The Lake Wobegon Effect in education and elsewhere

The tendency for people to like their local public schools more than they like schools across the nation comes through repeatedly and clearly through annual polls conducted by PDK International, a professional association for educators, and Education Next, a journal focusing on K-12 education issues. Education Next asks survey respondents to grade public schools using a letter grade system. Responses from parents of children in public schools are invariably in line with the Lake Wobegon effect. In 2017, 27% of parents gave “public schools in the nation as a whole” a grade of A or B, but 62% of parents gave “public schools in your community” such a rating. The numbers for the general population were 23% for schools in the nation overall and 54% for local schools (West et al. Reference West, Henderson, Peterson and Barrows2017). Similar results have occurred in the annual survey in previous years, dating to 2007, as well as in PDK International surveys (see Moe Reference Moe2001; Peterson, Henderson, and West Reference Peterson, Henderson and West2014).

Survey respondents’ general optimism about schools in their community does not translate into broad satisfaction with the amount of money being spent on them, however. As one example, 63% of respondents in the 2012 Education Next survey felt that “government funding for public schools in your district” should increase, while just 9% thought that it should decrease (Howell, West, and Peterson Reference Howell, West and Peterson2013). Even if the results concerned schools nationally, people were of the opinion that more spending was needed. For instance, 55% of respondents in the 2011 poll thought that teacher salaries should be increased, while just 7% said that they should be decreased. Perhaps not surprisingly, once the subject in these polls turned to raising taxes, respondents were quicker to close their wallets. The 2011 survey asked people whether “local taxes to fund public schools in your district should increase, decrease, or stay about the same.” A solid majority, 57%, thought that they should remain as they are with just 28% feeling they should be increased (Howell, West, and Peterson Reference Howell, West and Peterson2011).

The Lake Wobegon effect is not limited to attitudes toward public schools. Corporate managers and team leaders in 120 companies surveyed across nine countries tended to give higher marks to their organizations compared to competitors, with disparities decreasing only as respondents learned more information about their competitors (Betts, Croom, and Lu Reference Betts, Croom and Lu2011). High school students indicated that they possessed positive traits to a greater extent and negative traits to a lesser extent than others on average (Hoorens Reference Hoorens1995). College students with below average grades and test scores also were likely to overstate their performance (Maxwell and Lopus Reference Maxwell and Lopus1994). In addition, people tend to express unrealistic optimism about their health risks compared to others (Dunning, Heath, and Suls Reference Dunning, Heath and Suls2004). The illusion of superiority also can vary by the level of difficulty of tasks. Moore and Small (Reference Moore and Small2007) found that individuals rate themselves as being above average on easy tasks and below average on difficult tasks.

Psychological basis for illusory superiority

Hoorens (Reference Hoorens1993) defines illusory superiority as the degree to which an individual reports possessing positive characteristics to a greater extent than the average person, and having negative characteristics to a lesser extent than the average person. This way of thinking can apply to personality traits, abilities, conformity to group norms, or life circumstances. Illusory superiority is one of a collection of “self-related superiority biases” (Hoorens Reference Hoorens1993, 114). These biases also can include false consensus, in which an individual over-estimates the number of others who share an attribute with that individual; false uniqueness, in which an individual under-estimates the number of others who share an attribute with that individual; and unrealistic optimism, in which one over-estimates the likelihood of experiencing desirable events and under-estimates the likelihood of experiencing negative events. While illusory superiority can apply to assessments of one’s own abilities, this type of thinking can also extend to groups or organizations of which one is a member (Betts, Croom, and Lu Reference Betts, Croom and Lu2011). The authors caution that engaging in this type of thinking can pose a barrier to honest assessment and improvement, either at the individual or the organizational level.

Hoorens (Reference Hoorens1993) identifies several motivations for illusory superiority, including the need to maintain self-esteem, and to reassure oneself that one will be shielded from undesirable events. Presentation of oneself in a positive light, or “impression management,” also may be motivated by social desirability bias when responding to survey questions (Hoorens Reference Hoorens1995, 814). Mutz (Reference Mutz1998) cites personal motivations and cognitive errors as possible explanations for why individuals tend to rate their personal circumstances more positively than that of collective experience. People may be motivated to hold optimistic views of their situation in order to stave off fear or protect their egos. They may also be prone to misperceptions of collective experience if they tend to focus on “prototypes of high-risk individuals, which inevitably make the self look better in comparison” (Mutz Reference Mutz1998, 129).

Illusory superiority as a predictor of the Lake Wobegon effect

Scholars argue that the prevalence of comparing local schools more favorably to schools in general may be explained in part by familiarity. Loveless (Reference Loveless1997) contends that people typically assess social institutions in general according to harsher standards than they apply to institutions with which they are familiar. “The public’s relationship with the educational system writ large consists primarily of voting in elections and paying taxes. Local schools forge more intimate ties. Even nonparents may know friends, family, neighbors or colleagues at work who have children attending local schools” (154). Moe (Reference Moe2001) cites parental familiarity with public schools based on direct experience as a factor in overly positive assessments of school performance compared to nonparents. Moe finds that the parents tend to rate their local schools higher than nonparents even in school districts that are “disadvantaged,” as defined by socioeconomic factors and poor test performance (75). In addition to direct experience, parents may give high ratings to their local schools out of psychological necessity. Peterson, Henderson, and West (Reference Peterson, Henderson and West2014) refer to this necessity as “buyer’s delight,” observing: “It takes hard-nosed realists to admit that they are sending their child to a failing school or that the community in which they live has low-quality schools” (48).

Whether individuals are equipped to make accurate assessments of the performance of public schools is a subject of much discussion in scholarly research. Jacobsen, Snyder, and Saultz (Reference Jacobsen, Snyder and Saultz2014) argue that information asymmetries complicate citizens’ attempts to accurately evaluate the performance of schools in their communities. Most individuals have no direct contact with their local public schools. Even parents of children enrolled in the schools may have contacts that are limited to informal interactions with a small group of teachers on a regular basis (Jacobsen, Snyder, and Saultz Reference Jacobsen, Snyder and Saultz2014, 3).

Studies have generated mixed evidence regarding parents’ ability to accurately assess local school performance. Chingos, Henderson, and West (Reference Chingos, Henderson and West2012), analyzing national survey data and data from an oversample in Florida, found a significant and positive relationship between student performance on standardized tests and the public’s perception of local school quality nationally and in Florida. While the authors found a relationship, they did not test specifically for survey respondents’ knowledge of student or aggregate school performance on standardized tests. Howell (Reference Howell2006) tested for such knowledge and found significant gaps. A survey of 1,000 parents of students in public schools in Massachusetts showed that while 25% of parents who responded to the survey had children in schools that were rated as “underperforming” based on student scores on standardized tests, only 12% of parents correctly identified that their children attended underperforming schools (Howell Reference Howell2006, 152). At the same time, more than half of parents whose children attended higher performing schools were more likely to correctly characterize the schools their children were attending. Howell speculated that part of the gap may have been due to social desirability bias in responding to the survey. “After all, who wants to admit, especially to a stranger on the telephone, that they send their child to a public school that is underperforming?” (153). But Howell said some of the responsibility may also lie with the schools and disparities in the amount and effectiveness of communication between schools and parents. The study found that parents of students in higher performing schools were more likely to accurately identify the size of the school and the name of the principal, suggesting that those schools did a better job of communicating with parents.

Motivation may play a role in whether parents have accurate knowledge of school performance. While one might assume higher socioeconomic characteristics would predict accurate knowledge of school performance, Teske, Fitzpatrick, and Kaplan (Reference Teske, Fitzpatrick and Kaplan2006) found that in areas where school choice policies had become well established, perceived knowledge of school performance did not vary dramatically between low- and moderate-income parents. Only parents with extremely low annual household incomes, below $10,000, felt like they were not well informed about their local schools. Bickers and Stein (Reference Bickers and Stein1998), however, found that accurate perception of local school test scores was a function of parents’ levels of education, and how recently parents had moved to the community. Favero and Meier (Reference Favero and Meier2013) found that parent and teacher assessments of the performance of New York City’s public schools tended to be consistent with objective measures of school quality, including standardized test scores and crime statistics. The authors attributed this consistency to the motivation arising out of the school choice program in the New York City schools, in which parents relied on quality measures to make decisions as to where to enroll their children.

Whether this motivation, informed by factors such as standardized test scores, is widespread or unique to specific school systems also is a subject of debate. The Carnegie Foundation (1992) found that non-academic factors often played a key role in parents’ decisions. More recent studies confirm this tendency. Harris and Larsen (Reference Harris and Larsen2015), in a study of public schools in New Orleans, concluded that while academic factors matter somewhat for elementary school parents, “practical considerations such as distance and availability of extended school days seem especially important,” and for parents of high school students, “extracurricular activities such as band or football seem especially important” (3). Not all studies agree with these conclusions. Kelly and Scafidi (Reference Kelly and Scafidi2013) surveyed Georgians who were recipients of scholarships to attend a private school and found that “better student discipline,” “better learning environment,” and “smaller class sizes” were the most common bases for deciding where to go (1). Pride (Reference Pride2002) argues that for the community overall, it is not test scores that influence parents’ assessments. Instead, Pride contends, “critical events,” such as referenda on education policies, intervention by politicians, and even violent events such as school shootings, do more to shape the public’s assessment of schools than standardized test scores.

Does having accurate knowledge of student performance on standardized tests matter when parents choose a community based on the quality of its schools? Bickers and Stein (Reference Bickers and Stein1998) argue that even with imperfect knowledge homeowners can rely on “informational heuristics,” or psychological shortcuts, to correctly choose communities that provide the services that they deem desirable given what they can afford. For example, parents in search of high-quality schools might gravitate toward affluent communities based on reputation or property values.

While accurate knowledge of measures of school performance may or may not affect decisions about moving to a specific community, when individuals are exposed to actual performance indicators, they might revise their views of their local schools. Barrows et al. (Reference Barrows, Henderson, Peterson and West2016) found that average evaluations of local schools declined after survey respondents were exposed to information about their schools’ performance relative to other schools in their state, the nation, and in other developed countries. In sum, accurate information about school performance has the potential to reduce the incidence of illusory superiority when it comes to assessing the quality of local public schools. This suggests that there is a factual underpinning to illusory superiority, and that perceptions of facts shape views about school performance. It is possible, then, that misperceiving facts could lead to illusory superiority, and that misperceptions of facts could contribute to the Lake Wobegon effect when individuals compare their local public schools to public schools in general.

Factors that may shape views of school performance

Previous research suggests that parents of children in public schools may be more likely than others to overstate school performance, either due to a desire to cast their children’s schools in the most positive light to survey researchers, or to reassure themselves that their children attend good schools, or a mix of both motivations. Moe (Reference Moe2001) argues that direct contact with the schools may contribute to positive assessments, so the longer that parents have been involved with the schools the more likely they may be to think highly of the schools. Thus, it is possible that parents of students in public high schools might have more positive views than parents of students in middle or elementary schools.

In addition, individuals who own homes in a community may err on the positive side in evaluating their local schools because of the link between school performance and property values. Figlio and Lucas (Reference Figlio and Lucas2004) found that public school “report cards” issued by the state of Florida, with grades ranging from A to F, had a significant effect on home values, although the effect tapered off over time (see also McCabe Reference McCabe2013 for an extensive summary of research linking school performance to property values).

Length of time living in the community also might contribute to the overstatement of school quality due to ingrained community pride. It is plausible that individuals who are unhappy with the community will vote with their feet if they have the resources to move elsewhere. Those who stay behind may be satisfied with their community, and may develop feelings of pride out of a genuine appreciation for the community or as a rationalization for living there. Previous research suggests there are significant differences in attitudes toward public schools based on longevity of living in a community. Berkman and Plutzer (Reference Berkman and Plutzer2004), in a study of more than 9,000 school districts across 40 states, found a positive association between per pupil public school expenditures and concentrations of older residents ages 60 and up who had lived in a community for six years or more. The relationship was negative for 60-plus residents who had lived in the community for five years or fewer. The findings were “consistent with the idea that loyalty – an emotional bond between residents and their community’s institutions – competes with and often trumps instrumental self-interest” (Berkman and Plutzer Reference Berkman and Plutzer2004, 1190). It may be that the longer one lives in a community, the greater the possibility that one’s emotional attachment to institutions might get in the way of forming an accurate assessment of local schools. Research has found that more recent arrivals tend to have a more accurate sense of school performance, perhaps as a result of gathering information before deciding to move. Bickers and Stein (Reference Bickers and Stein1998) found that newcomers were more likely than long-term residents to correctly identify whether standardized test scores for the community were above, below, or the same as the county average.

The potential effects of other demographic characteristics are less clear cut. Previous research has found that African-Americans are less likely than the public as a whole to assign high ratings to local public schools, while Hispanics are more likely than the general public to do so (Peterson, Henderson, and West Reference Peterson, Henderson and West2014; see also Moe Reference Moe2001). In terms of education, Moe (Reference Moe2001) notes that one might reasonably expect individuals with high levels of education to have self-selected into more advantaged school districts, and therefore be more likely to express satisfaction with those schools. But Moe’s research found the opposite, that the least educated were the most likely to express satisfaction with public schools. Moe speculated that people with lower levels of education might also have lower expectations for their public schools, and consequently may be more likely to be satisfied with their schools.

Moe (Reference Moe2001) also discusses a concept called “public school ideology” that might play a role in shaping views on local public schools. The ideology centers on the idea that public schools, with their open access and egalitarian principles, are a cornerstone of the community and a manifestation of local democracy. Consequently, according to the ideology, local public schools deserve the community’s commitment and support (86–7). Moe speculates that public school ideology may be higher among individuals from low socioeconomic backgrounds, and that in the political sphere Democrats would be more likely to hold these views than Republicans or independent voters. Given the intersection between party identification, race and ethnicity, and to some extent socioeconomic status, predicting the effects of these characteristics on school ratings is difficult. We include them as controls in our models, but without a clear predictive sense of the direction of their influence.

Our focus instead is on the effects of having a child or children in the public schools, owning a home, and long-term residency on perceived school performance, and the role that those factual perceptions may play in comparing the quality of local schools to that of schools overall. These factors give rise to the following hypotheses:

Overstating local school performance on standardized tests also may shape views on school-related policies. It is possible that an inflated sense of school performance might affect how citizens view the need for additional resources for schools. If individuals view their local schools as performing better than schools overall in the state, they might conclude that the local schools have sufficient resources and that additional funds are not needed. On the other hand, one could argue that positive views of public schools might prompt support for additional resources to maintain and further enhance the quality of the schools. For example, Jacobsen, Snyder, and Saultz (Reference Jacobsen, Snyder and Saultz2014) argue that positive performance ratings of a public agency, such as a public-school system, confer a sense of legitimacy for the agency that can lead to support for more funding through taxation. We hypothesize that:

Methods and data

We test these hypotheses using data from a random-digit-dial telephone survey of 401 adults in Massachusetts that we conducted through a university-based survey research center (February 24 to March 2, 2014).Footnote ¹ We build on previous research by moving beyond simple comparisons of assessments of local schools to schools in general by comparing evaluations of local schools to actual school performance on standardized tests, and then by comparing that performance to average scores across the state of Massachusetts.

We asked survey respondents to compare the performance of their local public schools to that of schools across the state using the following measures:

1. 1. Using a grading scale of A, B, C, D, or failing, what grade would you give to the quality of education that the public schools in your local school district are providing?

2. 2. Using a grading scale of A, B, C, D, or failing, what grade would you give to the quality of education that public schools across Massachusetts are providing overall?

3. 3. How confident are you that the public schools in your local school district are providing a high-quality education to students? Are you very confident, somewhat confident, not very confident, not at all confident?

4. 4. How confident are you that public schools across Massachusetts are providing a high-quality education to students? Are you very confident, somewhat confident, not very confident, not at all confident?

5. 5. Using a grading scale of A, B, C, D, or failing, what grade would you give to the quality of teaching that occurs in your local school district?

6. 6. Using a grading scale of A, B, C, D, or failing, what grade would you give to the quality of teaching that occurs in public schools across Massachusetts overall?

Interviewers were trained to also accept “don’t know” and “refused” as voluntary responses for each question. We interspersed other questions between some of the measures to reduce question order effects. We also rotated questions about local schools with the matching statewide assessments to reduce the possibility of one question consistently anchoring responses to the next question (the full text of the questionnaire is available from the authors on request).

The pairs of questions in which respondents assigned letter grades to their local schools and schools across the state, expressed confidence regarding the local schools and schools statewide, and assigned letter grades to the quality of teaching locally and statewide, provided ample evidence of the Lake Wobegon effect (Table 1).

Table 1. Assessments of local schools and schools statewide in Massachusetts

Source: Massachusetts Statewide Survey, February 24 to March 2, 2014.

Note. N = 401. Figures are column percentages, and may not sum to 100% due to rounding.

When asked to assign letter grades for the overall quality of schools in their communities and statewide, respondents were more likely to give higher grades to local schools. About two-thirds of the sample assigned an A or a B to their local schools, while 50% of the sample assigned grades of A or B to schools overall in the state. A similar disparity emerged when respondents rated the quality of teaching, with 70% of respondents giving an A or a B to the teaching in their local schools, while 54% did so for teaching in schools statewide. In terms of confidence that public schools are providing a high-quality education, 31% of respondents said they were very confident in their local schools, while only 12% said they were very confident in schools across the state.

The aggregate percentages give some sense of the disparate views of survey respondents when they compare their local schools to schools statewide. But the differences reflected in the aggregate numbers might mask interesting patterns in the data. Do survey respondents as a group rate local schools higher than schools in general, or do some give equal rankings to both local schools and schools statewide, and do some rate their local schools lower than schools overall? Cross-tabulating the local and statewide assessments yielded the following patterns (Table 2).

Table 2. Patterns of ratings of local schools and schools statewide

Source: Massachusetts Statewide Survey, February 24 to March 2, 2014.

Note. N = 401. Figures are row percentages.

While large percentages of survey respondents gave higher marks to their local schools than to schools overall across the state, in no instance did this reflect the view of a majority of respondents. Thirty-six percent assigned a higher letter grade to their local schools compared to schools overall, while 35% gave the same grade, 15% gave lower grades to local schools compared to schools statewide, and 14% responded that they did not know or declined to give a response for one or both of the ratings. When asked about levels of confidence in schools locally and statewide, about 30% of the sample had higher confidence in local schools than in schools statewide, while a majority – 53% – had equal levels of confidence in schools locally and across the state. One-third of respondents gave higher letter grades to their local schools for the quality of teaching compared to schools statewide, while 46% assigned the same grade, and 9% assigned a lower grade to teaching in local schools compared to the state overall.

These findings give some context to the conventional wisdom that individuals tend to rate their local schools higher than schools in general. While this may be true, the results indicate that it is the case for only about one-third of the respondents to our survey. The larger question, and in some ways the more interesting one, is to what extent does misperception of school performance (the basis for illusory superiority) drive these ratings? We asked the following survey question to measure respondents’ perceptions of their local schools’ performance on a standardized test that the state administers annually in public schools:

“Next, the state measures the progress of students in public schools using a test known as MCAS. The test evaluates students in language arts, math, and science. Based on what you have read or heard, in general, do the schools in your local school district perform better on the MCAS tests than the state overall, worse than the state overall, or about the same as the state overall?”

The Massachusetts Comprehensive Assessment System (MCAS) tests students in public and charter schools in grades 3–8 and grade 10 in math, English language arts, and science and technology/engineering, under the provisions of the Massachusetts Education Reform Law of 1993. Following the adoption of the national No Child Left Behind law in 2001, Massachusetts began using MCAS results to assess whether schools were making adequate yearly progress toward meeting objectives laid out in No Child Left Behind. In addition, students in 10th grade must earn passing grades on the math, English language arts, and science and technology/engineering MCAS exams in order to partially meet requirements for receiving a high school diploma in Massachusetts (Massachusetts Department of Elementary and Secondary Education n.d.a). The tests are administered in the winter and spring, with individual results reported to parents over the summer and aggregate results released to the news media and the public in the fall. The state reports aggregated results at the state and school system level, as well as for individual schools and grades within each school, with percentages of student scores falling into one of four categories for each test: advanced, proficient, needs improvement, and warning/failing.

In response to the survey question on their local schools’ performance on the MCAS exams, 48% said their local schools performed better than the state overall on the exams, 8% said worse, 33% said about the same as the state overall, and 11% said they did not know or declined to answer the question.

In the demographic section of the survey, we asked respondents for their zip codes so that we could correctly place them in the public school system for their community.Footnote ² Using that information, we examined the 2013 MCAS scores for the public schools in each respondent’s community to assess the accuracy of respondents’ perceptions of local school performance on the standardized tests (Boston.com 2013). This validation process was a rough approximation of schools’ performance given that the overall scores for the school systems are aggregations of scores from different grade levels. In 2013, 69% of students across the state scored at the advanced or proficient level in English language arts, 61% in math, and 53% in science and technology/engineering (Massachusetts Department of Elementary and Secondary Education n.d.b). Table 3 shows the distribution of scores for advanced or proficient in the 207 communities in which the survey respondents resided.

Table 3. Comparison of MCAS scores in local schools and statewide in Massachusetts

Source: Massachusetts Department of Elementary and Secondary Education (n.d.a, n.d.b).

The mean scores for all of the respondents’ communities were within 1 percentage point of each of the statewide scores, with considerable variation in terms of maximum and minimum scores for the communities.

Matching community-level scores to survey respondents’ perceptions of the scores, however, revealed the extent to which respondents tended to overstate how their local schools performed. Using the survey responses and the local and state-level MCAS scores, we divided respondents into the following groups: those who had overstated their local schools’ performance compared to the state overall, those who had understated their local schools’ performance, and those who had been relatively accurate in their assessment of their local schools’ performance compared to the state overall. We also allowed for respondents who said they did not know or who declined to answer the question.

We calculated the difference between the scores on each exam in each respondent’s local public school system with the statewide score on that respective exam. We summed the differences across the three exams for each community, allowing for the possibility that the local schools might not have had consistent outcomes relative to the state overall across the three exams. For example, schools in a given community might have outscored the state on one or two exams, then scored below the state on the remaining exam(s). We coded communities that scored a total of at least three points higher than the state (theoretically scoring at least one point higher than the state on each of the three tests) as having scored higher than the state overall, and we coded communities with a negative aggregate difference of three points or more as having scored lower than the state overall. Allowing for close results, we coded communities that came within two points in either direction of the total state score as having scored about the same as the state overall. Among our sample of respondents, 41% lived in communities that scored lower than the state overall, 5% lived in towns that scored about the same as the state overall, and 55% lived in communities that scored better than the state overall (the percentages sum to 101 due to rounding).

We compared these results to respondents’ assessments of how their communities performed, and coded whether respondents rated their school performance lower than the actual results, about the same as the actual results, or higher than the actual results. In estimating the percentage of respondents who had overstated their local schools’ performance, we focused on survey respondents who fell into one of two categories. The first category encompassed respondents who said their local schools scored better than the state overall in percentages of students who scored advanced or proficient, when in fact their schools scored about the same or worse than the state overall. The second category consisted of respondents who said their schools scored about the same as the state overall, when in fact their local schools scored worse than the state overall. Respondents who had understated their local schools’ performance were those who had said their schools scored below the statewide numbers for schools overall, when their schools’ scores actually met or surpassed the statewide mark, or respondents who said their schools had scored about the same as the state overall when in fact their schools had exceeded the statewide scores.

We found that 14% understated their schools’ performance, 45% provided an accurate assessment, and 30% overstated their schools’ performance. Another 11% said they did not know how their local schools performed on the tests. Based on these results, we created a dichotomous measure of whether the respondent overstated school performance (1 = yes, 0 = no) as an indicator of illusory superiority. The variable reflects the extent to which individuals believe that their schools performed better than the state overall on the MCAS tests when the schools were lower than or level with the state’s performance, or that their schools were comparable to the statewide performance when the schools actually fell short of the statewide mark. Validating this phenomenon using the test scores for the local public schools helps to quantify the illusory part of illusory superiority in a way that is rare for scholarly treatments of the subject.

Multivariate models

Having illustrated indicators of illusory superiority in the sample, we now turn to multivariate models to test our hypotheses. In testing the models, we use a dependent variable that represents the traditional Lake Wobegon effect, and a dependent variable that quantifies the disconnection between assessments of schools and their actual performance on MCAS tests.

The traditional measure of the Lake Wobegon effect consists of a scale that we constructed from the three pairs of variables in which survey respondents compared local schools to schools statewide: the letter grades that respondents assigned to local schools and schools around the state; the levels of confidence respondents expressed in the quality of education that occurs in their local schools compared to schools statewide; and the letter grades respondents gave to the quality of teaching in local schools and in schools across the state. For each pair of measures, we assigned a value of one if the respondent rated local schools higher than schools in the state overall, and a value of zero otherwise. We summed the scores for the three pairs of measures, creating a four-category variable that reflected whether a respondent rated local schools higher than schools in the state overall for zero, one, two, or all three pairs of measures. Sixteen percent of respondents rated local schools higher than schools in the state overall for all three sets of measures, while 14% did so for two of the sets of measures, 24% did so for one of the pairs of measures, and 46% did not rate local schools higher than schools in the state overall on any of the sets of measures.

The other dependent variable that we used was based on the validated assessments of local school performance on the MCAS tests compared to the statewide performance, and captured the extent to which survey respondents overstated their local school performance on the standardized tests.

To test our hypotheses, we included in the models a measure of whether the respondent said s/he had one or more children enrolled in local public schools (23% of respondents). We also controlled for whether the respondent owned a home (77% of respondents). To capture longevity of residence, we created a four-category variable for years of residence: five years or fewer (18% of the sample), 6–15 years (27%), 16–25 years (18%), and more than 25 years (38%), and included the measures as dummy variables with the lowest category omitted as the reference category.Footnote ³ We chose to treat this category as categorical instead of continuous to try to distinguish between views of newcomers as opposed to long-term residents. Newcomers may have gathered specific information prior to moving and experiencing life in a community firsthand. Prior expectations and lived experience might vary, leading to changes over time, but not necessarily in a linear fashion. We also included a three-category measure of education: high school or less, some college, and college graduate, with dummy variables for some college and college graduate.Footnote ⁴ The model also controls for party identification (Democrat and Republican, with independents, third party voters, and nonvoters as the reference category), gender, and age, and dummy variables for respondents who identified as Black, Hispanic, or another race other than white. Given that we are measuring attitudes toward local schools, which are administered at the city and town level in Massachusetts, we estimated the models with robust standard errors clustered for the 207 cities and towns in which the respondents lived (Table 4).

Table 4. Predictors of rating local schools higher than schools in the state overall

Note. Dependent variable is a four-category measure of whether a survey respondent rated local schools higher than schools overall in the state on zero, one, two, or three dimensions. Coefficients are maximum likelihood estimates using ordered logistic regression. Standard errors are in parentheses. Models were estimated with robust standard errors clustered at the town or city level (207 clusters). N = 401. χ ² = 44.37, 14 df, p < 0.01 (Model 1); χ ² = 46.07, 15 df, p < 0.01 (Model 2). Pseudo-R ² = 0.04 (both models).

* p < 0.05.

^** p < 0.01 (two-tailed tests).

^† p < 0.10.

The model featuring the dependent variable scale constructed from three pairs of measures of school quality at the local and state levels provided some support for Hypothesis 1. Parents with one or more children enrolled in local public schools were more likely to rate local public schools highly than were respondents with no children in local public schools. The results also offer some evidence, in support of Hypothesis 2, that positive feelings about public schools will increase over time, with parents of students in high schools being more likely to favorably assess their local schools than parents with children in middle or elementary schools. Parents of students in public high schools were more likely than others to offer a positive assessment, but the effect was significant only at the level of p < 0.10 (p = 0.08). Parents of students in middle and elementary schools were not more likely than others in the sample to rate their local schools more positively than schools in the state overall.

The models do not provide evidence in support of Hypothesis 3, that homeowners would be more likely than others to assess local schools more positively than schools in the state overall. But there is evidence that, as longevity of residence increases, so too does the likelihood of rating local schools more positively than schools overall in the state (Hypothesis 4). With residential longevity of five years or fewer as the reference category, residents who have lived in communities for 16–25 years, or for more than 25 years, were significantly more likely to rate their local schools higher than schools in the state overall.

Education and race also had significant effects on whether survey respondents rated local schools higher than schools statewide. Using dummy variables for levels of education, we found that college graduates were more likely to rate local schools more positively than schools overall compared to respondents in the omitted category, individuals with a high school diploma or less. While Moe (Reference Moe2001) found that individuals with less education were more likely to express satisfaction with their local public schools, the opposite result appears here. It may be that as educational attainment increases, individuals self-select into more affluent communities with greater funding for public schools, and form more positive views of those schools. We also found that Black/Non-Hispanic respondents were more likely to rate their local schools higher than schools across the state compared to White/Non-Hispanic respondents. The controls for Hispanic respondents and those of other nonwhite races were not significant.

It is difficult to quantify the effects of the significant independent variables in an ordered logistic regression model, so we calculated the predicted probabilities for rating local schools higher than schools in the state overall for each of the significant independent variables. Using the predicted probabilities, we calculated the average marginal effects of the variables of interest (Table 5).

Table 5. Predicted probabilities of rating local schools higher than schools in the state overall on up to three measures (Models 1 and 2)

Parents who have one or more children in the local public schools had a predicted probability of 22% of rating their schools higher on all three dimensions, compared to 14% for those without children in the local public schools, for an average marginal effect of 8 percentage points. The average marginal effect for rating local schools higher on two dimensions was 4 percentage points higher for parents compared to non-parents. Residential longevity had similar effects, with individuals who had lived in their communities for 16–25 years being 8 percentage points more likely to rate their local schools higher on all three dimensions compared to those who had lived in their communities for five years or fewer. Among those who had lived in their communities for more than 25 years, the average marginal effect was 9 percentage points for rating local schools higher across the three sets of measures. The average marginal effect for having a college degree compared to a high school diploma or less was 7 percentage points. Race provided the largest marginal effect. Blacks were 40 percentage points more likely to rate their local schools higher than schools in the state overall on all three dimensions compared to White/Non-Hispanics.

In terms of overstating local school performance on standardized tests, our models generated mixed results (Table 6).

Table 6. Predictors of overstating local school performance on standardized tests

Note. Dependent variable is a dichotomous measure of whether respondent overstated local school performance on standardized tests. Coefficients are maximum likelihood estimates using logistic regression. Standard errors are in parentheses. Models were estimated with robust standard errors clustered at the town or city level (207 clusters). N = 401. χ ² = 37.69, 14 df, p < 0.01 (Model 3); χ ² = 37.71, 15 df, p < 0.01 (Model 4). Pseudo-R ² = 0.08 (both models).

* p < 0.05.

^** p < 0.01 (two-tailed tests).

^† p < 0.10.

One of the three groups of key stakeholders, long-term residents, were more likely to overstate how their local schools performed on standardized tests. Parents of students in public schools and homeowners were no more likely than others to overstate local school performance compared to the state overall. The effects for long-term residents varied by tenure, with those living in the community for more than 25 years the most likely to overstate local school performance compared to residents who lived in the community for five years or fewer, providing further support for Hypothesis 4. Residents of the community for 6–15 years also were more likely to overstate local school performance in Models 3 and 4, as were residents of 16–25 years, but the effects were significant only at p < 0.10.

As was the case with rating local schools in comparison to schools in the state overall, education was a significant predictor here, but in the opposite direction. College graduates were less likely to overstate local school test performance compared to those with a high school diploma or less. In contrast, in Models 1 and 2, college graduates were more likely to rate their local schools higher than schools in the state overall. Hispanic respondents also rated their local schools higher on test performance compared to White/Non-Hispanic respondents.

The average marginal effects provide additional details (Table 7).

Table 7. Predicted probability of overstating local school performance on standardized tests

Residents who had lived in their communities for more than 25 years were 22 percentage points more likely to overstate local school performance than residents who had lived in the community for five years or fewer. This may stem from a growing sense of connection to the community and pride in the community’s schools. Hispanic respondents also were 27 percentage points more likely to overstate local school performance. While there is little in the previous research to suggest why this might be the case, Howell (Reference Howell2006) found that Hispanic parents of children in public schools were less likely than white parents to accurately state whether the schools their children attended were making “adequate yearly progress” (154). Whether the same connection exists in our data is difficult to test, given the small number of Hispanic parents of children in public schools in the data (4 out of the 11 Hispanic respondents in the survey said that one or more of their children attended public schools).

The effect by education provides an interesting puzzle. College graduates were 18 percentage points less likely than those with a high school diploma or less to overstate local school performance. Yet college graduates were seven points more likely than those with a high school diploma or less to rate local schools higher than schools in the state overall in the three measures of educational quality (see Table 5). It may be that college graduates are more likely to view their local schools as being of higher quality than schools overall in the state, but may also have a more accurate view of school performance on standardized tests. When we re-ran these models with a dependent variable measuring whether respondents accurately described local school test performance, we found that education was a significant and positive predictor. Respondents with a college degree were more likely to accurately describe school performance compared to respondents with a high school diploma or less, with an average marginal effect of 26 percentage points (model results can be found in Table A2 in the Appendix).

Our results show that about 30% of survey respondents overstate local school performance on standardized tests compared to schools in the state overall. Length of time spent living in the community and Hispanic ethnicity are positive predictors of overstatement, and educational attainment is a negative predictor.

Is overstatement of school performance related to rating one’s local public schools higher than schools in the state overall? In other words, does misperception of local school performance positively predict a belief in the superiority of one’s local schools? We added perceptions of local school performance to our models’ predicting ratings of local schools relative to schools overall in the state (Table 8).

Table 8. Perceptions of local test performance as predictors of rating local schools higher than schools in the state overall

Note. Dependent variable is a four-category measure of whether a survey respondent rated local schools higher than schools overall in the state on zero, one, two, or three dimensions. Coefficients are maximum likelihood estimates using ordered logistic regression. Standard errors are in parentheses. Models were estimated with robust standard errors clustered at the town or city level (207 clusters). N = 401. χ ² = 117.78, 19 df, p < 0.01 (Model 5); χ ² = 120.08, 20 df, p < 0.01 (Model 6). Pseudo-R ² = 0.12 (both models).

* p < 0.05.

^** p < 0.01 (two-tailed tests).

^† p < 0.10.

The dependent variable in Models 5 and 6 is the same as in Models 1 and 2 in Table 4, a four-category measure in which respondents could rate local schools higher than schools overall on zero, one, two, or three dimensions. We added to the models’ dummy variables for whether respondents overstated local school performance on standardized tests, understated local school performance, or offered an accurate assessment. Further, we broke out accurate assessments by whether respondents correctly stated that their local public schools scored worse, about the same, or better than schools in the state overall. Respondents who said they did not know served as the omitted reference category. The results did not provide evidence of a link between overstatement of local school performance and rating local schools higher than schools overall in the state (Hypothesis 6). Instead, two categories of accurate assessment of local public schools had significant effects, but in opposite and interesting directions. Respondents who accurately described their local schools’ performance as worse than schools in the state overall were less likely to rate their local schools higher than schools across the state compared to people who responded “don’t know.” Those who correctly described their local schools as scoring higher than the state overall were more likely to rate their local schools higher than schools in the state overall. When controlling for perceptions of school performance, parents of students in public schools were not significantly more likely than others to rate their schools higher than schools in the state overall.

The results, then, do not provide a factual basis for illusory superiority, in that overstating local school performance does not have a significant effect on rating local schools more highly than schools in the state overall. Instead, accurate perceptions of local school performance are related to rating local schools differently than schools overall (Table 9).

Table 9. Predicted probabilities of rating local schools higher than schools in the state overall on up to three measures using assessments of school performance (Models 5 and 6)

Among respondents who correctly reported that their local schools scored higher than the state overall, the predicated probability of rating local schools higher than all schools in the state on all three measures was 0.32, compared 0.07 for all others, for an average marginal effect of 25 percentage points. Among those who correctly stated that their local schools scored worse than schools in the state overall, the probability of rating local schools higher than the state overall on none of the three dimensions was 0.78, compared to 0.44 for all others, an average marginal effect of 34 percentage points. The results indicate that accurate perception of school performance is related to comparative assessments of local schools and schools across the state.

While overstating school performance is not related to rating local schools higher, there is still the question of whether an overly rosy view of local school performance shapes policy preferences, particularly when it comes to additional funding for local schools. We asked survey respondents, “Would you be willing to pay higher taxes to improve the quality of public schools in your local school district?” Sixty-three percent said yes, 32% said no, and 5% said they did not know. We predict that respondents who overstate local school performance on standardized tests are more likely to view increasing taxes as further investment in strongly performing schools, and therefore may be more likely to be willing to pay higher taxes to improve the quality of their schools (Hypothesis 7).

We tested the hypothesis with willingness to pay higher taxes as the dependent variable (1 = yes, 0 = no). We excluded the 19 respondents who said they did not know or declined to answer the question, reducing the sample to 382 respondents. We used overstatement, understatement, and accurate descriptions of school performance as predictors, along with other variables used in the earlier models: whether the respondent has a child or children in the public schools, years of residence, whether the respondent is a homeowner, party identification, gender, age, education, and measures for Hispanic ethnicity and members of other nonwhite races. We had to omit the control variable for Black/Non-Hispanic respondents because all eight of those respondents indicated they would be willing to pay higher taxes to improve their local schools (Table 10).

Table 10. Predictors of willingness “to pay higher taxes to improve the quality of public schools in your local school district”

Note. Dependent variable consists of responses to the following question: “Would you be willing to pay higher taxes to improve the quality of public schools in your local school district?” 1 = yes, 0 = no. Coefficients are maximum likelihood estimates using logistic regression. Standard errors are in parentheses. Models were estimated with robust standard errors clustered at the town or city level (203 clusters). N = 382 (omits 19 cases in which respondents answered “don’t know” or refused to answer the question). χ ² = 41.82, 18 df, p < 0.01 (Model 7); χ ² = 43.92, 19 df, p < 0.01 (Model 8). Pseudo-R ² = 0.10 (Model 7) and 0.12 (Model 8).

* p < 0.05.

^** p < 0.01 (two-tailed tests).

^† p < 0.10.

Overstating local school performance was a positive predictor of willingness to pay taxes, as we hypothesized. Understating school performance also was a positive predictor, while accurately describing school performance had no effect across any of the three conditions of accuracy (correctly stating that local schools performed better, the same, or worse than schools in the state overall). Being the parent of a child or children in public schools had no significant effect for parents of all public-school students. But parental status was a significant predictor of willingness to pay higher taxes for respondents with children in middle and/or elementary schools (Model 8). Having a child in high school was not a significant predictor. Length of residence in the community was a negative predictor, while identifying as a Democratic voter had a positive effect. Republican Party identification had a negative effect, but the effect fell short of statistical significance (p = 0.10 for both models). Having a college degree exerted a positive influence, reaching significance at p = 0.047 in Model 7, but falling short of significance in Model 8 (p = 0.06).

Calculating the average marginal effects of the significant variables showed that overstating and understating school performance, as well as parenthood, had the largest effects (Table 11).

Table 11. Predicted probability of willingness to pay higher taxes to improve the quality of local public schools

Overstating and understating local school performance on standardized tests increased the probability of supporting higher taxes by an average marginal effect of about 21 percentage points. It might be the case that overstating school performance reflects a belief that the schools are doing a good job, and that additional tax dollars would be well spent to further improve the quality of local schools. The positive effect of understating school performance on support for raising taxes may rest on a different logic, that local schools are not doing well, and could use the additional financial support.

Parental views on a hypothetical tax increase also provided interesting results. The probability of being willing to pay higher taxes to improve public schools was 0.85 for parents of children in public middle and elementary schools, compared to 0.64 for those with children in private schools or who did not have school-aged children, an average marginal effect of 21 percentage points. Why would parents of middle school and elementary school students be more willing to pay higher taxes, while the same was not true for parents of high school students? We speculate that parents of students who still have several years to go in the public schools may view higher taxes as an investment with a direct payoff for their children. Parents of students who may be graduating soon may be less willing to pay higher taxes given that their children might benefit from the additional resources for a shorter period of time.