwestern community of the United States, and Mann's (1972) study included undergraduate psychology students from the University of Texas-Austin.

³ Residents of the Yukon and Northwest Territories, as well as full-time residents of institutions are excluded from the sample. The response rate for the survey is 83.3 percent (Statistics Canada 1986).

⁴ In additional analyses, the sample was truncated to those under age 55 to reduce the differential mortality risk. These results are similar to those presented in this article and are available upon request.

⁵ Of the 3,346 that were dropped, 1,275 were students. In the remaining sample of individuals older than age 25, only 184 (less than 2 percent) were students and so were retained.

⁶ This eliminated a total of 1,275 respondents from the remaining sample.

⁷ From this remaining sample, 3,214 (about 29 percent) stated they did not know their father's education, and 2,704 (about 24 percent) did not know their mother's education level. I include dummy variables indicating this in all analyses (see Vach 1994).

⁸ A suitable continuous measure that captured number of years from kindergarten to beyond high school was not available in these data.

⁹ Also in Table 1, we see a definite inverse relationship between age cohort and residential mobility.

¹⁰ A better comparison may be with interstate movers from the United States. For example, Schachter (2004) reports that less than 3 percent of Americans aged one and over moved between states in a year; a figure which has been quite constant since the mid-1990s.

¹¹ In analyses not shown, a linear measure of mobility was also used. These results also show a positive relationship with education; however it was weaker. Also to allow for possible curvilinearity, a squared term was used, which was not significant once demographic controls were added. The findings using the squared term, however, reflect those presented in that the squared term becomes insignificant at the same time that five or more communities drop out of significance. I use the current categorical measure to more clearly illustrate this curvilinear effect on education.

¹² See Equation 5, 1 − e^(−0.202) = 0.18.

¹³ e^(0.465) = 1.59; e^(0.483) = 1.62.

¹⁴ 1 − e^(−0.324) = 0.28; 1 − e^(−0.221) = 0.20; 1 − e^(−0.177) = 0.16.

¹⁵ I only discuss the control variables from the logistic regression results from Table 3; those from Table 4 are substantively similar.

¹⁶ Age is only grouped in five-year categories and so this is the youngest age range at which high school completion can be measured; 15- to 19-year-olds include those in school. The analysis on ages 20 to 24 is not presented but is available upon request.

¹⁷ This could have been due to a smaller sample size as the age 20 to 24 sample only had 1,736 respondents.

¹⁸ In other analysis (not shown) that examined high school completion restricted to an even older sample (age 35 and older), the positive effect of greater residential mobility is even more pronounced: The positive effect of living in five or more communities was never reduced to nonsignificance.

¹⁹ The results of the logit regression are available upon request.

²⁰ It should be pointed out, however, that only three interaction terms are significant at the.05 level: high school × three communities, high school × five-plus communities, and at least some postsecondary × five-plus communities.

²¹ The results of the logit regression are available upon request.

²² Movers were defined as children who moved to a suburb of one of Canada's largest urban areas (Toronto) from some other place in Canada. This place could have been within the same province or from across the country.

²³ Population per square mile is 8 in Canada and 79 in the United States (Population Reference Bureau 2004).

²⁴ Long's (1992) study did not include Canada, however; the comparison with the United States was with New Zealand, Great Britain, Japan, Belgium, and Ireland.

²⁵ This is a challenging endeavor because of the confounding influences of distance and population size. In trying to disentangle these issues within the United States, Herting, Grusky, and Van Rompaey (1997) point out the complexities of modeling migration propensities across states of vastly different population size and density.