The One-Eyed Parent May Be Enough

MarkSchneider;PaulTeske;MelissaMarshall;ChristineRoch

AmericanJournalofPoliticalScience,Vol.42,No.3.(Jul.,1998),pp.769-793.

StableURL:

http://links.jstor.org/sici?sici=0092-5853%28199807%2942%3A3%3C769%3ASFSITL%3E2.0.CO%3B2-0AmericanJournalofPoliticalScienceiscurrentlypublishedbyMidwestPoliticalScienceAssociation.

YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat

http://www.jstor.org/about/terms.html.JSTOR'sTermsandConditionsofUseprovides,inpart,thatunlessyouhaveobtainedpriorpermission,youmaynotdownloadanentireissueofajournalormultiplecopiesofarticles,andyoumayusecontentintheJSTORarchiveonlyforyourpersonal,non-commercialuse.

Pleasecontactthepublisherregardinganyfurtheruseofthiswork.Publishercontactinformationmaybeobtainedathttp://www.jstor.org/journals/mpsa.html.

EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprintedpageofsuchtransmission.

TheJSTORArchiveisatrusteddigitalrepositoryprovidingforlong-termpreservationandaccesstoleadingacademic

journalsandscholarlyliteraturefromaroundtheworld.TheArchiveissupportedbylibraries,scholarlysocieties,publishers,andfoundations.ItisaninitiativeofJSTOR,anot-for-profitorganizationwithamissiontohelpthescholarlycommunitytakeadvantageofadvancesintechnology.FormoreinformationregardingJSTOR,pleasecontactsupport@jstor.org.

http://www.jstor.orgThuOct2513:10:192007

Shopping for Schools: 󰀁In the Land of the Blind, 󰀁The One-Eyed Parent May Be Enough* 󰀁Mark Schneider, SUNX Stony Brook Paul Teske, SUNX Stony Brook Melissa Marschall, University of South Carolina Christine Roch, SUNX Stony Brook Theory: Market-like reforms, such as school choice, can work effectively in spite of the 󰀁low levels of information commonly found among citizen/consumers. 󰀁Hypothesis: We hypothesize that (1) parental knowledge of school characteristics is a 󰀁function of ability, incentives, and whether parents believe a particular school attribute to be important; (2) parents will select schools for their children that rate high on the dimen- sions that they value; (3) the \"marginal consumer\" will be more knowledgeable about schools than other parents; and (4) marginal consumers will have more accurate informa- tion on dimensions that they value and are more likely to select schools for their children that rate high on these dimensions. 󰀁Methods: Multiple regression is used to predict parents' knowledge of reading test scores, 󰀁racial composition of the school, and number of violent incidents. Multiple regression is 󰀁also used to determine the extent to which consumers have enrolled their children in 󰀁schools that are high on the dimensions about which they care. 󰀁Results: We find that on average low-income parents have very little accurate informa- 󰀁tion about objective conditions in the schools. However, even in the absence of such ob- 󰀁jective knowledge there is evidence of a matching process in which children are enrolled 󰀁in schools that are higher on the dimensions of education that their parents think are im- 󰀁portant. We then shift our analysis away from the behavior of the \"average\" parent and 󰀁identify a subset of parents who are in fact informed about the conditions of the schools. 󰀁We demonstrate that there is a tighter match between what these parents want and the 󰀁conditions of the schools in which their children are enrolled. 󰀁Shopping for Schools Empirical research has shown consistently that citizens know very little about politics and public policies. Yet, many contemporary policy reforms seek to introduce market-like mechanisms that give \"citizen/consumers\" greater choice in the public services they consume. Theoretically, introducing *The research in this paper was supported by the National Science Foundation, Number SBR94070. Paul Teske thanks the National Academy of Education's Spencer Foundation Post- Doctoral Fellowship for support on this project. Mark Schneider thanks the Russell Sage Founda- tion, where he was a visiting scholar during 1997-1998. Replication note: The data to replicate this study are available from the lead author. American Journal of Political Science, Vol. 42, No. 3, July 1998, Pp. 769-793 O 1998 by the Board of Regents for the University of Wisconsin System 770 M. Schneidel; f! Teske, M. Marschall, and C. Roch market-like reforms will increase efficiency in two ways: first, there will be a better match between what citizenlconsumers want and what they get. Sec- ond, the suppliers of public goods will be under competitive pressure to in- crease the quality of their product and control their costs. Many studies of such market-like reforms have focused on how the number of suppliers (and the competition between them) can be increased. The success of these re- forms and the returns they promise have a demand-side element to them, spe- cifically, the ability of citizen/consumers to make intelligent choices about the expanded set of options they face. This demand-side element leads us to a fundamental question: can citi- zenlconsumers make good choices given the levels of information they have? In this article, we explore this question by investigating citizenlcon- sumer behavior in the choice of the schools their children attend-a choice that is highly salient for many parents-and a policy domain in which the range of choice individuals have is expanding rapidly. While most previous educational reforms focused on internal curricu- lum and teaching methods, today's reforms center on issues of governance. Perhaps, the single most common ingredient in current proposals to reform schools is the expansion of the enrollment options available to parents (Anhalt et al. 1995; Chubb and Moe 1990; Henig 1994; Ravitch and Viteritti 1996; Smith and Meier 1995). These new reforms are based on a belief that education cannot be improved unless power is shifted toward parents, changing the way in which educational policy decisions are made. Propo- nents of school choice identify both supply- and demand-side effects. On the supply-side, proponents argue that choice will increase competi- tive pressure on schools to deliver higher quality education. On the demand- side, proponents argue that by giving parents more control over the schools their children attend, choice creates incentives for parents to become more informed about the schools. Further, because choice is designed to increase the range of schooling options available, parents also have incentives to learn more, so that they can choose the school that best matches the needs of their children. In this article, we examine how much information inner-city parents have about schools. We demonstrate that on average these parents have very little accurate information about the objective conditions in the schools. We then show that even in the absence of such objective knowledge there is evi- dence of a matching process in which children are enrolled in schools that are high on the dimensions of education that their parents think are impor- tant. This creates a puzzle: how can such matching take place in the absence of information? To solve this puzzle, we shift our analysis away from the behavior of the \"average\" parent to identify a subset of parents who are in fact informed about the conditions of the schools. We demonstrate that there is a tighter SHOPPING FOR SCHOOLS 771 match between what these parents want and the conditions of the schools in which their children are enrolled. We then explore the efficiency and equity implications of a market for schools driven by these informed parents, who we argue play the role of the \"marginal consumer\" in the choice process. Citizen Information Levels Beginning at least as far back as Lazarsfeld, Berelson, and Goudet (1944) political scientists have shown that most citizens know very little about politics. The evidence is so consistent that Bartels (1996, 194) de- scribes the current consensus about political information levels in no uncer- tain terms: \"The political ignorance of the American voter is one of the best documented data of modern political science.\" Since electoral politics are distant and removed from the daily world of most citizens, the knowledge of candidates might be lower than knowledge about specific policies of government that affect them. Exploring this possi- bility, Kuklinski et al. (1996,4) note that previous research on citizen knowl- edge has been \"overwhelmingly\" weighted toward the study of processes, institutional structures, and the names of current politicians, and that \"re- searchers have rarely asked for factual knowledge about policy.\" Yet even following the accident at the Three Mile Island nuclear power plant, Kuklinski himself found that few citizens were well-informed about basic issues of nuclear power (Kuklinski, Metlay, and Kay 1982). Moreover, in re- cent work examining citizen knowledge of welfare, Kuklinski found a citi- zenry woefully misinformed about basic aspects of that policy (Kuklinski et al. 1996). More generally, Zaller dismisses the suggestion that citizens are likely to learn more about matters that are important to them, arguing that the \"tendency appears not to be very great or very widespread\" (1992, 18. See also Price and Zaller 1993; but see Delli Carpini and Keeter 1996). One reason for such low citizen information levels is a simple benefit1 cost calculation: while accurate political decisions may have beneficial con- sequences for individual citizens, the search for information upon which to base such decisions is costly. According to Simon (1986, S210-S21 I), cog- nitive effort is a scarce resource, and the knowledge and computational power of the decision-maker are always limited. All decision making in- volves both benefits (decision accuracy) and costs (cognitive effort), and in- dividual decision-makers can be viewed as \"cognitive misers\" who seek to minimize the costs involved in cognitive effort while trying to maximize the rewards of accurate decisions (see, for example, Fiske and Taylor 1991; Sniderman, Brody, and Tetlock 1991 ;or Lodge and Stroh 1993). Proponents of expanded choice in markets for public goods argue that re- form will fundamentally rearrange the benefits and costs that citizen/consum- ers face and give them incentives to become more informed about public goods. This argument is central to debates over school choice. Most notably, 772 M. Schneidei; I? Teske, M. Marschall, and C. Roch Chubb and Moe argue that the frequently voiced criticism that school choice programs will founder on high levels of parental ignorance is wrong. Instead, they argue that present ignorance is a function of present institutional ar- rangements: \"In a system where virtually all the important choices are the re- sponsibility of others, parents have little incentive to be informed or involved. In a market-based system, much of the responsibility would be shifted to par- ents (their choices would have consequences for their children's education), and their incentives to become informed and involved would be dramatically different\" (1990, 5). Coons and Sugarman (1978, 188) make the same point more colorfully: \"In a system with no options, ignorance might be bliss. In a system based on choice, ignorance is ruin.\" Thus, according to proponents of choice, by changing the incentives to gather information, parents' benefitlcost calculations will be positively af- fected: given choice, there will be increased benefits associated with being informed; hence, parents will be more likely to search for information. Research Design Because critics of choice have been most concerned about the ability of low-income parents to gather information about schools and make appropri- ate choices, we focus our analysis on this group. We study the behavior of parents with children enrolled in public elementary schools in two inner-city school districts in Manhattan, New York. Using telephone interviews con- ducted with a random sample of parents, we analyze the knowledge and choice behavior of parents whose children attend schools in these districts. The first district, Community School District 4, is located in East Harlem and has a long history of alternative schools and school choice. The district main- tains twenty-nine elementary schools, of which eleven are alternative schools. It also maintains eighteen intermediate (junior high) schoo1s.l The other district we study, District 1,is located on Manhattan's Lower East Side. Compared to District 4, District 1 was much later in starting school choice, and its program is neither as developed nor as successful. District 1 has fif- teen elementary schools, including four alternative elementary schools and six intermediate schools. Since these districts are relatively compact and ser- viced by extensive public transportation, parents can visit any or all of the schools in their district if they so choose. Because there are fundamental differences in the choice mechanisms at the elementary school and intermediate school levels, before looking at par- ent choice behavior, we describe in more detail the context of choice in these two districts. At the elementary school level in Districts 1 and 4, and similar to the choice programs implemented in a number of other cities (including Alum Rock, Milwaukee, San Antonio, and St. Louis), alternative schools 'In New York City, junior high schools are usually referred to as intermediate schools. SHOPPING FOR SCHOOLS 773 exist alongside traditional neighborhood schools. In these types of \"option demand choice systems (Elmore 1991), parents must choose to send their child to one of the schools of choice, otherwise the child is assigned to a \"default\" neighborhood school. Parents with children in grades seven and eight are faced with a different type of choice. In District 1,parents have ba- sically no choice at the intermediate school level, since residential location determines which intermediate school a student attends. In contrast, the \"universal choice\" program in place at the intermediate school level in Dis- trict 4 requires all parents to choose a school for their children. These differences in choice mechanisms have important implications for both parents as consumers of education and schools as suppliers (Elmore 1991). We believe that the two systems of choice must be analyzed sepa- rately, even when they are operating in the same school district. In this ar- ticle, we focus on the behavior of parents with children in elementary sch~ols.~In addition to the fundamental theoretical justification for our de- cision based on the difference in choice protocols, our focus on these parents is also dictated by a very practical concern: our sample reflects the prepon- derance of schools and grades at the elementary school level. Thus, while we interviewed more than 500 parents of public elementary school students in these two districts, only 121 parents of junior high school students were in 2The larger project of which this is a part also studied parental behavior in Montclair, New Jer- sey, a suburb of New York with a long history of universal choice, and Momstown, New Jersey, a suburb with strict residential catchment zones. However, we do not include these respondents in our analysis of matching because the environment of choice in Montclair is radically different from the environment of choice in New York. First, the number of options is much fewer than in either of the New York districts-there are only two middle schools in Montclair and only five schools at the grade 3-5 level. In addition, while alternative schools in New York emphasize unique themes (for example, bilingual education, performing arts) the Montclair schools largely do not. Moreover, Montclair has a strong policy commitment to minimize differences in the performance and the de- mographic composition of its schools. For example, across the two middle schools, the percent of students reading at grade level is within 1 percentage point and there is only about a 5 percentage point difference in the percent black (the largest racial minority group in Montclair). Across the five schools containing grades 3-5, the range in percent black is from 33% to 44% and the range in read- ing scores is from 80 to 95%. Moreover, even though there is choice, there is a strong tendency to have children follow established patterns from feeder schools at the pre-K through 2 level into the next level of school and then into intermediate schools. Thus the market for schools, the range of op- tions, and the extent of choice in Montclair (as well as its demographic composition) are all radically different than found in New York. And finally very few parents in New Jersey (only 5%) ranked di- versity or safety as important concerns. Despite these major environmental differences, when we replicated the New York analysis that we report later in this article for reading scores in New Jersey, the results are remarkably similar. Parents whose children are in grades 3-5 in Montclair, the level at which choice is the most meaningful, and who say high scores are important are over 6 points closer to the real reading scores of their schools than other New Jersey parents, and they place their children in schools whose performance on reading tests are .67 standard deviations higher than other schools. Thus even in this more limited choice environment, there is evidence of a matching process by parents who actively choose their children's schools. 774 M. Schneidel; f! Teske, M. Marschall, and C. Rock our sample. Given the distribution of missing values in the various indicators we use in this study, the replication of the results we report for elementary school parents using intermediate school parents would be based on unac- ceptably small numbers of observations-as few as forty-one parents in the analysis of school safety and at most eighty-two in the othem3 Faced with this data constraint, we are better able to examine the issues of information and choice of public goods in the option demand system. In addition, it is the option demand system around which much of the debate surrounding school choice has been based. Thus, we focus our analysis ex- clusively on parents whose children attend elementary school^.^ an attempt to explore this issue further while maintaining a sufficient sample size, we com- bined the two sets of parents into a single sample and introduced dummy variables for level of schooling (intermediate school vs. elementary school). We also introduced interaction terms to ex- plore further the questions motivating our research. The results for this pooled analysis are much more complicated because of the additional interaction terms, but are virtually identical to the results we report for the elementary school sample alone. We believe, however, that the right way to study the issue is to have a larger sample of parents in a universal choice system. The results of the pooled analysis are available from the authors. 4We contracted Polimetrics Laboratory for Political and Social Research, a survey research firm at Ohio State University, to interview 400 residents in each district. To start, Polimetrics identi- fied the zip codes in each of the four school districts. All listed telephone numbers for each zip code were identified. From this, a list was developed using random generation of the last two digits of the appropriate telephone exchanges, so that unlisted numbers were included as well. All known busi- ness telephone numbers were removed as they were not eligible to be interviewed. Then, they took a random sample of the remaining numbers. To be eligible to be interviewed, respondents needed to: live within the school district, have children between grades K-8, be the adult responsible for decisions affecting that child's education, and identify the school their child attended (which could be either a private school or a district pub- lic school). To randomize, respondents were asked to answer the school-specific questions based on the child in grades K-8 whose birthday came next on the calendar. For the purpose of this analysis, we are only examining K-6 public school children in the New York City districts. The actual interviews were conducted in the period March-June of 1995. The interviewers were given extensive training, and some interviews were done in Spanish. Interviews were moni- tored randomly and, to insure validity, 15% of all completed interviews were verified with respon- dents by the supervisors. The goal was to obtain 400 completed interviews in each of the districts. The following table shows the call dispositions in each district. Disposition of Survey Telephone Calls District 4 Completed Refusals No final disposition Non-household Ineligible 400 District 1 401 113 225 5,237 5,722 522 1,2 17,883 13,469 In addition to these New York City districts, we also interviewed parents in two New Jersey suburbs, Montclair and Morristown, but these results are not reported here-see footnote 2. An ap- pendix describing the sample and giving more details about the survey is available upon request. SHOPPING FOR SCHOOLS 775 In addition to the parental survey data, we collected objective indica- tors of performance for each school and program in the two school dis- tricts. With these data, we have a means of determining not only the extent to which parents have accurate knowledge of school performance, but also whether parental choice behavior is congruent with their information about school performance. In studying school choice, we must address a fundamental fact: educa- tion is a multifaceted good and parents differ on the attributes of schools they feel are important for their children (see Delpit 1995; Hirsch 1996; Schneider et al. 1998). Cognitive resources are limited and even the most in- volved parents will find it difficult if not impossible to learn about every di- mension of the schools in which they can enroll their children. Thus levels of information and choice behavior must be weighted by the specific attributes of schools parents value-that is, if a parent feels that academic perfor- mance is important and multiculturalism is not, the parent's knowledge should be more accurate on the first dimension than on the second; and the school helshe chooses for hislher child should be higher on the first dimen- sion as well. In our survey, we presented parents with a list of school attributes about which previous research has demonstrated that parents care.5 These specific attributes fall into three domains. The first domain consists of attributes asso- ciated with the fundamental educational \"product\" of the schools-math and reading scores, the quality of teachers, and class size. The second domain re- flects the reality of local schools in inner cities: safety and discipline. A third set of attributes focuses on the make-up of the student population in terms of race and income. In the analysis that follows we explore one indicator from 5During the survey, interviewers read the following statements to parents: \"Different parents believe that different things are important for their child's education. We are interested in knowing which things you think are important. In this next section I will read to you a list of some of the things that parents believe are important in a school, and I'd like to know what you believe to be most important to your child's education.\" They were then told: \"From the following list of qualities about schools, first tell me which is the most important to you\" and were read a list of eleven at- tributes: quality of teachers and staff, a student body that is mostly the same races as [child's name], values of the school, a racially diverse student body, safety, economic background of students, loca- tion, high math or reading scores, special programs, discipline code, and class size.\" After selecting an attribute as most important, they were then read the list again, this time without the previously named attribute, and again they were asked to name the attribute they thought most important. This procedure was repeated four times and the order in which the list was presented to the respondents was randomized to control for any order effects. Across these four queries, teachers were named important by 77% of the respondents, high scores by 55%,class size 31%, and special programs 25%. Not surprisingly given the inner-city mi- lieu of our study, safety (70%) and discipline (44%)were also frequently cited. In contrast, diversity (16%),same race, and economic makeup of student body were not frequently cited (both less than 5%). Values was chosen by 33% of the parents and location by 22%. 776 M. Schneidel; J! Teske, M. Marschall, and C. Roch each of these clusters: reading scores, the number of incidents in a school, and the raciayethnic diversity of the student population. In analyzing paren- tal knowledge, preferences, and choice behavior, we combine the data from the two school districts, using several techniques to control for the unique and unmeasured characteristics of each distri~t.~ We look first at levels of knowledge parents have about each of these characteristics and then we look at parents' choice of schools. Our research question is evident: are parents placing their children in schools that reflect their preferences over these attributes? How Accurate Are Parents about the Conditions of Schools? Table 1 presents some simple descriptive data about the objective condi- tions in the schools in our study and the average \"error\" between the objec- tive conditions found in a child's school and the estimate of that condition given by the parent. Clearly, the average distance between estimates and ac- tual conditions can be quite large. Note for example that the average distance between the percent of students reading at grade level estimated by parents and the actual performance is 25 points. Similarly, parents were \"off' an av- erage of more than 15% in their estimate of the racial composition of their child's scho01.~ In the next stage of analysis, we examine whether or not characteristics of parents affect accuracy. We model accuracy, measured by the distance be- tween the estimate and the true condition, as a function of a set of parental characteristics that have been found to affect school knowledge: Accuracy =f (race, SES, length of residence in district, church attendance, district, importance) 6As noted below, we employ a dummy variable for the district in the knowledge analyses, while utilizing a district z-score in the matching analyses. In the latter type of analysis, an additional dummy variable would not have any clear meaning. 7We experimented with two other measures of accuracy. Delli Carpini and Keeter (1996)argue that political scientists have placed too much emphasis on pinpoint accuracy, without giving citizens enough credit for being \"near\" correct. To investigate how \"near correct\" parents were, we took the score of each school on each measure and created an interval of +I-20% around the actual score. We then counted a parent as correct if her estimate of her child's school performance fell within that in- terval. Using this technique, less than 5% of the parents were correct concerning the number of inci- dents; less than 25% were correct in their estimate of reading scores and percent black; and less than 45% were accurate about the percent Hispanic. There is an obvious relationship here between the mean and the percent near correct-as the mean increases, so does the size of the interval. We also measured the ability of parents to rank their schools relative to other schools in the district-and again the number of parents who could accurately place their child's school was low. SHOPPING FOR SCHOOLS 777 󰀁Table 1. Objective Conditions and Aggregate Levels of Parent Information No. of Schools No. of Parents Actual Mean Range of Parents' Estimates Average Distance Score* Indicator % Reading at grade level % Hispanic % Black Number of incidents ** 68 67 30 408 440 468 436 31% 62% 30% 3.6 13-93% 17-100% 1-83% 1-1 1 25% 17% 15% 7 *Distance is the absolute distance between the estimate of a condition given by a parent and the ob- jective measure of that condition for that child's school or program. **The number of schools/programs for which data are available. Incident data is available only for school buildings, while other indicators are available at the school and the program level. Source: Objective data: New York City Board of Education School Report Cards. Parent estimates: survey of parents in two New York City community school districts. where: Race is a series of dummy variables indicating the respondent's self- reported racial identity-Black, Asian, Hispanic, and White (the excluded categ~ry).~ SES is represented by the years of schooling of the re~pondent.~ Note that the race and education variables allow us to address one of the major issues in the debate over school choice: are there systematic differ- ences in the knowledge and behavior of parents who are less educated and are from minority racial groups that distinguish them from better educated parents and parents who are White? Length of residence is measured by the number of years the respondent has lived in the school district. Public goods, including education, have both a search and experience dimension-some knowledge can be gained by searching for information prior to the purchase decision, but other knowl- edge can be gained only through experience with the good (Teske et al. 1993; on this distinction in general see Weimer and Vining 1992 and Wilde 1981). 8There were twelve respondents who did not fall into any of these categories and are not in- cluded in the analysis. 9We also ran the models with the self-reported income level of the respondent and the results are identical to the results we report using years of schooling only. We chose to leave income out of the analysis because we lose a large number of observations when we include income in our mod- els-over fifty respondents refused to answer the income question. In addition, income and years of schooling are highly correlated, and most studies have shown that in studying parental behavior in school related issues, education is a more important indicator of SES than is income (e.g., Martinez et al. 1995). 778 M. Schneidei; f! Teske, M. Maischall, and C. Roch Church attendance is included as a general measure of involvement with the social life of the community and has been found to be related to involve- ment with a range of school-based activities (Schneider et al. 1997a). The last two measures in the model are dummy variables. First, we in- clude a dummy variable for district to account for any gross differences be- tween behavior in the districts, net of the specific conditions we measure (District 4 = 1). Second, given the multidimensional nature of education, we include a dummy variable indicating whether or not the parent rated the par- ticular measure of school performance important. We begin with the hypothesis that parents who believe a particular di- mension is important should be more accurate on that dimension than par- ents who do not rate the dimension important. To test this hypothesis, we regressed the distance measure for each of our indicators of school perfor- mance against these characteristics. The results in Table 2 indicate that er- ror rates do not vary significantly with the characteristics of parents that we measure, including the importance they assign to the issue. Indeed, very few of the independent variables have any statistically significant effect on error rates and none of the four equations reach acceptable levels of statis- tical significance. lo Matching Children with Schools In the study of elections, some scholars recently have argued that, de- spite low levels of information, citizens can still get \"enough\" information to make good choices (see, for example, Althaus 1995; Iyengar 19; Lupia 1992, 1994; Popkin 1991 ; Sniderman, Brody, and Tetlock 1991; Zaller 1992). Lupia and McCubbins (1998) have suggested that rather than being overpowered by complex problems, people are actually quite good at ob- serving the world around them and estimating the consequences of a wide range of actions. According to Lupia and McCubbins, while citizens rarely demonstrate detailed \"encyclopedic knowledge\" about public issues, they often demonstrate \"ability knowledgeH-which they define as the capacity of a citizen to identify the dimensions of choice needed to inform the actual behavior helshe needs to undertake to achieve desired goals. We have demonstrated that parents do not have encyclopedic informa- tion about school performance. However, can they still make reasonable de- cisions? Do parents enroll their children in schools that are high on the di- mensions of education they value? 'Owe conducted similar analyses using the two other measures of accuracy (near correct and relative) described in note 7. The results using these different measures were identical to the results we report using the distance measure. SHOPPING FOR SCHOOLS 779 Table 2. Do Parents have Objective Knowledge of School Conditions? Accuracy as a Function of Parental Characteristics Reading Asian Black Hispanic Years of schooling Length of residence Church attendance Is issue important? District 4 Constant Incidents Black percent Hispanic percent 2.46 7.60 -8.81* 3.75 -7.40* 3.46 -.08 .28 .0 1 .13 -.08 .43 -3.34 1. 2.23 1.92 34.4* 5.52 .34 2.40 1.21 1.39 .87 1.26 -.I5 .10 -.02 .04 -.00 .15 -.48 .68 .84 .69 7.53* 2.08 -3.82 5.36 -.22 2.83 3.39 2.63 .OO .21 -.I7 .09 -.23 .30 -1.05 1.96 -1.70 1.36 17.17* 3.95 -6.29 5.82 -6.62* 2.97 -1.71 2.73 .09 .21 -.04 .10 -.04 .32 -1.78 2.13 -1.3 1 1.43 20.59* 4.17 N.of cases F statistic Probability Adj. R~ *p< .05. The dependent variable is the distance between the estimate of each condition and the actual objec- tive condition in the school. The number in the first line of each entry is the regression coefficient, while the number in the second line is the standard error of the estimate. Source: Survey of parents in two New York City school districts. To investigate the answers to these questions, we look at the perfor- mance of the schools in which children are enrolled. Our analysis here is aimed at assessing the degree of the match between what parents value in education and the performance of the schools on that dimension. In this stage of our analysis, we use as the dependent variable the per- formance of a child's school relative to other schools in the district. Our procedure here is straightforward. We collected information on reading scores, ethnic composition, and the number of incidents from each school in each district. We then converted each school's performance into a z-score, using the mean and standard deviation of the performance of all 780 M. Schneidel; f! Teske, M. Marschall, and C. Roch schools in the district on that specific measure. After computing these z- scores within each district, we combined the observations from both dis- tricts into a single data set, allowing the analysis of comparative perfor- mance across different locales (on the use of z-scores for comparative analysis see, e.g., Schneider 19). Recall we are focusing on three objective measures of school perfor- mance-one measure in each of the domains of education with which we are concerned. We measure test scores as the percentage of students in the school who are reading at or above grade level, as measured by the New York City Chancellor's Achievement Test. For safety and discipline, we use the number of incidents in the school, as reported by the New York City Board of Education. Finally, as a measure of diversity we use the ratio of the percent of the student population that is White to the percent Hispanic. All data are for the 1994-95 school year, the period during which our survey was administered. Note that the reading scores and diversity measure are available for both neighborhood schools and for alternative schools and pro- grams, while the incident data are available only at the level of the school building. Since a single building may house several alternative programs or combine both a traditional neighborhood school and one or more alternative programs, as will become evident, this presents a measurement problem that affects our analysis. The results in the three analyses presented in Table 3 all point in the same direction-even though levels of objective information held by parents are low, their actual choice of schools reflects theirpreferences in education. For example, parents who say that high scores are important enroll their children in schools that are more than a quarter of a standard deviation above the dis- trict mean in reading scores. similarly, parents who value safety and those who value diversity also enroll their children in schools that are significantly higher than the average schools in their district on these dimensions.12 This raises an intriguing puzzle: if parents in general do not have accu- rate information about these dimensions, how can the matching process demonstrated in Table 3 take place? \"Because the dependent variable is standardized by district, we no longer include a dummy variable for district location on the right-hand side of the equation-the standardization procedure corrects for the conditions that the dummy variables would otherwise control. 12Note that the R2 statistic for incidents is much lower than for either of the other two dimen- sions. We believe that this is a function of the level of reporting for this indicator, as well as a mea- surement problem: incident data are available for school buildings only and not available at the pro- gram level. We have mapped the school building data onto the programs by noting which buildings the programs are in. However, because most alternative programs are in buildings that also house neighborhood schools there is no way to compare alternative schools to neighborhood schools or to correlate this indicator with other program indicators. SHOPPING FOR SCHOOLS 781 Table 3. Do Parents Sort Themselves into the \"Right\" Schools? Reading Asian Black Hispanic Years of schooling Length of residence Church attendance Is issue important? Accuracy(**) Accuracy 2 Constant Incidents Diversity N. of cases F statistic Probability Adi. R~ *p < .05. **Accuracy: The distance measure between the actual score and the score estimated by the respon- dent. In the diversity equation, there are two measures of knowledge: one the distance score for the percent black (Accuracy), the other for the percent Hispanic (Accuracy 2). The dependent variable is the z-score of each school on the specific indicator of performance. The number in the first line of each entry is the regression coefficient, while the number in the second line is the standard error of the estimate. Source: Survey of parents in two New York City school districts. One possibility is that parents have developed a set of shortcuts to in- formation that allow them to identify schools high on the dimensions that they value without actually having detailed information about performance. For instance, parents may rely on the judgments of friends, family, and co- workers about the quality of schools when choosing. Or, they may rely on cues about the performance of schools provided by the news media. Rather than rely on the importance of shortcuts such as these, however, we pose an alternative explanation. 782 M. Schneidel; f! Teske, M. Marschall, and C. Roch We demonstrate that the same puzzle we have just documented in school choice-that of matching preferences in the face of low levels of in- formation-has been posed in the study of many private goods markets. Most research into this phenomenon has demonstrated that the solution to the puzzle lies in a small group of consumers who actually gather informa- tion about products. Thus, while on average, information about products is low, a small group of buyers in markets tend to be more informed. More im- portantly, these studies also show that this small percentage of buyers can effectively drive a market toward a competitive outcome.13 Paralleling these findings based on the study of private sector markets, we argue that the matching process we have demonstrated in Table 3 is driven by the behavior of a subset of parents. We go further and argue that for the market for schools to function effectively, only a subset of parents need be informed about the various packages of goods and services their schools offer. l4 Shortcuts to Decisions We recognize that our position emphasizing the role of a subset of par- ents differs from a current thrust in political science that seeks to solve the problem of how people make decisions in the face of low levels of informa- tion by emphasizing the importance of cues as shortcuts to information. For example, in his discussion of the various shortcuts that voters use when evaluating information, Popkin argues that: \"Voters rely on the opinions of others as a shortcut in evaluating the information they have, because even when they do know about an issue, they are unaware of many relations between government and their lives\" (199 1, 44). Similarly, Lupia (1994) demonstrated the importance of relying on friends, coworkers, and groups as cue givers in a referendum on insurance reform in California. Lupia and McCubbins (1998) argue that individuals develop strategies to screen out important information from the vast flows of information to which they are exposed. In particular, if an individual listens to the right people, then reasoned decision-making does not require extensive (\"encyclopedic\") knowledge. Thus, we could explain the match in preferences and enrollment in the absence of encyclopedic information by demonstrating that parents rely on 13For example only 30%of buyers of nondurable products shop at more than one store before making their purchases (Katona and Mueller 1955); for larger items, Claxton, Fry and Portis (1974) found that only 5% of furniture buyers and 8% of appliance buyers gathered extensive information. I4Note that parents cannot maximize on all dimensions by choosing the same schools. In Dis- trict 1, the correlations between diversity, safety, and reading scores are not significant. In District 4, reading scores and the number of incidents are (negatively) correlated, while the other correlations are not. SHOPPING FOR SCHOOLS 783 cues provided by others when choosing a school. There is little evidence, however, that the majority of inner-city residents engage in extensive inter- personal communications about schools. For example, we asked respon- dents to name up to three people with whom they talked about schools. In this set of inner-city parents, the modal number of educational discussants in New York was 0 and the mean just over 1. Moreover, not only are these net- works limited in size, most of these parents are not linked to high quality sources of information. In fact, patterns of discussion about schools are so stratified by class and segregated by race that Schneider et al. (1997b) called them \"networks to nowhere.\" Conceivably, parents could rely on the evaluation of schools that appear in the media. However, such evaluations appear infrequently at best. We did an extensive search of newspapers to find stories about schools in the dis- tricts that we study. We found less than a dozen news stories in the last five years, and many of these were about the turmoil on District 1's community school board. Moreover, television news programs are even less likely to carry coverage of the performance of individual elementary schools. Thus, in the next section we develop an explanation for matching in the absence of general knowledge that is parsimonious and congruent with other work on the behavior of consumers in markets for private goods. We then show that the match found in Table 3 may in fact be a function of the behav- ior of these more highly informed shoppers for schools. The \"Average\" Consumer versus the \"Marginal\" Consumer Studies of private sector markets show that in many competitive markets only a subset of consumers are likely to gather information about their pur- chases, but that these consumers can drive a market towards a competitive outcome (Claxton, Fry, and Portis 1974; Katona and Mueller 1955; Newman and Staelin 1972;Thorelli and Engledow 1980). Rhoads (1985, 144) argues that in many markets these informed or marginal consumers are the most careful and well-informed shoppers and that their actions generate: \"com- petitive pressures that help keep prices reasonable for less-informed, non- searching consumers as well.\" Schwartz and Wilde (1979, 638) argue that \"the conventional analysis asks the wrong question. Rather than asking whether an idealized individual is sufficiently informed to maximize his own utility, the appropriate normative inquiry is whether competition among firms [here, read schools] for particular groups of searchers is, in any given market, sufficient to generate optimal prices and terms for all consumers\" [1979, 638. Emphasis added]. Thus, competitive markets require at least some consumers to be informed enough to pressure producers to deliver ser- vices efficiently. Wilde and Schwartz (1979, 55 1) also note that \"the likeli- hood of competitive equilibria obtaining varies directly with the number of 784 M. Schneidel; I? Teske, M. Marschall, and C. Roch consumers who visit more than one firm [again, read schools] and with the number of firms such 'comparison shoppers' visit.\" Empirical studies of private markets do find a group of consumers that search for more information than the average consumer. These consumers are more interested in and \"involved with\" the product (Claxton, Fry, and Portis 1974; Katona and Mueller 1955; Newman and Staelin 1972; Slama and Tashchian 1985; Wilde and Schwartz 1979, 543). Two sets of studies have focused on the critical importance of these informed consumers. First, Thorelli and Engledow (1980) identify \"Information Seekers\" who make up 10-20% of the population and help police the market by their comparative shopping. Second, Feick and Price (1987) labeled the upper third of informa- tion seekers \"market mavens,\" and Slama and Williams (1990) confirm that market mavens provide comparative product information to others for many products and services. Teske et al. (1993) extended this work from the private market to the public goods market, identifying \"marginal consumers\" who are informed about schools and who exert pressure on local schools to be more efficient.15 (See the exchange between Lowery et al. 1995 and Teske et al. 1995; also Dowding, John and Biggs 1994; John, Dowding, and Biggs 1995). While proponents of school choice, like Chubb and Moe (1990), Coons and Sugarman (1978), and others, imply that the full benefits of choice at the systemic level will be generated by high levels of information across all par-ents, we argue that the response of a smaller group of involved, motivated, and informed parents may be sufficient to produce wider systemic benefits. This explanation is related to, but quite different from, the two-step flow of information that has been identified in decision-making about voting (e.g., Berelson, Lazarsfeld, and McPhee 1954; Huckfeldr and Sprague 1995). In both cases, a subset of the population has extensive or accurate in- formation about choices. In the two-step model of political decision-making, other citizens talk to and learn from these knowledgeable voters, who drive the choice process through their own actions and by influencing others. We have shown that large numbers of the parents in our sample do not talk to anyone about their decision. In our model, the marginal consumer, by mak- ing the best choices for him or herself, can provide a positive externality to other consumers by histher market behavior, even without directly commu- nicating information to less informed citizens. I5In the suburban schools Teske et al. (1993) studied, informed consumers exert pressure by manipulating their locational decisions, creating Tiebout-like forces to increase efficiency. Propo- nents of school choice argue that the monopolistic power of schools based on rigid attendance zones is one of the major causes for the failure of inner-city schools. SHOPPING FOR SCHOOLS 785 The Marginal Consumer in the Market for Education To investigate the role of the marginal consumer, we replicate the analy- sis we presented above for all parents by introducing a term for parents who have demonstrated high levels of involvement with the schools by actively choosing the schools their children attend, rather than using the default op- tion of neighborhood schools. To examine the role of the informed, marginal consumers, we introduce a dummy variable for whether or not a respondent was a chooser and an interaction term between that variable and whether or not the parent indicated a particular dimension of education was important. We then rerun the previous models from Tables 2 and 3 with these two terms included. In Table 4, we reestimate the effects of parental attributes on knowl- edge, including the two terms specific to the marginal consumer. In three of the four models, we find that this group of parents is significantly more knowledgeable about schools than other parents. For example, in Table 4, note that the estimates of reading scores of choosers who think that high scores are important are almost 17 percentage points closer to the actual reading scores than are the estimates of other parents, ceteris paribus. Similarly, choosers who are interested in diversity are 13 points closer to the objective measure of the size of the Hispanic population and 8 points closer to the size of the Black population in their children's schools. Note that the one measure where this pattern is not evident is the measure of incidents. Here we find that on average choosers are more accurate than other parents (there is a main effect), but the interaction effect of chooser*importance is not associated with greater accuracy about safety. But as noted, this may be the result of a measurement problem-we are re- lying on school building data, not on program data, and hence there may be substantial measurement error in the objective data we are using as the de- pendent variable. Do Marginal Consumers Match with Schools? In Table 5, we examine the extent to which marginal consumers have enrolled their children in schools that are high on the dimensions of educa- tion that they care about. In general we find evidence of a strong match in the two domains where our data are most reliable. Note that choosers who think that diversity is important are in schools that are more than one stan- dard deviation above the district average, and choosers who care about high scores similarly are in schools that are far above the mean level of perfor- mance for schools in the district. Note that once we introduce the marginal consumer into the model, the match between other parents (the \"average 786 M. Schneidel; fl Teske, M. Marschall, and C. Roch Table 4. The Marginal Consumer Knows More About Schools Reading Asian Black Hispanic Years of schooling Length of residence Church attendance Is issue important? Chooser Interaction: Chooser*Importance District 4 Constant Black % Hispanic % Incidents N. of cases F statistic Probability Adj. R~ *p 4.05. 󰀁\"\"p < .lo. 󰀁The dependent variable is the distance between the estimate of each condition and the actual objec- 󰀁tive condition in the school. The number in the first line of each entry is the regression coefficient, 󰀁while the number in the second line is the standard error of the estimate. 󰀁Source: Survey of parents in two New York City school districts. 󰀁consumer\") and school performance disappears in the models that are most reliably measured-once we account for the behavior of the marginal con- sumer, there is no longer a ~uzz1e.l~ 160nce again, the pattern for safety is somewhat anomalous. We again find that choosers, in general, are in safer schools, but we do not find a significant effect of the interaction term. This nega- tive result may be driven by the quality of the data and our inability to gather accurate data on inci- dents at the program level. SHOPPING FOR SCHOOLS 787 Table 5. Does the Marginal Consumer Pick Schools High on the Attributes They Value? Reading Diversity Incidents Asian Black Hispanic Years of schooling Length of residence Is issue important? Active Chooser Interaction: Active Chooser*Importance Accuracy Accuracy 2* * Constant N. of cases 󰀁F statistic 󰀁Probability 󰀁Adj. R~ 󰀁*** ~4.05. 󰀁Accuracy: The distance measure between the actual score and the score estimated by the respon- 󰀁dent. In the diversity equation, there are two measures of knowledge: one the distance score for the 󰀁percent black (Accuracy), the other for the percent Hispanic (Accuracy 2). 󰀁The dependent variable is the z-score of each school on the specific indicator of performance. The 󰀁number in the first line of each entry is the regression coefficient, while the number in the second 󰀁line is the standard error of the estimate. 󰀁Source: Survey of parents in two New York City school districts. 󰀁It is essential to show that the match between parental preferences and program performance is not simply a function of choosing alternative schools that are superior on all of these dimensions. As evident in Table 6, there is no difference in the mean level of performance of alternative schools and neighborhood schools on reading scores and in racial composition. In addition, reading scores and diversity are not significantly correlated with one another (r = .13, sig. = .67). 788 M. Schneidel; l? Teske, M. Marschall, and C. Roch Table 6. Objective Conditions in Alternative 󰀁Versus Neighborhood Schools 󰀁Neighborhood Schools % Reading at grade level Black % Hispanic % Alternative Schools/Programs Significance of Difference 33 27 65 Note: Safety data is available only at the building level and is thus unavailable for alternative schools 󰀁and programs. 󰀁Source: New York City school report cards. 󰀁Conclusion: Efficiency and Equity in the Market for Schools These findings suggest that school choosers can act as marginal con- sumers by matching their preferences to appropriate schools. While these results are suggestive, particularly because they take place in a low-income, central city environment, we recognize that there are limits to our ability to generalize from them. First, we have focused on parents involved in an op- tion demand system of choice. While this is the most common form of choice found in the public schools, it is not the only form. The results may be different in a system of universal choice, in which all parents are choos- ers. Second, our results obtain in school districts marked by a considerable number of schools and a wide range of performance across the dimensions of education parents value. We do not know if these results would appear in school districts that have limited numbers of schools or that have used ad- mission policies that limit the range of choice. Thus our research findings may be context dependent-the more a choice system resembles a competi- tive market with many options and maximum incentives for parents to be in- volved, the more likely it is that our results will apply.17 Given these caveats, we argue that by changing the incentives of parents to gather information, public school choice can produce two flows of ben- efits. First, at the individual level, choice can allow parents to get more of what they want for their children from the schools. In addition, as parents gain information about the schools, they pressure the schools into being more efficient producers of these attributes of education. But how many par- ents are needed to increase both types of efficiency? \"As note 2 reports, we do find parallel results for universal choice in our matched New Jersey school districts, despite large differences in the mechanisms of choice. SHOPPING FOR SCHOOLS 7 Chubb and Moe (1990), Coons and Sugarman (1978), and other propo- nents of school choice imply that choice will lead to better schools by creat- ing the conditions under which all parents will have incentives to become informed about schools and hence provide competitive pressures on schools. This perspective overstates not only what we should expect from the average parent, but also what is necessary for effective competition. Competitive markets do not need all consumers to be informed-competitive pressures can result even if a relatively small subset of consumers engage in informed, self-interested search. Further, Schneider et al. (1997a) show how much more involved these marginal consumers are in school events than average parents, giving them both \"voice\" and \"exit\" pathways through which they can pressure schools to perform efficiently and effectively. We recognize that markets for public goods, like those created by public school choice, are much more complex than markets for private goods. Many critics of choice (e.g., Rose-Ackerman 1992; Henig 1996) will find them- selves in agreement with the part of our analysis showing that some parents will become more informed than others and will match their children with the best schools. However, these critics are concerned that these informed par- ents will make choices that will harm the children of less well-informed par- ents, in effect, by leaving them behind in the worst schools. But this ignores the possibility that competition can force all schools to improve. Thus, a criti- cal question is whether or not these marginal consumers do increase the effi- ciency of educational outcomes for everyone, or just for themselves. Few areas of social science are more complex and controversial than measuring educational outcomes. Still, there is evidence that schools in Dis- trict 4, where there are many marginal consumers of the type analyzed here, where choice has been ongoing for over 20 years, and where competitive pressures on schools have had a long period in which to play out, have im- proved substantially and now perform above their \"expected\" levels. For ex- ample, when choice started in 1974, District 4 schools ranked at or near the bottom of all thirty-two New York City districts in reading scores. By the mid-1980s, the ranking had improved to fifteen out of thirty-two and has maintained a ranking in the middle range since then (Fliegel and MacGuire 1993). Only one other New York district, a district that has undergone a large increase in income levels due to gentrification (which definitely has not hap-pened in District 4), showed a greater improvement over this period. In the early 1970s, only a handful of District 4's graduating students were accepted by the selective high schools across New York. By 1992, 20% of District 4 graduates were accepted into these high schools, compared to only 9% for the average district. Furthermore, the school with the highest reading scores in the entire city in 1996 is located in District 4, the Talented and Gifted 790 M. Schneidel; l? Teske, M. Marschall, and C. Roch School (made up of a majority of black female students).18 Given this perfor- mance, it is not surprising that parents in nearby districts are voting with their feet in favor of District 4: every year, many students from other districts seek entry into District 4 schools. Okun (1975) argued that all markets embody what he called the \"big trade-off' between efficiency and equity. Markets for public goods, such as ones created by school choice, embody this trade-off. While the perfor- mance of schools in District 4 appear to have been raised by the institution of choice, clearly some schools have been raised more than others and, in turn, some children have benefited more than others. In our sample, the marginal consumer is not randomly drawn from the population, rather helshe is likely to be better educated (31% of choosers had a college education compared to only 11 % of nonchoosers) and less likely to be a member of a racial minority (27% of choosers were White compared to only 3% of nonchoosers). This suggests that the benefits result- ing from the behavior of the marginal consumers may not be spread evenly across the entire market, and that more well-educated active choosers enroll their children in schools that children of other higher status parents also at- tend. But the extent to which this is a gain or a loss depends on the perspec- tive. Since average performance across all schools in District 4 has improved considerably since the implementation of choice, there are positive exter- nalities accruing to all students from the actions of the marginal con~umer.'~ Moreover, the two inner-city districts in our sample offer public schools that long-ago lost most of their White and middle-income families. Successful public school choice may bring back (or keep) these parents in the role of informed, marginal consumers, which is clearly better for the system than losing them to another form of choice-opting out of the public schools en- tirely for private schools or for suburban ones. Manuscript submitted 5 May 1997. 󰀁Final manuscript received 13 October 1997. 󰀁18There is some controversy, but very little data, about the nature of District 4's sustained im- provements. We have analyzed 1996 sixth grade reading and math scores for all schools in New York City. In a regression analysis of more than 400 schools, using control variables for poverty, limited English speaking, school student turnover rates, and racial characteristics of schools (nearly all of which have significant effects in the expected directions), a dummy variable for District 4 indicates reading scores 8 points above that otherwise predicted and math scores over 4 points higher. 191n our analyses of school performance over time, we found that the performance of neighbor- hood elementary schools in District 4, which are filled with students from the zoned local area and are not \"choice\" schools, has not declined and generally has improved relative to citywide averages in both math and reading scores over the past 20 years. This suggests that choice has put pressure on these schools to improve, even though active choosers are not attending these schools. SHOPPING FOR SCHOOLS 791 REFERENCES Althaus, Scott L. 1995. \"The Practical Limits of Information Shortcuts: Public Opinion, Political Equality and the Social Distribution of Knowledge.\" Presented at the annual meetings of the American Political Science Association, Chicago. Anhalt, Bari E., Alan DiGaetano, Luis Ricardo Fraga, and Jeffrey R. Henig. 1995. \"Systemic Re- form and Policy Effort in Urban Education.\" Presented at the annual meetings of the Midwest Political Science Association, Chicago. Bartels, Larry. 1996. \"Uninformed Votes: Information Effects in Presidential Elections.\" American Journal of Political Science 40(1): 194-230. Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1954. Voting: a Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press. Chubb, John, and Terry Moe. 1990. Politics, Markets and America's Schools. Washington: Brookings Institution. Claxton, J., J. Fry, and B. Portis. 1974. \"A Taxonomy of Prepurchase Information Gathering Pat- terns.\" Journal of Consumer Research 4:3542. Coons, John, and Steven Sugarman. 1978. Education by Choice: the Case for Family Control. Ber-keley: University of California Press. Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know about Politics and Why It Matters. New Haven: Yale University Press. Delpit, Lisa D. 1995. Other Peoples' Children: Cultural Conflict in the Classroom. New York: New Press. Dowding, Keith, Peter John, and Stephen Biggs. 1994. \"Tiebout: a Survey of the Empirical Litera- :767-97. ture.'' Urban-Studies 31 Elmore, Richard F. 199 1. \"Choice as an Instrument of Public Policy: Evidence from Education and Health Care.\" In Choice and Control in American Education Vol. I., ed. William H. Clune and John F. Witte. New York: Falmer Press. Feick, Lawrence F., and Linda L. Price. 1987. \"The Market Maven: A Diffuser of Marketplace Infor- mation.\" Journal of Marketing 51:83-97. Fiske, Susan, and Shelley Taylor. 1991. Social Cognition. 2nd ed. New York: McGraw Hill. Fliegel, Seymour with James MacGuire. 1993. Miracle in East Harlem: The Fight for Choice in Public Education. New York: Times Books. Henig, Jeffrey. 1994. Rethinking School Choice: Limits of the Market Metaphor. Princeton, NJ: Princeton University Press. Henig, Jeffrey. 1996. \"The Local Dynamics of Choice: Ethnic Preferences and Institutional Re- sponses.\" In Who Chooses? Who Loses?: Culture, Institutions, and the Unequal Effects of School Choice, ed. Bruce Fuller, Richard F, Elmore and Gary Orfield. New York: Teachers College Press. Hirsch, E. D., Jr. 1996. The Schools We Need and Why We Don't Have Them. New York: Doubleday. Huckfeldt, Robert, and John Sprague. 1995. Citizens, Politics, and Social Communication. New York: Cambridge University Press. Iyengar, Shanto. 19. \"How Citizens Think About National Issues: A Matter of Responsibility.\" American Journal of Political Science 33:878-900. John, Peter, Keith Dowding, and Stephen Biggs. 1995. \"Residential Mobility in London: a Micro- Level Test of the Behavioural Assumptions of the Tiebout Model.\" British Journal of Political Science 25: 379-97. Katona, George, and Eva Mueller. 1955. \"A Study of Purchase Decisions in Consumer Behavior.\" In Consumer Behavior, ed. Lincoln Clark. New York: New York University Press. Kuklinski, James H., Daniel S. Metlay, and W. D. Kay. 1982. \"Citizen Knowledge and Choices on the Complex Issue of Nuclear Energy.\" American Journal of Political Science 26:615-42. 792 M. Schneidel; l? Teske, M. Marschall, and C. Roch Kuklinski, James H., Paul J. Quirk, David Schwieder, and Robert F. Rich. 1996. \"Misinformation and the Currency of Citizenship.\" Presented at the annual meeting of the American Political Science Association, San Francisco. Lazarsfeld, Paul F., Bernard Berelson, and Hazel Gaudet. 1944. The People's Choice: How The Voter Makes Up His Mind In A Presidential Campaign. New York: Duell, Sloan and Pearce. Lodge, Milton, and Patrick Stroh. 1993. \"Inside the Mental Voting Booth: an Impression-Driven Process Model of Candidate Evaluation.\" In Explorations in Political Psychology, ed. Shanto Iyengar and William McGuire. Durham: Duke University Press. Lowery, David, William E. Lyons, and Ruth Hoogland DeHoog. 1995. \"The Empirical Evidence For Citizen Information And A Local Market For Public Goods.\" American Political Science Re- view :705-7. Lupia, Arthur. 1992. \"Busy Voters, Agenda Control and the Power of Information.\" American Politi- cal Science Review 86:390-404. Lupia, Arthur. 1994. \"Short Cuts versus Encyclopedias: Information and Voting Behavior in Califor- nia Insurance Reform Election.\" American Political Science Review 88:63-76. Lupia, Arthur, and Mathew McCubbins. 1998. The Democratic Dilemma: Can Citizens Learn What They Need to Know? New York: Cambridge University Press. Martinez, Valerie R., Kenneth Godwin, Frank Kemerer, and Laura Pema. 1995. \"The Consequences of School Choice: Who Leaves and Who Stays in the Inner-city.\" Social Science Quarterly 76(3):485-501. Newman, James, and Richard Staelin. 1972. \"Prepurchase Information Seeking for New Cars and Major Household Appliances.\" Journal of Marketing Research 9:249-57. Okun, Arthur. 1975. The Big Trade-Off: Eficiency and Equity. Washington, DC: The Brookings In- stitution. Popkin, Samuel L. 1991. The Reasoning Voter: Chicago: University of Chicago Press. Price, Vincent, and John Zaller. 1993. \"Who Gets the News? Alternative Measures of News Recep- tion and Their Implications for Research.\" The Public Opinion Quarterly 57:133-. Ravitch, Diane, and Joseph Viteritti. 1996. \"A New Vision for City Schools.\" Public Interest 122:3-16. Rhoads, Steven. 1985. The Economist's View of the World: Government, Markets and Public Policy. New York: Cambridge University Press. Rose-Ackerman, Susan. 1992. Rethinking the Progressive Agenda: The Reform of the American Regulatory State. New York: Free Press. Schneider, Mark. 19. The Competitive City. Pittsburgh: University of Pittsburgh Press. Schneider, Mark, Paul Teske, Melissa Marschall, Michael Mintrom, and Christine Roch. 1997a. \"In- stitutional Arrangements and the Creation of Social Capital: The Effects of School Choice.\" American Political Science Review 91:82-93. Schneider, Mark, Paul Teske, Christine Roch, and Melissa Marschall. 1997b. \"Networks to No- where: Segregation and Stratification in Networks of Information about Schools.\" American 1201-23. Journal of Political Science 4 1: Schneider, Mark, Melissa Marschall, Paul Teske, and Christine Roch. 1998. \"School Choice and Culture Wars in the Classroom: What Different Parents Seek from Education.\" Social Sciences Quarterly. Forthcoming. Schwartz, Alan, and Louis L. Wilde. 1979. \"Intervening in Markets on the Basis of Imperfect Infor- mation: A Legal and Economic Analysis.\" University of Pennsylvania Law Review 127:630-82. Simon, Herbert A. 1986. \"Rationality in Psychology and Economics.\" In Rational Choice: The Con- trast Between Economics and Psychology, ed. R. M. Hogarth, and M. W. Reder. Chicago: Uni- versity of Chicago Press. Slama, Mark E. and Armen Tashchian. 1985. \"Selected Socioeconomic and Demographic Charac- teristics Associated with Purchasing Involvement.\" Journal of Marketing 49:72-82. SHOPPING FOR SCHOOLS 793 Slama, Mark E., and Terrell G. Williams. 1990. \"Generalization of the Market Maven's Information Provision Tendency Across Product Categories.\" Advances in Consumer Research 17:48-52. Smith, Kevin, and Kenneth Meier. 1995. The Case Against School Choice: Politics, Markets, and Fools. Armonk, NY M. E. Sharpe. Sniderman, Paul M., Richard A. Brody, and Philip E. Tetlock. 1991. Reasoning and Choice: Explo- rations in Political Psychology. New York: Cambridge University Press. Teske, Paul, Mark Schneider, Michael Mintrom, and Samuel Best. 1993. \"Establishing the Micro Foundations of a Macro Theory: Information, Movers, and the Competitive Local Market for Public Goods.\" American Political Science Review 87:702-13. Teske, Paul, Mark Schneider, Michael Mintrom, and Samuel Best. 1995. \"The Empirical Evidence for Citizen Information and A Local Market for Public Goods.\" American Political Science Review :705-09. Thorelli, Hans, and Jack Engledow. 1980. \"Information Seekers and Information Systems: A Policy Perspective.\" Journal of Marketing. 44:9-27. Weimer, David L., and Aidan R. Vining. 1992. Policy Analysis: Concepts and Practice. 2nd ed. Englewood Cliffs, NJ: Prentice Hall. Wilde, Louis L. 1981. \"Information Costs, Duration of Search, and Turnover: Theory and Applica- tions.'' Journal of Political Economy : 1122-41. Wilde, Louis L., and Alan Schwartz. 1979. \"Equilibrium Comparison Shopping.\" Review of Eco- nomic Studies 55543-53. Zaller, John. 1992. The Nature And Origins of Mass Opinion. New York: Cambridge University Press.