Rural Youth Residential Preferences: Understanding the Youth Development-Community Development Nexus

Abstract

Residential aspirations of youth are influenced by family, school, and community characteristics. This study uses data from 7th and 11th grade students in 10 rural Pennsylvania school districts to identify community characteristics associated with rural youth aspirations to live in their home community when they are 30-years-old. Multinomial logistic regression models comparing rural home community to other places and don't know responses show that 7th and 11th grade youth who like their community “a lot” and who perceive their community to offer viable career opportunities are more likely to want to stay in their home community. Higher educational aspirations are associated with higher odds of wanting to leave the home community. Just over 60% of the youth in this study want to live somewhere other than their rural home community as an adult.

Keywords:

• community development
• youth development
• youth out-migration

Notes

Note. *Significant at the .05 level; **significant at the .01 level; ***significant at the .001 level.

Note. Observations = 844; log-likelihood = −799.67; df = 44; likelihood ratio test = 212.52/p = .000; deviance statistic for good fit chi-sq = 159.91/p = .78; Cox and Snell's pseudo R ² = .22; classification accuracy = 54.4 > proportion by chance = 44.2.

*Significant at the .05 level; **significant at the .01 level; ***significant at the .001 level; ^∧ approaches significance at .05 level.

Note. Observations = 692; log-likelihood = −591.98; df = 48; likelihood ratio test = 154.05/p = .000; deviance statistic for good fit chi-sq = 1183.96/p = .99; Cox and Snell's pseudo R ² = .20; classification accuracy = 62.00 > proportion by chance = 52.43.

*Significant at the .05 level; **significant at the .01 level; ***significant at the .001 level; ^∧ approaches significance at .05 level.

Weighting is required because the RYE sampling frame of school districts was stratified into eight subgroups based on the school district typology. The population of students in rural Pennsylvania school districts is not distributed equally across these categories and school district size varied across schools in each category recruited to the study. Weights for the RYE sample were calculated based on the proportion of all students in the 7th or 11th grade in all of the schools in the state in each of the eight categories compared to the proportion in the sample in 7th and 11th grade from each category. Separate weights were calculated for the 7th and 11th grade samples. Item nonresponse was less than 5%, but to compensate for missing data, the missing data were imputed using a single imputation Markov Chain Monte Carlo method.

We include a discussion of bivariate findings among key variables for students who indicated aspirations to remain in their rural home area in order to inform concluding interpretations and discussion. Tables provided upon request.

We report the following standard fit indices: likelihood ratio test for model fit; deviance statistic for goodness-of-fit; Cox and Snell's pseudo R ²; and classification accuracy rate. The likelihood ratio test tests the significance difference between the likelihood ratio for the proposed model minus the likelihood ration for a reduced model (intercept only). The deviance goodness-of-fit test determines how well the model fits the data, and a small p value indicates a good fit. The classification accuracy rate is compared to the proportion of a chance criteria; when the classification rate is equal to or above the chance criteria, prediction is accurate.