Learning Citizenship? How State Education Reforms Affect Parents’ Political Attitudes and Behavior

Other kinds of education accountability policies exist. For example, some states have school choice policies that allow students to enroll in a variety of different types of schools; while others have student accountability policies that require students to pass exit examinations as a condition of receiving a diploma. I focus on content standards, assessments, and school accountability policies in this paper because, unlike the other reforms described here, these three types of policies have been adopted in every state.

Of course, it is possible that these policies also affect the attitudes and behaviors of other groups, such a students, teachers, and the public at large. Future research should examine the scope and direction of the effects of school accountability policies on these other groups.

For the purposes of this article, I set aside the extremely fraught question of whether the actual impact of these policies on student achievement also affects parental attitudes and behaviors.

5.8 % were “unsure”.

Critically, because these policies are instituted at the state level (rather than locality by locality), the costs to parents and children of exit are very high. While parents routinely make decisions about where to live within a given community in part based on local school quality, they cannot easily move to another state if they are dissatisfied with state standards, testing, and accountability policies. Consequently, parents may feel “stuck” with state-instituted reforms they oppose, exacerbating feelings of disempowerment.

Polimetrix is regularly used by social science researchers to conduct public opinion surveys. For example, the firm conducts the surveys for the Cooperative Congressional Elections Studies project (Cooperative Congressional Elections Study 2012).

When verifying information is not provided, EPE staff attempt to locate adequate information from publicly accessible records.

For more information on the Common Core of Data, see http://​nces.​ed.​gov/​ccd/​.

Estimating separate models for each of the variables comprising the index also sheds light on whether a particular subset of items is influencing estimates of results for the index as a whole.

I thank an anonymous reviewer for the encouragement to use these measures in my analysis.

Importantly, granular knowledge about how standards-based reforms affect particular modes of parental involvement is consistent with broader trends in research on how parents participate in their children’s education. Whereas earlier research tended to clump together various different forms of parental involvement into broad indices, more recent research has focused greater attention on the causes and consequences of specific modes of involvement (e.g. Hill and Tyson 2009).

The 10th grade exam was not offered in 2011.

This was the most recent national election prior to the period (January 2012) during which the survey was administered.

Recent research (Arceneaux and Nickerson 2009; Angrist and Pitsche 2009) demonstrates that robust clustered standard errors are equally effective as more complex methods (random effects, HLM) in producing precise estimates of group-level effects when the number of clusters is above 20, as it is in my research.

The means and standard deviations of the variables constituting the Political Participation Index were: Contact Official (mean = 0.37, SD = 0.48); Attend Rally (mean = 0.13, SD = 0.35); Contribute to Campaign (mean = 0.17, SD = 0.37); Worked for Campaign (mean = 0.09, SD = 0.29); Wore Button (mean = 0.25, SD = 0.43); Joined Internet Group (mean = 0.11, SD = 0.32); Joined Political Organization (mean = 0.16, SD = 0.36).

The means and standard deviations of the variables constituting the Parental Involvement Index were: Contact Teacher (mean = 0.74, SD = 0.44); Attend Open House (mean = 0.76, SD = 0.43); Attend Parent-Teacher Association Meeting (mean = 0.44, SD = 0.50); Attend Parent–Teacher Conference (mean = 0.71, SD = 0.45); Attend Parental Advisory Meeting (mean = 0.30, SD = 0.45); Attend School Event (mean = 0.71, SD = 0.45); Volunteer at School (mean = 0.33, SD = 0.47); Help at Fundraiser (mean = 0.46, SD = 0.50); and Attend School Board Meeting (mean = 0.27, SD = 0.45).

In the raw data, State Standards appears to have a modest negative effect on School Quality, but not the other dependent variables. However, when I matched the data on State Standards and re-estimated the models, the effect disappeared. Thus, I consider this result highly tentative. This analysis is available from the author on request.

Detailed information about post-matching balance is available from the author on request.

In an attempt to assess whether and to what extent the matched samples reflected or departed from the full sample, I compared the variable means for the treatment and control groups across the matched and full samples for each of the five matched datasets. I did not observe large differences across the two samples, though (as expected) overall balance across the treatment and control groups was better in the matched samples. However, as Imbens and Wooldridge (2008) note, it is possible that matched samples differ from full samples across unobserved characteristics, and such differences (by definition) cannot be evaluated. This is a primary reason why care must be taken in discussing the generalizability of results obtained from matched samples.

Notably, Rosenbaum sensitivity analysis makes the extremely strong assumptions that the unobserved covariate(s) “exhibit a strong, near perfect relationship with the response” (Rosenbaum 2002, p. 111) and are independent from the covariates used in matching, which are unlikely to hold in most practical applications (see also Mayer 2011, p. 642).

The results suggested that the state standards, assessments, and accountability policies had no effect on any of these variables, with the exceptions that School Accountability was positively associated with the probability that parents would report wearing a button (p < .05) and joining an internet-based political group (p < .05). To further investigate the relationship between School Accountability and the political participation variables, I employed genetic matching. I recoded School Accountability as a dichotomous variable with values 0/4 recoded as 0, and 5 recoded as 1 (I selected this cut-point because the average value of School Accountability was 4.04 and because this coarsening divided the sample almost exactly in half). Then, I genetically matched this coarsened measure on the other independent variables in each of the five datasets, and re-estimated the logistic regression models on the political participation variables in the genetically matched datasets. The results of these analyses provided no support for the hypothesis that School Accountability was positively associated with the probability of participating in these political activities. I concluded that the relationship between School Accountability and these participation variables was tenuous.

While the multivariate model for Attend Open House also appears indicate a statistically significant negative effect for State Assessments, I was unable to replicate this result using the rbounds command. Consequently, I consider this result more tentative. The discrepancy in this case may be due to the fact that the genetic matching procedure for my multivariate analysis followed best practice by matching with replacement, while proper implementation of sensitivity analysis with the rbounds command requires matching without replacement. Notably, the results for the other variables with statistically significant effects held regardless of whether matching was done with or without replacement, or whether the analysis was conducted with multivariate methods or using rbounds.

Full results are available from the author on request.