Jump to ContentJump to Main Navigation
School Choice and the Question of AccountabilityThe Milwaukee Experience$

Emily Van Dunk and Anneliese M. Dickman

Print publication date: 2003

Print ISBN-13: 9780300099423

Published to Yale Scholarship Online: October 2013

DOI: 10.12987/yale/9780300099423.001.0001

Show Summary Details
Page of

PRINTED FROM YALE SCHOLARSHIP ONLINE (www.yale.universitypressscholarship.com). (c) Copyright Yale University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in YSO for personal use (for details see http://www.yale.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 21 June 2018

What Parents Know

What Parents Know

An Examination of Informed Consumers

Chapter:
(p.74) Chapter Four What Parents Know
Source:
School Choice and the Question of Accountability
Author(s):

Emily Van Dunk

Anneliese M. Dickman

Publisher:
Yale University Press
DOI:10.12987/yale/9780300099423.003.0004

Abstract and Keywords

This chapter examines whether parents are able to send a clear message to schools by the act of choosing a school. It discusses the existence of informed consumers, focusing on the level of knowledge parents have that influences their decision in choosing or leaving a school. The chapter examines the following educational options for parents in the Milwaukee area: both aspects of the interdistrict Chapter 220 program (that is, suburban students choosing city schools as well as city students choosing suburban schools); the intradistrict three-choice selection process within Milwaukee public schools; and traditional tuition-based private schooling.

Keywords:   informed consumers, educational options, Milwaukee, Chapter 220 program, suburban students, city schools, city students, private schoolling

Until now we have been focusing on accountability by probing the responsiveness of both voucher and public schools to parents. In each case we have found this response to be deficient. We turn our attention in chapters 4 and 5 toward understanding whether parents are able to send a clear message to schools by the act of choosing a school. We examine first the existence of informed consumers.

In traditional, assignment-based public school systems school choice proponents believe parents have little incentive to gather information about schools because they do not select the schools their children attend; educators and elected school boards control school assignment. School choice proponents hope school choice will eliminate these impediments to parental control over education. The belief is that the empowering of parents to choose where their children attend school also allows parents to learn more about and actively participate in schools.1

School choice can take many forms and may therefore differ in the degree to which it induces parents to gather information. For instance, tuition-paying parents selecting private schools make up the largest competitive education market in the country. At the other extreme are public school district students, who are assigned to schools solely on the basis of their residency. Some public school districts allow (p.75) parents limited choices, which can include specialty or magnet schools. Other public school districts give first preference to neighborhood students and then open the remaining seats to students from across the district. Finally, a few public school districts have open enrollment throughout the district, in which parents are able to select across all available options and a lottery is held for seat assignment.

Despite the fact that school choice exists widely in private education and to some degree in various public school districts, we know little about the incentives these different choice systems provide parents when gathering schooling information. We would hypothesize that the more open the market, the more likely parents are to act as true consumers. That is, the fewer restrictions on a parent's ability to choose, the more likely the parent is to make a rational, beneficial decision. By rational and beneficial we mean a decision that satisfies the varying needs of parents in selecting a school. To make such a decision, however, parents need information. The debate over accountability is undeniably linked with the adequacy of information parents have about schools. To test our hypothesis, in this chapter we examine across various types of choice systems the amount of information parent consumers have about the schools they select, and whether these parents send consistent signals when choosing.

Accountability and School Choice

Research suggests that if parents' choices are to serve as an adequate school accountability system, they need to be based on high levels of information. Yet even Chubb and Moe acknowledge that markets are inevitably subject to all sorts of real-world imperfections, and one of these is that consumers may be too poorly informed to make choices that are truly in their best interests (1990: 34). Others point out that individuals having lower incomes and less education are less likely to possess adequate information (Henig 1999: 74). In addition, survey (p.76) research has demonstrated that parents are rather uninformed about their education choices, and many are unaware of choice programs that exist for their children (Public Agenda 1999).

Within these real-world limitations, Schneider and his coauthors studied parent-consumers of public school choice in four school districts. The major focus of their work was the ways in which school choice altered parents' incentives and in turn affected their behavior (2000). They addressed a number of issues relating to parental behavior, paying particular attention to the argument put forth by choice proponents that school choice increases the incentives for parents to obtain information about the schools their children attend. They found overall that most parents are uninformed about their schools; however, they found a subset of parents who are informed consumers. In their study, these informed consumers are more likely to be parents who actively chose their children's schools rather than having the district assign them a school. They identify these active, informed parents as marginal consumers. They argue that the marginal consumer, by making the best choices for him or herself, can strengthen overall parental accountability in two ways. First, marginal educational consumers influence other consumers by their market behavior, even without directly communicating with the less-informed parents. Second, marginal consumers pressure schools to improve. Therefore, for the education market to function effectively, only a subset of parents needs to be well informed about the various educational options their schools offer (Schneider et al. 1998: 784; for a more detailed discussion, see Schneider et al. 2000).

Other research on parental decision making suggests that parents who choose do not always send clear signals about their decisions to leave one school for another. Manna examined data from parent surveys conducted from 1990–91 to 1994–95 as part of an evaluation of the MPCP. In exploring these questions, Manna described the possible interplay between the decisions parents made in choosing or leaving a (p.77) school and the inferences school officials (and other parents) might draw from them. The author concluded that the messages parents send by their decisions to leave the public schools are not always clear (Manna 1999). Manna's research questions whether such choosers really influence other consumers or pressure schools to improve. Neither of the above studies directly examines school choice accountability, yet because of their focus on parent-consumers, their findings have implications for how schools participating in a choice system are held accountable.

Research Design

The research design we use borrows from the above analyses of choice by Schneider et al. (2000), Manna (1999), and also Witte (1991–95).2 We used data from long-established Milwaukee choice programs other than vouchers because data on participating voucher parents are not available to the public and because most growth in the Milwaukee voucher program occurred within the last few years. Our aim was to identify informed consumers who send consistent signals.3 If an efficiently operating market system of accountability, in the form of informed consumers, can exist, evidence of it is more likely in choice programs that are well established. Finding such evidence would support placing the responsibility for accountability in the hands of parents.

We surveyed parents who chose their schools in Milwaukee and the surrounding suburbs under various choice systems. As noted in chapter 1, Milwaukee parents can select among a number of choice programs for their children. For this study we examined the following educational options for parents in the Milwaukee area: both aspects of the interdistrict Chapter 220 program (that is, suburban students choosing city schools as well as city students choosing suburban schools); the intradistrict three-choice selection process within MPS; and traditional tuition-based private schooling.

(p.78) Chapter 220 Interdistrict Public School Choice

The Chapter 220 program, the oldest of Milwaukee's parental option programs, was established in 1976 to promote racial integration of Milwaukee and its surrounding suburban districts. The implementation of the program coincided with a federal court order for desegregation in Milwaukee, the result of declining nonminority enrollment in the city. It is a targeted choice program because only a minority pupil residing in a district with a concentration of minority pupils of 30 percent or greater may transfer into an adjacent district with 30 percent or less minority pupils. Likewise, majority students residing in districts with low numbers of minority students may transfer into an adjacent district with the higher specified concentrations.4 In the case of minority students each applicant can list up to three suburban districts but cannot choose specific schools within these districts. MPS conducts a random selection process to pick students and match them with their chosen districts, where they are assigned a school. If a student is not accepted at a preferred suburban district, the student is placed on a waiting list, from which he or she may be admitted until mid-December. A student must reapply for the program the following year if he or she is not selected off the waiting list. In our analysis we refer to this group of parents as Milwaukee 220. Suburban parents applying to MPS can apply to any of the more than 150 schools. We identify this group of parents as Suburban 220.

Intradistrict Public School Choice

The intradistrict public school choice system is made up of parents in MPS who chose to send their children to either a citywide specialty or regular school. Unlike parents in many public school systems, MPS parents must participate in a three-choice selection process implemented in 1991–92 when enrolling their children. Therefore, only a small percentage of parents whose children currently attend MPS did not choose the schools their children attend. The MPS three-choice selection (p.79) process is not a “controlled choice” process, as that term is used by other cities in the nation. Although Milwaukee's elementary students are limited to choosing among schools in particular areas of the city—areas that are drawn to be as racially balanced as possible—a student's choice of a particular school is, for the most part, not restricted by racial parameters. There are some exceptions to this general rule, however, and waiting lists for the citywide specialty schools are maintained by race. In addition, the school board has a policy of ensuring that any child who desires a racially balanced school has the opportunity to attend one. According to an April 1999 report on school selection in MPS, of the 20,109 students indicating their top three school choices for 1999–2000, 16,512, or 82 percent, were accepted to their first choice school. Another 1,329 students were accepted to their second or third choice school. Our intradistrict choice system is divided into two different categories depending on whether a parent selected a citywide or a regular MPS school.

Tuition-Based Private School Choice

The final choice system that is under study involves parents who send their children to private schools.5 Approximately 20,000 students who reside in Milwaukee attend private school. Traditional tuition-based private school choice is not a targeted program; parents must have the financial ability to pay for tuition.

As table 4.1 illustrates, 144,381 Milwaukee area students participate in these choice systems.

Survey Methodology

Do these choice systems provide different incentives to parents to gather information? To answer this we conducted a phone survey in fall 1999 of parents whose children participated in one of the choice (p.80)

Table 4.1. Schooling Options in the Milwaukee Metropolitan Area

Program

Parent Eligibility Milwaukee resident

Number of Participating Schools or Districts

Number of Participating Milwaukee Students

Regular Milwaukee Public Schools*

Milwaukee resident

119 (schools)

76,086

Interdistrict Chapter 220 Program

      Milwaukee minority parents (Mil-waukee 220)**

Minority Milwaukee resident

140 (schools)

4,859 (FTE)

      Suburban nonminority parents (Suburban 220)**

Nonminority resident of participating suburban district

160 (schools)

588 (FTE)

Charter schools***

Milwaukee resident

11 (schools)

5,048 (FTE)

Statewide open enrollment***

Resident of any Wisconsin public school district

426 (districts)

840

Milwaukee Parental Choice Program***

Low-income Milwaukee resident

103 (schools)

9,200 (FTE)

Traditional tuition-based private school (includes MPCP)**

Determined by independent schools

117 (schools)

27,208

(*) Data from 1999–2000 school year.

(**) Data from 1998–99 school year.

(***) Data from 2000–01

(p.81)

Table 4.2. Completed Phone Interviews by Category

Category

Completed Interviews

Percent of Total

Milwaukee 2201

201

30

Attend MPS2

234

35

Suburban 2203

63

9

Attend private school

180

27

    Total

678

101

(1) MPS 220 participants are those parents who reside in Milwaukee and participate in the 220 program by sending their children to a participating suburban district.

(2) Includes both MPS citywide and MPS regular.

(3) Suburban 220 participants include parents who reside in a district outside of Milwaukee and sent their children to MPS by participating in the Chapter 220 program.

Note: Total does not add up to 100 percent because of rounding.

programs (table 4.2). The survey included eighty questions and took approximately twenty-one minutes to complete. To reduce the number of total questions asked of any one respondent, fourteen of the eighty questions were asked only of the odd-numbered respondents and another set of fourteen were asked only of even-numbered respondents. This not only ensured that all groups would be equally represented in these questions, but also reduced the overall length of the survey. The survey included eighteen questions recorded for statistical purposes only. These questions allow for an understanding of the characteristics of the survey respondents. In addition, they make it possible to compare respondents by controlling for a number of personal characteristics theoretically related to education decisions.

Survey questions were designed to fit into one or more of four themes: perception of particular schools/programs, accessibility of information, satisfaction with school and the schooling option program, and parental involvement. In this chapter we focus on questions related to accessibility of information.

The survey was conducted as part of a contract with MPS. In designing the survey we borrowed the wording of questions from surveys used by Schneider and his coauthors (2000) and by Witte (1990–95) (p.82) in order to allow for replication of the analyses used by these authors and by authors who rely on their data. The majority of questions were designed to assist MPS in learning more about the actions of parents who participated in various choice programs. After it was determined which choice program the family participated in, that parent was asked how many of their children were enrolled in the program.6 If more than one, the parent was asked to limit his or her answers to the child who was having the next birthday. This ensured a random sample of children in all grade levels.7 It also ensured that answers were unique to one child. The total number of completed surveys was 678, a response rate of 20 percent.8

Table 4.3 illustrates the distribution of sampled respondents on several social characteristics. Parents whose children attend MPS or participate in the Milwaukee 220 program have very similar characteristics in terms of age, marital status and race, but Milwaukee 220 parents are wealthier and more likely to be college educated. Suburban 220 participants are more likely to be college graduates and have the highest income of all surveyed groups.

Findings: A First Look at the Information Level of Parent Consumers

We examined how much knowledge parents have about certain key characteristics of the schools their children attend. We asked half of our survey respondents questions about several school characteristics. The specific wording of the questions is as follows: (1) What do you think is the percentage of students in [INSERT SCHOOL] who are eligible for free or reduced-price lunch? Less than 50%, 51%–75%, Greater than 75%. (2) What do you think is the percentage of students in [INSERT SCHOOL] who have reading scores at or above grade level? Less than 50%, 51%–75%, Greater than 75%. (3) We'd like you to think about the racial makeup of [Name of Child] school. In particular we'd like you to tell us (p.83)

Table 4.3. Percentage Distribution of Sampled Respondents on Selected Social Characteristics

Average Age in Years of Respondents

Percent Married

Percent College Graduate or Higher

Percent Nonwhite

Percent Catholic

Percent Indicating

Milwaukee 220

37

57

30

81*

20

40

Attend MPS

39

56

22

70

21

32

Suburban 220

42

86

57

10*

48

67

Attend private school

40

79

35

29

28

50

    Total

39

68

33

57

28

54

(*) The data are on the parent, not the child. In order for a student to participate in the Chapter 220 program they have to represent the minority population in Milwaukee and the majority population in the suburbs. Biracial and transracial families can explain the 19 percent of parents of MPS 220 participants who are white and the 10 percent of Suburban 220 parents who are nonwhite.

(p.84) what percentage of students in [INSERT SCHOOL] you think are African American? 0–9%, 10%–25%, 26%–50%, 51%–75%, 76%–100%. We used a broad measure of knowledge by allowing the survey respondent to select within a range of percentages of a school's population on all three questions. These variables were included in the survey because we felt they would give us an opportunity to measure accuracy of information for a subset of the population. Schneider et al. (2000) included similar questions in their survey of school choice in four different districts.9 We compared this selection with the actual data for the child's school.10 Table 4.4 illustrates by program the accuracy of parental information on these characteristics.

Accuracy is highest for Milwaukee 220 parents on the percentage eligible for free or reduced-price lunch. Accuracy is lowest for MPS citywide parents who are more likely to believe that the percentage eligible for free or reduced-price lunch is lower than actual. On average, half the parents do not possess accurate information on these school characteristics. This is similar to Schneider et al.'s (2000: chap. 7) findings concerning only public school choosers. What is unique about our findings is that they are a direct comparison of several choice programs, both private and public. If inherent differences in incentives to gather accurate information exist in different choice systems, it should be obvious from our data.

In fact, we found no such differences. Given that all parents surveyed can be considered active choosers, perhaps it should not be surprising that parents who selected private schools were no more likely to have knowledge about those school characteristics than MPS parents participating in the public school three-choice selection system. Or that those same parents who pay private school tuition are no more knowledgeable than parents who could merely select a district for their child rather than a specific school, as in the Milwaukee 220 program. In one instance, Milwaukee 220 parents knew more than tuition-paying private school parents—84 percent of Milwaukee 220 parents were aware of (p.85)

Table 4.4. Levels of Parent Accuracy in Information

Number of Schools

Number of Parents

Percent of Parents Who Do Not Possess Accurate Information on School Characteristic

Percent African Americans in School

Milwaukee 220

59

81

57

Suburban 220

18

33

45

MPS

69

107

50

       Citywide

30

56

55

      Regular

39

51

44

Private

33

50

42

    Total

271

50

Percent Reading at Grade Level

Milwaukee 220

42

75

63

Suburban 220

16

32

47

MPS

66

104

58

      Citywide

32

53

70

      Regular

34

51

45

Private

NA

NA

NA

    Total

211

58

Percent Eligible for Free and Reduced Lunch

Milwaukee 220

48

96

16

Suburban 220

17

34

44

MPS

58

110

64

      Citywide

23

51

71

      Regular

35

59

58

Private

23

30

43

    Total

270

42

(p.86) the percentage of the student body of their school eligible for free or reduced-price lunch compared to 57 percent of private school parents. But for the most part, the majority of parents in both private and public school choice systems did not have accurate knowledge about their children's schools. The unanswered question is whether ignorance of details by half of the choosing parents supports entrusting school accountability to parents.11

Findings: A Closer Look at the Information Level of Parent Consumers

Before considering this question, we examined another means for assessing the knowledge level of parent-consumers. This entailed measuring responses to a survey question that does not pertain to specific facts about the demographic makeup of the student body or achievement level of students but merely asked for general school information. We know that in making most decisions people do not possess encyclopedic information about the matters in question (Lupia and McCubbins, 1998). Therefore, we believed it was important to examine whether parents had broad, general knowledge of the school. To do so, we asked whether parents know the name of the principal at their child's school. We did not verify if a parent was correct on this answer.12 We trusted that a positive answer to the question meant the respondent did know the name of the principal.

We found that the majority of parents knew the name of the principal; however, there were some differences across choice systems.13 At the high end, 88 percent of Suburban 220 parents knew the name of the principal compared to 83 percent of private school parents, 61 percent of Milwaukee 220 parents, 62 percent of MPS Citywide parents, and 58 percent of MPS Regular parents.

Several things are apparent from these data. Between 10 and 40 percent of parents in the different choice programs do not know the (p.87) names of the principals in their children's schools. Second, the difference in knowledge levels does not correlate with being a public or private school parent. Private school parents and Suburban 220 parents who choose a Milwaukee public school are equally likely to possess knowledge about this school fact. Third, the differences do not vary by whether a parent selects a district or a particular school. Milwaukee 220 parents, who can select only a school district, are just as likely to know the name of the principal as parents who participate in the MPS three-choice selection process. In light of these findings it is important to again highlight that all parents surveyed can be considered active choosers. For this reason, perhaps our expectations should be that their knowledge about their school would be similar. The data appear to support this assertion.

Because support for parental accountability hinges on the belief that choice empowers parents to gather information, it is important to examine whether the disparities we find in general school knowledge are due to differences in the choice programs or to other characteristics of the parent. To test the hypothesis that it is the choice program itself that matters, we developed a logistic regression model that analyzes the probability of knowing the name of the principal while controlling for other variables that may help explain the variation in this knowledge.

The independent variables that could help explain the variation in knowledge levels include the parent's race, educational level, household income, church attendance, and length of residence in the district. The race and/or ethnic group variable was coded one if the parent is white, zero if not.14 The education variable was coded one for parents who have a four-year college degree or higher, zero if not.15 Our income variable was coded into eight separate categories from “Under $10,000” to “Above $70,000.”16 Church attendance ranged from “Attend more than once a week” to “Never attend.”17 Length of residence in the district was a continuous variable measuring years in the district.18

The results demonstrate a statistically significant difference between two choice programs (table 4.5).19 Milwaukee 220 parents are the (p.88)

Table 4.5. Logistic Regression Model of Knowledge of Principal by Different Forms of Choice

Coefficient

Standard Error

Chi-Square

Log Odds Ratio Estimates

Intercept

1.08

0.22

23.52**

Type of choice1

    Milwaukee 220

−0.42

0.21

4.06*

0.45

    MPS Citywide

−0.42

0.27

2.42

0.52

    MPS Regular

−0.29

0.25

1.38

0.46

    Suburban 220

0.76

0.46

2.73

1.5

Education

    Less than college degree

−0.29

0.14

4.52*

0.56

Nonwhite

−0.38

0.14

7.17**

0.47

Length of time in district

0.008

0.01

0.4

Church attendance2

    More than once a week

0.73

0.26

8.01**

4.7

    Once a week

0.05

0.2

0.07

2.4

    Less than once a week but at least once a month

0.007

0.24

0.001

2.3

    Less than once a month but at least once a year

−0.03

0.29

0.01

2.2

    More than once a year

0.06

0.48

0.02

2.4

Household income3

    $0–10,999

−0.34

0.37

0.87

0.7

    $11,000–19,999

−0.08

0.28

0.07

0.92

    $20,000–29,999

−0.37

0.24

2.42

0.68

    $30,000–39,999

0.29

0.28

1.09

1.3

    $40,000–49,999

−0.03

0.28

0.01

0.96

    $50,000–59,999

0.12

0.3

0.16

1

    $60,000–69,999

0.39

0.4

0.92

1.5

Likelihood ratio chi-square test statistic

74.94

19 df

0.0001

Somers' D

0.48

Note: Dependent variable is 1 indicating individual knows principals name

(1) Reference group for log odds ratio is parents in tuition-based choice

(2) Reference group for log odds ratio are parents who are not church attendees

(3) Reference group for log odds ratio are respondents with income 〉 $70,000

(*) p. 〈 .05

(**) 〈 .01

(p.89) only group that is significantly different from private school parents after controlling for the individual characteristics of the parent.20 We used a statistic called the log odds ratio to measure the likelihood of a parent having knowledge of the principal as compared to a parent in the reference group (private school parents).21 The log odds ratio of .45 denotes that knowledge of the principal's name is nearly half as likely among Milwaukee 220 parents as among parents in tuition-based private school. There are no other significant variations between parents in any of the other choice systems and private school parents. We do find, however, that nonwhite parents and parents without a college degree are half as likely to know the names of their children's principals as white parents and college-educated parents, respectively. Finally, parents who attend church more than once a week are four and a half times more likely than parents who never attend church to know the names of their children's principals. Interestingly, parents who attend church more than once a week are just as likely to have their children enrolled in public school as in private school, therefore diminishing any anticipated connection between church attendance and knowledge of a private school.

Thus, in terms of general knowledge levels, the type of choice program matters in only one case. Parents who participate in traditional tuition-based private school choice are more likely to know the name of their children's principals than parents whose children attend suburban public schools under the Chapter 220 program. After controlling for race and education, we find no significant differences in knowledge level between parents who select private school and parents who select MPS. Therefore, with one exception, the parents choosing in the most marketlike choice system—traditional tuition-based private school—have no greater tendency to gather accurate information than parents choosing in a public school choice system.

Overall, in these first two examinations of school knowledge we find that a large number of parents are uninformed about the schools (p.90) their children attend. Fewer than half of the parents surveyed possess accurate information on specific school characteristics such as student achievement and student makeup. More parents do possess general information about the school, but, in contrast to our hypothesis, differences in general knowledge levels appear to be more a factor of race, education, or church attendance than the type of choice system. The case can be made that since parents in all choice systems are rather uninformed about their children's schools, additional accountability mechanisms are necessary to ensure adequate information is available for all of a parent's schooling options.

Findings: The Consistency of the Signals Parents Send When Choosing

We examined parents knowledge levels based on characteristics we think demonstrate adequate information about a school. Some may argue, however, that parents may not be interested in these characteristics and perhaps should not be expected to know about factors they do not deem important in a school. Thus, the next step in our analysis looked at what parents viewed as the important factors when selecting a school and how they communicated these factors to other parents and to other schools in the marketplace. The survey data indicate how clear these signals are, that is, whether parents actually sought information on the factors they said were important to them in selecting a school. For accountability to function accurately based on the actions of informed parent-consumers, parents should send clear messages about what they seek in a school, allowing competing schools to react and relay their responses to other parents. Ultimately, those schools that meet parents' desired characteristics should prosper, and those that do not will close.

In order to understand whether or not parents are sending consistent signals, we asked parents if they requested information on several factors when choosing a school. We then compared those responses with (p.91) the three factors they felt were the most important when choosing a school. The question on what factors were important was open-ended. The question asking if they gathered information on a particular question was closed-ended. For this reason, it was not possible to ask respondents if they gathered information on every important variable. The closed-ended question included those categories that we hypothesized would be very important in selecting a school. If we had been able to ask the data-gathering question for all thirty-five categories mentioned by parents, the likelihood of getting mixed signals would have increased. The closed-ended question on obtaining information read, “Did you obtain any of the following information on any of the schools you applied to?” The list included student achievement, class size, teacher qualifications, administrator/principal qualifications, what is being taught, method of teaching, incidents of discipline at the school, special programs, incidents of safety at the school, opportunities for parental involvement, and athletics and other extracurricular activities.

A mixed signal results where the parent indicated a variable was an important factor when choosing a school but did not gather information on it. For example, the interviewee with ID number 8 mentioned that what is taught at the school, safety and discipline at the school, and the administrators' qualifications were the three most important factors in choosing a school. But respondent 8 did not seek information on administrators or discipline at the school; therefore, this respondent presented a mixed signal. If a respondent did not gather information on any one of the three characteristics they deemed important in a school, the response was categorized as a mixed signal. Likewise, if a respondent did not gather information on more than one characteristic they deemed important we did not weight that as a higher-level mixed signal. Parents either gave a mixed signal or they did not.

There are two ways of interpreting the data presented in table 4.6. The first conclusion is that there is a remarkable degree of consistency between what parents say they look for in a school and what data (p.92)

Table 4.6. Number of Consistent and Mixed Signals when Gathering Information on Important Factors in a School

Consistent Signals: Sought information on factors parents said were important in selecting a school

Mixed Signals: Did not seek information on factors parents said were important in selecting a school

Number

Percent

Number

Percent

Total

Milwaukee 220

124

65

67

35

191

MPS

151

79

39

21

190

  Citywide

59

79

16

21

75

  Regular

92

80

23

20

115

Suburban 220

53

87

8

13

61

Private

138

78

38

22

176

  Total

466

75

152

25

618

they gather on a school. Depending on the choice program, between 65 and 87 percent of all parents are actually gathering information about a school based on the factors important to them. The second conclusion is that the mixed signals are troublesome. Mixed signals cause one to question how parents can hold schools accountable when they do not actually seek information about the factors they say are important in a school.

To find out whether the differences in the frequency of mixed signals are related to the type of choice program or to individual characteristics of the parents, we again used logistic regression, analyzing the same demographic variables as the previous analysis. We find parents who participate in the Milwaukee 220 program are more likely to send mixed signals than parents whose children attend traditional tuition-paying private schools (table 4.7).22 An examination of the log odds ratio estimates reveals that the occurrence of a consistent signal is one-third as frequent among Milwaukee 220 parents as it is among parents who choose tuition-based private schools. However, there are no significant (p.93)

Table 4.7. Logistic Regression Model of Different Forms of Choice and Gathering Information on Important Factors in a School

Coefficient

Standard Error

Chi-Square

Log Odds Ratio Estimates

Intercept

1.45

0.21

47.31**

Type of choice1

    Milwaukee 220

−0.86

0.2

18.25**

0.34

    MPS Citywide

0.16

0.28

0.33

0.93

    MPS Regular

−0.1

0.25

0.16

0.72

    Suburban 220

0.57

0.37

2.41

1.4

Education

    Less than college degree

−0.15

0.13

1.36

0.75

Nonwhite

0.21

0.14

2.5

1.5

Length of time in district

−0.006

0.01

0.27

Church attendance2

    More than once a week

−0.02

0.24

0.005

1.1

    Once a week

0.09

0.2

0.25

1.2

    Less than once a week but at least once a month

−0.03

0.23

0.01

1.1

    Less than once a month but at least once a year

0.09

0.29

0.1

1.2

    More than once a year

−0.06

0.43

0.02

1

Household income3

    $0–10,999

0.56

0.5

1.24

2

    $11,000–19,999

0.39

0.32

1.42

1.7

    $20,000–29,999

−0.59

0.24

6.08**

0.63

    $30,000–39,999

−0.24

0.26

0.89

0.89

    $40,000–49,999

−0.03

0.27

0.01

1

    $50,000–59,999

−0.45

0.27

2.85

0.72

    $60,000–69,999

0.5

0.39

1.67

1.9

N = 516

Likelihood ratio chi-square test statistic

34.8

19 df

0.01

Somers' D

0.33

Dependent variable is 1 indicating individual consistently gathered information on factors they said were important in selecting a school. A negative coefficient suggests individuals possessing characteristics were less likely to be consistent.

(1) Reference group for log odds ratio is parents in tuition based choice

(2) Reference group for log odds ratio are parents who are not church attendees

(3) Reference group for log odds ratio are respondents with income 〉 $70,000

(*) p. 〈 .05 ** p. 〈 .01

(p.94) differences between private school parents and other public school parents. All other individual characteristics of the parents are insignificant except for income; church attendance, race, and length of time in the district are not significantly related to sending mixed signals.23

The question remains whether an education market provides enough incentives for gathering information and sending clear, consistent signals to entrust accountability entirely to parent-consumers. The number of consumers who are informed and send consistent signals in the education marketplace varies; but in only one case is this variation related to the type of choice program. This is true of all our tests of the existence of informed and consistent consumers—knowledge of school characteristics, knowledge of principal, and clarity of signals—implying that the degree to which choice provides sufficient incentive for parents to gather information for holding schools accountable is limited.

Conclusion

School choice has the potential to empower parents to collect information about schools and select schools on the basis of the factors that are important to them. Choice thereby offers the opportunity for parents, rather than the government, to be ultimately responsible for holding schools accountable. Schools that meet parents' desires will succeed and the others will fail.

Achieving accountability of this nature requires that parents possess adequate information about the schools they chose. In addition, it requires the signals parents send when they select a school to be clear. We hypothesized that in choice systems, where there are few outside accountability mechanisms, parents would have greater incentive to gather accurate information and would be more consistent in the messages they deliver to the schools and to other parents. We expected to find more informed parent-consumers in traditional tuition-based private (p.95) schools than in the various public school choice systems because the private school system operates most like a true market.

Instead, we find choice systems differ only slightly in the extent to which they motivate parents to seek information. We are no more likely to find informed and consistent consumers in traditional tuition-based private school choice than in public school choice systems, with one exception—parents sending their children from Milwaukee to suburban public schools through an interdistrict integration program. In terms of general knowledge of a school we find that such individual characteristics of a parent as race, education, and church attendance matter more.

Therefore, it seems clear that structuring a school choice system on the pure competitive market theory that well-informed, knowledgeable parents and their resulting actions are all that is necessary to hold such schools accountable falls short in the reality of the Milwaukee experience. Private school parents, who are most comparable to voucher parents and who could be expected to have a deeper interest in their schools than public school parents for whom outside accountability mechanisms already exist, are no better informed or consistent in their behavior than their public school counterparts. At this point the market theory does not hold to its promise of accountability.

Notes:

(1.) An earlier version of this chapter appeared in Urban Affairs Review (July 2002).

(2.) John F. Witte was appointed in 1990 as the State of Wisconsin evaluator of the Milwaukee voucher program. He collected the data used by Manna and others (see http://dpls.dacc.wisc.edu/choice/).

(3.) Charter schools and open enrollment are relatively new in Milwaukee. The Wisconsin Charter School Law was created in 1993. From 1993–97 only one charter school existed in the city. In 1997 the state legislature authorized the City of Milwaukee, the University of Wisconsin-Milwaukee, and the Milwaukee Area Technical College to grant charters. During the time of our survey, 186 students were attending one of four charter schools in Milwaukee. Open enrollment started in Wisconsin in the 1998–99 school year. During the time of our survey, 119 Milwaukee students were participating in this program.

(4.) The statute defines “minority group pupil” as “a pupil who is Black American, a Native American, a Spanish-surnamed American or an Oriental American.” See Julie Mead's report: “Publicly Funded School Choice Options in Milwaukee: An Examination of the Legal Issues,” June 2000, Public Policy Forum, Milwaukee, for more information on this program.

(5.) The list of private school parents came from a list MPS maintains for transportation services. The list includes parents whose schools have worked out transportation agreements with MPS, and parents who have individually requested information from MPS. It is only a small sample, less than 10 percent, of those students who attend private schools in Milwaukee. This list included seventeen parents who identify themselves as having children who receive vouchers.

(6.) We surveyed the parent or guardian who was responsible for making the educational decisions in the household.

(7.) We were unable to receive from MPS phone numbers of any of the parents. For this reason, we used a different strategy to get a sample of phone numbers. We asked for and obtained mailing labels for 5,100 students attending or applied to attend suburban schools under the Chapter 220 program, 540 suburban students attending MPS under 220, 4,970 students attending MPS, and 2,122 students attending private school. Since we were conducting a phone survey, the mailing addresses were processed through a reverse directory to obtain telephone numbers. This process involved matching addresses with phone numbers located on a web-based data set. We were able to get phone numbers for 44 percent of the mailing labels. For a number of reasons, phone numbers will not be available for some addresses. First, unlisted numbers were not available. Second, in other cases there was no working phone connected with an address. Finally, the labels represent students, not households; due to the presence (p.196) of siblings approximately 38 percent of the addresses are duplicates. We contracted Lein/Spiegelhoff, Inc., a survey research company in Brookfield, Wisconsin, to interview the parent or guardian who makes the decisions about the education of the children living in the household. If the person was not available a callback was set up. Three callbacks were conducted. The completion rate for phone interviews was about 20 percent over the total sample. This means that for every five numbers, one interview was completed. In order to complete 678 interviews, approximately 4,700 phone numbers were needed. Approximately 22 percent, or 1,330 of the numbers, were not accurate for the choice groups. Of the 3,270 accurate phone numbers approximately 7 percent of the attempts resulted in refusal; other interviews were not completed because the interviewer reached an answering machine or voice mail, a busy signal, fax machine, a business number, or a privacy manager.

(8.) Given the small number of parents we interviewed who participated in the Suburban 220 program we interpret these results with caution. The final numbers in this choice group are useful only for descriptive purposes.

(9.) We also asked parents if they had knowledge about class size. But we could not examine accuracy of information for this variable because data on class size does not exist for any of these schools. The ratio of teachers to students does exist, but this measure is not a good substitute for class size (see Educational Series on Class Sizes, www.weac.org/sage/research/CLASSIZE.HTM).

(10.) Reading data for suburban school districts were obtained from the 1999 School Facts Book, published by the Wisconsin Taxpayers' Alliance. These scores include the percentage of students proficient or above on Wisconsin's Knowledge and Concepts Exam, given to fourth, eighth, and tenth graders. The scores are districtwide for Chapter 220 school districts. For MPS schools the data were taken from the 1998–99 MPS Accountability Report published by MPS Division of Research and Assessment. Private schools are not required to take these standardized exams and therefore are not included in this analysis. Data for public schools on percentage of school that is African American are taken from the same sources listed above. For private schools these data were taken from the schools themselves, many of whom report racial makeup on the Empowering Parents for Informed Choices in Education website. This information was also obtained from the 2000 Legislative Audit Report for schools in which MPCP students made up at least 90 percent of the student body. Finally, data on students receiving free or reduced-price lunch were taken from the Wisconsin DPI. This information is for both private and public schools.

(11.) Controlling for the individual level characteristics of parents, we examined logistic regression models for accuracy of information. None of the models approached statistical significance. Because these models are not significant, the fact that these questions were asked of half of the survey respondents, and the (p.197) lack of reading score data for private schools we do not include the models in this chapter.

(12.) We acknowledge that the error would be higher if we tested whether the respondent was correct.

(13.) We also examined these differences by grade level and found no significant differences.

(14.) Our sample was made up of 43 percent white, 39 percent African American, 8 percent Hispanic, 5 percent Asian, and 3 percent other. In addition, 3 percent of our survey respondents refused to answer. We ran the model using all five categories; there were no statistically significant differences between the nonwhite categories.

(15.) Our survey question on education provided us with five categories. Six percent of our total sample had less than a high school degree, 24 percent a high school degree, 36 percent some college/tech school, 33 percent graduated from college and/or had some postgraduate work. We ran this analysis with separate dummy variables for each grouping and did not find statistically significant differences for categories less than a college degree. For this reason the final model is a dummy variable.

(16.) The distribution of respondents across our income variable is as follows: 0−$10,999 = 5 percent, $11,000−$19,999 = 9 percent, $20,000−$29,999 = 15 percent, $30,000−$39,999 = 13 percent, $40,000−$49,999 = 13 percent, $50,000−$59,999 = 11 percent, $60,000−$69,999 = 10 percent, and 〉 $70,000 = 10 percent.

(17.) We asked, “How often do you go to a place of worship?” Our distribution of responses included 18 percent who go more than once a week, 35 percent who go at least once a week, 21 percent who go at least once a month, 13 percent who go at least once a year, and 10 percent who never go or go less than once a year.

(18.) Length of residence was included as an independent variable because parents who have resided in a district may know more about the schools their children will attend than parents who are new to the district.

(19.) Besides exploring the impact of the independent variables, it is instructive to examine several goodness of fit measures that the logistic regression procedure generates. In logistic regression, the analog of the global F test is a likelihood ratio chi-square test statistic, which is often referred to as the model chi square. We test H0, all the betas equal 0, against H1, at least one beta is not 0. This test statistic is 74 with 19 df significant (p. 〈 .0001) for both models. Thus we can conclude that at least one of the betas in the model is nonzero. The logistic regression technique also produces predictions of the actual dependent variable. The results indicate that the model explaining knowledge of the principal is correct 74 percent of the time. The model also includes a measure of how well the model reduces prediction error using the variables included in the (p.198) model versus when predicted by chance alone. We report the Somers' D statistic. We reduce error by 48 percent (see Damaris 1992 and Hosmer and Lemeshow 1989 for a further discussion of logit modeling).

(20.) Selectivity bias is a concern whenever the assignments to the treatment and control groups are not random (Barnow 1980: 43–45). One method to examine the possible impact of selectivity bias is to run a two-staged least squares (2SLS) regression model. This method has been used when the factors related to selection into a group are known. It is also useful when examining one treatment group and one control group (Barnow et al. 1980; Heckman 1980; and see Schneider, Teske, and Marshall 2000 and Witte 2000 for use of 2SLS in choice studies). This method is not necessary for the models examined in this book. First, parents are selecting themselves into separate programs, not one. Second, all the individuals studied are active choosers. As Schneider, Teske, and Marshall found, when all the parents in the program must choose, nonrandom selfselection does not apply (2000: 83). Third, the final model is not an OLS, but a logistic regression model. The two-stage regression procedure when modeled for this analysis produced estimates that were substantively meaningless. For these reasons we report the final logistic model. This does not mean that selectivity bias has not been adequately addressed. What it does mean is that the best way to minimize the impact in this study was to ensure that factors related to being an informed consumer were included as control variables. As Barnow (1980) notes, the most common way to deal with selectivity bias is to make a diligent effort to include a large number of independent variables that control for selection bias. Every attempt was made to include survey questions that would ensure no variables related to the dependent variable were excluded. In this way we feel confident an omitted variable would not account for the differences among these programs.

(21.) Because of the difficulty in interpreting logit coefficients, we report the log odds ratio.

(22.) The chi-square test statistic is 34.8 with 19df and is significant (p. 〈 .01). The model is correct 66 percent of the time in predicting a mixed signal. The Somers' D statistic suggests that we reduce prediction error by 33 percent.