lizZo Human Resources Development and Operations Poiloy The Wouid Sank August 1993 HROWPS8 M N I - AND PRiVATE SO""ECT"OuAR RESPONSE TO DECENTRALIZATION AND SCHOOL CHOICE Donald R. Winkler Taryn Rounds mbadmatiodn_; andhec use sha paperouldtakeacafItspoisl sdur. lbflpamn ce ipssd thX umtwlthse of th mabo) and shuddm be auims in any mu_ tD X. Wadd Ban, to 1 afflatped ar mesbua cteBau f euieDinusorth coiis hy eisas Muniidpal and Private Sector Response to Decentralization and School Choice: The Case of Chile, 1981-1990 by Donald R. Winder and Taryn Rounds Latin American and Caribbean Region Technical Department Abstract A large number of developing nations are in the process of decentalizing basic education, with the aim of diversifying revenue sources and introduction greater accountability and efficiency. This is especially true in Latin America, where Chile introduced the first significant reform in 1981. This reform kept most of the responsibility for educational finance with the Ministry of Education but transferred the responsibility for delivering services to municipalities and non-profit, private schools. In response to this reform, municipalities increased their finance of public schools, and the supply of subsidized-private education increased dramatically. By 1990, enrollment in subsidized-private schools represented about one-third of total primary-secondary school enrollments. This paper examines the effects of the reform. Municipal finance, which is closely tied to municipal fical capacity, has created inequities in school expenditures even though it rep-esents only ten percnt of total revenues. Variations in the private school market share across municipalities are principally explained by the ease of market entry, family socioeconomic status, and the relative performance of public and private schools; this model does not offer - satisfactory explanation of the growth in private school enrollments in Chile over time. The effect of the reform on cost-effectiveness is ambiguous. Ministry of Education non-teacher employment declined by over half, cognitive test also declined. There is some evidence that the growth in pnvate school enrollments may have improved overal efficiency since pnvate schools are found to be slightly more cost-effective than public schools. Table of Contents IN TODUCllON ................................................ 1 EDUCAllON REGORM IN CHILE ............................... . 1 MUNICEPAL.AND PRIATESECTORRESPONSE ...................... 2 THE EQUJ7 f AND EFFICIENCY CONSEQUENCES OF THE REFOR M........ 9 CONR CLUSIONS ............................................. 17 REFERENCES . ............................................... 19 INTRODUCTION Countries in Latin America are increasingly experimenting with educational reform which decentralizes the responsibility for providing primary and secondary education and introduces elements of school choice and competition. At present, school choice experiments are still in the p'anning stage, while decentralization is being implemented rapidly and chaotically. Among the features of decentralization in the region are: (1) the division of responsibilities among levels of government is poorly defined; (2) revenue transfers to subnational governments are ad hoc in nature and frequently politically negotiated; (3) subnational governments are ill-prepared to assume their new functions; and (4) little attention is paid to the design of mechanisms to promote accountability and consumer voice. Given the current popularity of decentralization and school choice policies, it is useful to evaluate the experience of one developing country--Chile--which undertook such reforms over a decade ago. As with any case study, unique features of the Chile reform bring into question its applicability to other countries. In particular, the fact the reform took place in a nondemocratic political context means the design and implementation of the reform might not be practical where political opposition, especially among teachers, would be freely expressed. Despite the various caveats one can make about the Chilean reforms, the unique nature of the policy change and the availability of data to measure its effects are sufficient reason to attempt a systematic assessment. EDUCATION REFORM IN CIELE As is true in many countries today, education reforms in Chile were part of a larger decentralization policy, which included assigning municipalities new revenue sources and new expenditure responsibilities.1 In terms of expenditures, the largest responsibilities given to municipalities were primary and secondary education and primary health care. The 1980 Decentalization Act transferred all school property from the Education Ministry to the municipalities. Teachers were terminated as central government employees, given severance pay, and transferred to municipal payrolls. One of the more unique features of the education reform was the introduction of central government school attendance grants (or vouchers) to finance primary-secondary education. Municipalities receive grants based on the number of students attending class each month, with the base grant level adjusted for differences in costs.2 In addition, the EducaWion Ministry directly provides textbooks to schools and directly contracts for the provision of school lunches I The Municipal Revenue Act of 1980 created new sources of municipal revenue, including a large block grat called the Municipal Common Fund, which attempted to equalize fiscal disparities. However, municipalities wae given only very limited powers to change their revenues by alteing tax rates. 2 The 1980 Law of Subventions, which remains relatively unchanged today, provides a per studet payment ajusted for education level and other school characteristics, including location, with nual schools and t oading schools receiving more. 2 to poor children, while municipalities can also receive grants for school construction and rehabilitation from the central government's Regional Development Fund. Also, in the early stages of decentralization, the Ministry provided municipalities with a 3-5% overhead ,,rant on salaries to cover stive expenses. As saed in early policy papers of the Education Ministry, municipalities were expected to supplement the school attendance grants they received. Another unique feature of the reform was the treatment of private schools. Chile has a long tradition of public subsidies to private schools offering free education; historically, most such schools had a religious affiliation. A 1951 law provided a per student subsidy equal to fifty percent of the cost of public education. Subsequent reductions in the real value of this subsidy forced the closure of a large number of private schools in the late 1960s and early 1970s; the military government attempted to reverse this pattem by increasing the subsidy level in the late 1970s. The 1980 reform further increased the subsidy to one hundred percent of the recurrent cost of public education, with the value of the attendance grant determined by the same formula used to determine municipal school grants.3 In effect, the reform introduced education vouchers which could be used in any school, public or private, which did not charge tuition.4 Perhaps the most important behavioral effect of Chile's educational reform was to directly tie school revenues to school enrollments, thus providing an incentive to schools to compete for students. Municipal schools, which often have excess physical and teaching capacity, find that eolling mn additional student generates revenues (the voucher) in excess of marginal cost, thus permitng reductions in municipal fmance. Subsidized, private schools, which have lower recurrent costs than e public schools, can also maximize profits (in the form of director/owner salaies) by enrolling students up to the school's physical capacity.5 MUNICIPAL AND PRIVATE SECTOR RESPONSE Revenues. The schooi attendance grant or voucher represents a minimum expenditure level which municipalities may choose to augment through transfers from their other, non- earmarked revenues. When the reform was initiated in 1980, the grant was set equal to the per student ependiures of the Education Ministry, but by 1990 the real value of the grant had decreased by almost forty percent. The result has been a rapid growth in municipal finance of mWicipal schools. By 1991 municipal finance represented 10.5 percent of total municipal 'Pivate schools are eligible to receive goverament-financed textboob and school lunches for poor child, but e ar not eligSble for capital investmet ga. " In pncile, n schools which charge tuition are eligible to receive the voucher, but 5 40 percnt tax rt on tuition tevenu (as reflected i reduced voucher levels) mea tat vey few schools choose to opely charg tuition. Some schools do receive significant asmunt of donated revenues or revenues i kind, especially for capital investment, which avoid tho tax. sThe rapid gowth in ite number of subsidized, pnvate schools since 1981 sugges tat vouchers, thou declining in real value over the decade, are adequate to generat profits*. 3 school revenues. Although municipal and private schools rece:.ve attendance grants of equal value, municipal schools receive three additional types of financial assistance: (1) in-kind transfers of school buildings from the Education Ministry to the municipalities; (2) cash transfers from the Regional Development Fund for municipal school construction and rehabilitation; and (3) cash transfers from municipal general funds.6 The result is that municipal school revenues per student exceed subsidized-private expenditures.7 Changing Market Share. As shown in Figure 1, the percentage of all primary-secondary school students in municipal schools has continuously declined since the introduction of the reform, and the share of all students in subsidized-private schools has continuously grown and currently represents about one-third of total enrollments. Meanwhile, the share of students in paid-private schools initially decreased and subsequently increased. Upon the introduction of the ll-cost voucher all paid-private schools which were spending less than the voucher amount had a clear incentive to accept the voucher and become subsidized-private schools. As a result, between 1979 and 1982, paid-private enrollments decreased by about 27 percent while subsidized-private enrollments increased 35 percent.8 However, as the real value of the voucher declined throughout the 1980s, paid-private school enrollments grew. By 1989, paid-private enrollments were larger than they were prior to the reform and were 50 percent higher than they were in 1982. !nterviews with subsidized-private school directors suggest the trend of conversions from subsidized-private to paid-private status will continue, largely due to the decrease in the real value of the voucher. Despite the voucher scheme, not all stdents in Chile can choose the school they attend. Of the total of 325 municipalities, 234 have subsidized-private schools, and only 72 have paid- private schools. Both types of schools are concentrated in urban areas, where private school enrollments are as high as public school enrollments. In 91 predominandy rural municipalities, students have no choice but to attend public schools.9 Scre subsizied private schools also receive othe financial assistance in the form of donated school buildings and fiancial or in-kind contributions from school paent-teacher associations. 7 We esdimate the difference is between 13.8 percent (the reported difference in expenditures per pupil in municipal compared to subsidized-private schools) and 36.8 percent (which includes, in addition, an estimate of the implicit rental subsidy received by municipal schools). The implicit rental subsidy is esfimated assuming that municipal and subsidized-private schools spend an equal percentage of their budget on materials and supplies. A recent study for the Education Ministry finds municipal schools spend 90 percent of their budget on salaries, compared to 67 percent for subsidized-private schools, which in most cases must also pay rent on their premises. a The Education Ministry budget increased 49 percent in real tems the year (1981) of the reform, but not all of this can be attributed to subventions to formerly private-paid schools. ' A municipality is defined as a geographic area; while it typically includes an urbanized area, it may be predominantly rural. 4 Figure 1: Market Sham in Prlmary.Secondary School Enrollhent - - - - -. - -JI ~ - - - - --. -- - - -- -- - - 0 1979 1982 1987 1989 * Public Subsidized Private El Paid Private Demand for Private Education. The demand for private education has been hypothesized to be directly caused by (1) excess demand resulting from less than comtplete coverage by the public school system anl (2) heterogeneity in tastes not fully satisfied by public schools oriented to the tastes of the median voter, in addition to the usual variables of tuition price and ability to pay. "~ While excess demand, as measured by comparing the size of the age cohort to the supply capacity of the public sector, is likely to explain demand in many developing countries, u is unlikely to be an important factor in Chile, where 80 percent of the pmimary-secondary school age cohort are enrolled in school and where public supply is most deficient in rural municipalities. The relatively homogeneous cultural and religious composition of the population also suggests these taste factors are likely to play a relatively unimportant role in explaining demand tbr private schooling. On the other hand, heterogeneity in tastes associated with socioeconomic status or educational levels of households, a household's ability to pay, and school quality are all likely to positively affect demand. Of these variables, the most difficult "O See Jamnes (1985. 1991) for further elaboration of these arguments. S Table 1 SnTDENT CHARACTERISTICS BY TYPE OF SCHOOL * Z . . - T Characteristics Public School Subsidized Private Paid Private School _____________ ____ _ ,Students School Students Students Years of Ed'lcation of Head of 7.8 9.3 NA Household' Monetary Income (Thousands of 110.6 153.3 NA Pesos per Menth) Househo!d Siz. 5.1 5.1 NA Percent of Households not 58.3% 76.7% NA Receiving any School Meals 5 Average Education Level of 2.18 2.41 3.63 Parents (1 low, 5 hgihr) Average Socioeconomic Level of 0.94 1.27 2.16 Parents (1 low, 4 highr Average Repetition Rater 9.7 7,8 2.2 Average Fourth Grade Math Scr 49.86 60.26 77.06 (1990)" Average Fourth Grade Sp9ish 5S.46 61.47 77.41 Score (1990) 'For 72 municipalities in the CASEN 3 (1990) sample. For 325 municipalities from the 1990 SIMCE test and teacher questiomaire. to measure is school quality. Among the perceived indicators of quality in Chile are the school name, the presence of school uniforms, and the background of school peers." Table 1 shows a number of household and school characteristics for children in public, subsidized-private, and paid-private schools in Chile. These data show: (1) public and subsidized-private schools appear to be more similar than do subsidized-private and paid-private " An unpublished experimet caried out by a Chilean thik tank found that ching the scool w a fom Spani to English and intoducing school uniforms in public schls tled in nced pulibc schol mrllim Per group composition, either in terms of socioeconomic status or achievemen, has beeu found by several studieS to be one of the more imporan school factors that determine larning. 6 schools; (2) household income, education, and socioeconomic status axe all higher in private than public schools; and (3) school peers in private schools are of higher socioeconomic status and higher educational achievement levels than peers in public schools.'2 Supply of Private Education. Private education in Chile is characterized by the nature of ownership and the price of the service. The traditional private school has an affiliation with some non-profit (religious or civic) organization; it often has an elite clientele and charges high tuition levels. The new, non-traditional private school is sectarian, frequently owned by former teachers in the public system, and typically receives most its revenues from the govemment voucher.'3 The taditional private school is similar to private education in other countries in its affiliation with non-profit organizations and its access to donated capital and volunteer labor, which reduce its costs. The supply of education by traditional schools is likely to be relatively inelastic with respect to input prices and voucher levels, although the voucher is likely to influence its decision to accept tuition or not. The new private school, while required to be non- profit, may be de facto profit seeldng with surplus revenues captured by the owner/director of the school. Hence, the supply of education by new schools is likely to be more elastic with respect to input prices and the voucher level. Since public schools already exist in all municipalities, private supply is also affected by the opportunity for market entry, which is likely to be easier in larger metropolitan areas which could support both public and private schools. Model of Private School Market Share. This paper attempts to estmate a reduced-form model of private school market share for the 234 urban municipalities that have both public and private schools.14 Since there are no data on paid-private school enrollments at the municipal level, market share is defined as the percent of publicly-financed students that enroll in subsidized private schools. The model hypothesizes the demand for subsidized-private schooling is determined, among other things, by the relative educational outcomes of public and private schools. Earlier empirical studies of variations in private market share across countries have not tested this hypothesis but have instead demonstrated that public school speding levels adversely affect private school demand.'5 In the case of Chile, educational outcomes would appear to be a better indicator to parents of the quality of schooling hn would public school spending, which does not always reflect the quantity of school resources received by public school 12 Educational achievement, as meaued by the grade four SIMCE test adminsterd in about 5000 of Chile's schools, is both a schooling outcome and an nut to leaning. Te Tse dif_eces an reflected in thest wcoes of sden erolled in the two types of schools. In 1990, subsidized private schools more than ten yeusi old hd an average Spani teat score of 63.8, compared to 60.5 for stbsidized private schools lss t ten years old. T4 he model is a teducod form equation of the supply of and demad for privat education. " See James (1993). One difficulty with the spending vasiable is that it is likely to reflect vardtions in input prices s well as the quantity of educational input received by childr 7 children."6 This model also assumes the subsidized-private school market share is determined by relative test scor but not vice versa, since government regulations prevent subsidized-private schools from using entrance examinations to select students. Of course, schools may use proxies for cognitive achievement in selecting students, but at least one study finds no evidence of such simultaneity. 17 Earlier cross-sectional empirical work has shown the private school market share to be larger where public school provision is small, where there is high cultural heterogeneity, and where governments subsidize private schools."8 These findings are not easily confirmed for the cross-section of Chilean municipalities as public school provision is consistently high, the culture is fairly homogeneous, and subsidy levels, adjusted for differences in costs, are uniform. The estimated model, reported in Table 2, findi a strong relationship between the private school market share and population density, which can be intexpreted as a proxy for ease of market entry."' Neither the subsidization or voucher level per student nor the teacher salary Ie 'el are significantly related to private supply. This is not surprising given the lack of variation in voucher levels, controlling for costs, and the lack of a measure of teacher salaries in private schools. Is nstad, it sometim reflects constsints on mnimcipalities in their ability to either release redundant teachers or to replace older, highly pad teachers. 17 In etimang simultaneous models of pr.vate school market share and public school spending, a possible proxy for tive educational quality, James (1993) found no evidence that private market share determines public seding. " See James (1986 and 1987). u A Huln probabilty model is adopted over a logit model, as its coefficients are more easily intrprted, and, in this case, does not yikld predicted values of percent private tiat lie outside the zero to one boundary. 8 Table 2 PRIV.TE SCHOOL MARKET SliARE (Standard Errors In Parentheses) - ndependent Variables Private Earollments as Percent of Total Intercept 15.551 10.720 Subsidization per Stdent -0.255 -0.232 (in thousands) (.369) (.371) Teach Salary in Public Schools 1.359 -1.301 (in thousands) (1.641) (1.655) Density 1.785" 1.786 (in thousands) (.344) (.347) Socioeconomic Status of Familis 8.573-m 8.672 with Children in School (3.615) (3.638) School AV P%7ulation0.141 0.147 School Age Population (.102) (.103) Poverty Rat 0.382' 0.370' (.170) (.173) Ratio of Public to Private Test -11.392" Score-Math (6.061) Ratio of Public to Pnivate Test -6.594 Score-Spaniah (6.553) R2 0.240 0.231 S.E. 14.583 14.675 Significant at 0.10 level. Significant at 0.05 level. 9 Three demand variables are found to be strongly related to the private school market share. Higher average household socioeconomic status, a proxy for household tastes, positively affects private school demand, a finding consistent with our expectations.20 However, the school-age population of the municipality, a proxy for heterogeneity of tastxs, is not statistically significant. A high poverty rate is also found to increase private school demand; since most poor students are found in the public schools, higher poverty rates may drive higher income students out of public and into private schools. Finally, there is weak evidence that student test scores affect school choice. A higher ratio of public to private school test scores in mathematics (but not Spanish) adversely affects private school demand. Since most parents do not have access to school-level test score data, these variables are a proxy for parental perceptions of student peer groups. These results help confirm some propositions but fail to shed light on others. They confirm the proposition that the supply of private education is larger in densely populated areas where there is likely to be greater ease of market entry; that households of high socioeconomic status are more likely to prefer private schooling; and that peer group characteristics, be they measures of achievement or economic status, affect school choice. But the model fails to explain the large increase in private market share between 1982 and 1989 at a time when the real voucher level was decreasing. Two untested hypotheses suggest themselves: (1) the demand for private education shifted sufficiently rapidly to offset the reduction in supply resulting from reduced real vouchers; and (2) input prices, especially teacher salaries, decreased even more rapidly than voucher levels, resultng in continued supply increases. THE EQUIrY AND EFFICIENCY CONSEQUENCES OF THE REFORM The major criticisms of both decentralization and privatization of education through voucher schemes is that both policies may lead to greater inequality of educational opportunity among children. Data constraints do not permit us to fully examine the equity consequences of Chile's educational refonn, but the data do permit some partial answers. Equay in Spending. Equity can be assessed in terms of either school inputs or school outputs. The voucher financed by the central government provides a high base level of minimum spending that either municipal or subsidized-private schools can increase from other revenue sources. As noted earlier, the voucher vanes among jurisdictions to compensate for differences in costs between levels of education and student and community characteristics, including needs of the. physically challenged. However, the formula that determines the voucher level does not adjust for the compensatory requirements of poor or low-achieving children. In fact, the only extra resources those children receive is in the form of free school lunches. 1 Given the Zero or ner-zeto pric, of private educatona, household ability to pay should not be strongly related to demad. 10 Fgure 2: Average Grats and Expenditures for Education In Chile Munidpalties LJc v *M cOv ISO - 140 ---------------------U --------------- 120 -12.7---------------- -------------- 104.6 100 -~-- - - - - -- - - - | _ 78.7 0 w----___ 60.9 -- - 40 - _ _ - - - 40- ._ __-__ __ _ _. . 20- -- -- -.- --- -. LO HI LO HI LO HI FOVETY AL CAP. St ACH 0 Desite the appret equity of the basic voucher, the rdatively smal averge amount of extra monies budgeted by the municipalities results in sigficant expenditure dispanties. Figure 2 compares the finance and expendiures per pupil for the bottom and top decile of municipalite ranied by poverty rates, fiscal capacity, and education test scores. As shown in the figure, the voucher alone provides more cash per pupil in high poverty than low poverty municipalities and in low achieving than high achieving municipalities. But disparities in municipal finance reverse the pattemn in terms of total spending, resulting in higher per pupil spending in low than high poverty municipalities and in high than low achieving municipaliies. Regarding fiscal capacity (as measured by per capita genera revenues), the voucher is larger for high than low fiscal capacity municipalie (reflecting cost variations), but adding the municipal finance results in ependitWes in high fiscal capacity municipalities more than double those in low fiscal capacity 11 municipalities.2 Equality in Outcomes. As shown in Table 1, educational outcomes as measured by repetition rates and mathematics and Spanish test scores differ greatly between public, subsidized-private, and paid-private schools, with public schools having the worst and paid- private schools the best outcomes. Table 3 disaggregates achievement scores by four levels of household socioeconomic status. These results show the same ranking of performance by school type controlling for average socioeconomic status in the schools. The differences between municipal and subsidized-private schools, however, is small compared to the differences in test scores between socioeconomic levels.' There is no convincing evidence that changes in test scores since the reform have favored one type of school over another. However, at least one study concludes that within public schools higher socioeconomic level students' performance improved and lower socioeconomic level students' performance worsened between 1982 and 1988.23 2a In addition to generating spending inequities between municipalities, municipal finance cts spending differences between public and subsidized-private schools. As of 1989, per pupil primary education and secondary education expelditures in mumicipal schools wore 8.7 percent and 22.3 percent, respectively, above those in subsidized-private schools. This compariso, however, is not complte as it does not include the implicit rental value of public school buildings or services and goods donated in-kind to private schools, nor does it take account of the different teacher's wages paid by public and private schools. A 1990 Education Ministry survey showed that subsidized private school salaries are 24.1 percent lower than public school salaries at the primary level and 17.5 percent lower at the secondary level. 12 The data reported are for mathematics achievement in large cities. Morales Frias (1990) repor seva other disaggregations of test results by subject matter, size of jurisdiction, and socioeconomic satus. Th reslt are in general consistent with those reported in Table 3 although there are cases where municipal schools have higher test scores than subsidized-private schools. : Espinola H. (199!). 12 Table 3 MATHEMATICS ACHIEVEMlENT (GRADlE 4) IN 1988 FYOR MEROPOLITAN AREAS School Type Socioeconomic Level Munlcipal Subsidized Private Paid Private 4 (highes) NA 65.3 74.9 3 56.7 58.3 67.2 2 48.4 51.8 NA 1 (lowest) 47.2 49.6 NA Source: Morales Frias (1990) Effects on Costs. The overall cost of primary-secondary education in Th'ie .s decreased considerably since the reform; real expenditures of all publicly-financed schools decreased 25 percent between 1981 and 1989 (see Figure 3).2 In addition, the overhead adriinistraive costs of the Education Ministry decreased, with employment decsing m 18,522 to 8,305 employees between 1981 and 1989 (see Figure 4).2s While much of qe. decrease in education expenditures has been the result of reduced teacher salaries, outlh.s Jn textbooks and school lunches have declined as well.' Cost-Fffecdveness. While the unit costs of primary-secoraary education have decreased since he reform, the evidence on educational outcomes is mix4d with some evidence of declines in test scores for all types of schools between 1994 and 1988.2? The ambiguity of the 2 This average hides te fact that per pupil expe4ituc acually increased in primay education, while they decreased in seconday education. See Espinola RI. (1991). 25 Tb Educati a. ?finistry overhead is. of coure, for all education level. 26 'b number of textbooks distibuted annually decreased by 43.7 percent betwen 1980 and 1990; in 1980 the averge student received 1.5 texts, while in 1990 this number declined to 0.9. Tho number of school breakfasts and lunches served dso decluied i the same m period. 27 However, diffences in tes construction over this timo period reu the tests not being stictly comparable over tim. 13 Figure 3: PublIly Fnaced ExpendItures Per Pupil 100 90 80 70 60 40 1981 1982 1983 1984 1985 1986 1987 1999 1989 Average Real Expenditures achievement data also make it difficult to determine changes in efficiency in the use of resources. Nothing can be concluded from the fact that education expenditures have declined more rapidly than test scores over the past decade, for while there has been some reduction in real school inputs, the principal consequence of reduced expenditures has been reduced teacher pay. Similarly, the observation that test scores are higher and per pupil expenditures are lower in subsidized-private compared to municipal public schools says nothing about the efficiency of these school types in producing new cognitive knowledge, for students in subsidized private schools also have home environments that are more conducive to learning and have a higher level of knowledge upon starting school than do students in municipal public schools. The education reform introduced competition between schools for students. As opposed to the usual monopoly model of public education, competition should increase efficiency in producing those educational outcomes which influence school choice.2' In the particular case of Chile, it could be argued that the incentives for maximizing performance are more transparent 21 See Levin (1991) for an elaboration of this thesis, including discussion of the possible social costs of private choice in education. 14 Figure 4: Mlnlstry of Education Employment 30 25 20 15 nl 1979 1980 1991 1982 1983 1984 1985 1996 1987 1998 1989 1990 *MInlstry of Educatlon and direct for private than public schools. The owner/directors of private schools can directly benefit from the increased enrollments which result from better performance, while the principal benefit to public schools from increased enrollments is reduced municipal finance; public school managers may in fact find that increased enroUments makes for extra work.2Y If this argument is true, one would expect to find private schools performing better than public schools. In what follows, we specify a crude model of student learning and attempt to estimate the relative effecLiveaaess of public and private schools in Chile. The performance of public and private schools has by now been examined in severl important empirical studies, beginning with Coleman, Hoffer, and Kilgore (1982). However, lack of adequate data have limited the number of such studies for developing countries." The data for Chile have several flaws that constrain I Municipal finance may be reduced to the extent the revenue from the voucher exceeds the marginal costs of au additional earolled student; this is likely to be true for many municipalities where increased private school entollments have resulted in excess capacity in the public schools. "' See Psacharopoulos (1987), Cox and Jimenez (1987). and Jimenez, Lockheed and Wattanawaha (1988). 15 the validity of the empirical analysis. First, our observations are at the municipal and not the individual student level. Second, the sample consists only of the seventy largest municipalities in the country.3' Third, the measures of current (grade 8) and lagged educational achievement (grade 4) are for different students within the same municipality,32 and we have school input data for only one year (1989). Thus, we miake the strong assumption that student cohorts do not differ and school inputs are perfectly correlated over time. Finally, our measures of home and school environments are not very rch. We assume the usual model in which achievement is determined by prior or lagged achievement, the home environment, and the school environment. Since the sample size is small, consisting of 70 municipal-level observations of public schools and 70 municipal-level observations of subsidized-private schools, the estimated parameters are constrained to be identical for public and private schools, excepting school expenditures, where including a separate variable for private school expenditures permits a test of the hypothesis that private and public schools are no different in their productivity. The estimated parameters of the model are given in Table 4.33 Models 1 and 2 show that the effects of the home environment are captured by a mee'ure of lagged achievement." Model 3 shows that, controlling for fourth grade achievement, school expenditures significantly affect eighth grade achievement; in addition, the magnitude of the effect is about 30 percent Larger for privatr, than public schools. Overall, however, the magnitude of the expenditure effect is small. To bring about a five percent improvement in eighth grade achievement would require a 32 percent increase in public school spending or a 27 percent increase in private school spending. Whlle aggregate data, a small sample, and weak measures of the student's home environment suggest caution in drawing strong conclusions from these findings for Chile, these results are consistent with other empirical work companng public and private schools, which typically finds private schools to have a small edge in cost-effectiveness. The principle t The data come from the 1990 household surey (CASEN 3), and the sample size is adequate to compute sepaate subidized-vate and public school mea only for the larges municipalties. 9 Th Chile achievent test (SIMCE) is admisterd at grade four in even-mbered yeas (e.g., 1988) and grade ight in odd-numbered yoers (e.g., 1989). " The model is estmted ung OLS. As noted earlier, susidied-pivaue shos anot permitted to select studens on the basis of test performance, but they might use proxies fbr test perfomnc in slecting stue Hence, we used instrumental techniques to estite an alternative model specification, which includes the percent of staudents in subsidized-private schools as an endogenous, independent vaiable. The coefficient on Percent Private was stati6tically insignificant, and other estimated coefficients were not materially affected in terms of either size or statistical significance. 34 Since the esimated results ar very similar for grade eigt mathematics and Spanish achievemn=t, adly the mathema esults are reported here. Table 4 EDUCATIONAL ACHIEVEI MODEIS (Standard Emrs in Parentheses) ModelI Mod 2 Model 3 Gradl108 e 8 Grade 8 Grade Grade 8 Mathematics Mathomatics Mathematics j Spnsh Adib_me Achieveame Achievnxt Acheeet 46. 15I4* 12.988** 19.264* 15.609* (9.148) (8.281) (8.511) (8.474) Yas of Education of Houseold Head 1.282** -0.011 -0.111 0.098 l__________________________________ (.281) (.271) (.269) (.289) Household Size -3.1414 -0.691 -1.087 -0.095 (.281) (1.200) (1.118) (1.229) Grade 4 0.582** Spanish Acheme- (.075) Grade 4 0.703** 0.637** Mathematics Achievement (.082) (.085) I Expediur Per Student 0.204* 0.1 18* 0.103** 0.098** (in thousands) (.033) (.028) (.028) (.029) Expediu Per Shudent in Private Schools 0.031** 0.036** (in thousands) (.013) (.013) R2 0.403 0.618 0.633 0.661 Stand Enror 4.658 3.724 3.651 3.808 Sigificant at 0.10 level. Signdficant at 0.05 leve. 17 difference is one of context-despite the fact that both public and private schools in Chile compete for students, private schools are still slightly more cost-effective. CONCLUSIONS Education reform in Chile had two principal characteristics which are popular among reformers in other countries today: decentralization of the provision of educational services and introduction of choice and competition. The reform had severl results: (1) Municipalities provide a small share of total public school financing, but inter- municipal variations in fiscal capacity generate inequities in per pupil school expenditures. (2) The relatively large size of the central government school attendance grant (almost 90 percent of total financing) ensures all pupils receive a relatively high minimum level of school services. (3) In the short-run, introduction of the voucher program led to large numbers of schools changing status from paid-private to subsidized-private, resulting in significantly higher government-financed education spending. (4) In the long-run, reductions in the real value of the voucher led to paid-private schools regaining their earnier market share. (5) The subsidized-private market share has consistently grown, despite reductions in the real value of the voucher and despite approximate parity with municipal schools in terms of test scores (controlling for socioeconomic status). The model estimated here suggests parents select private schools based on the characteistics of their students; since the demand for private schooling increases with socioeconomic status, private schools have a continuing advantage in terms of student background characteristics. (6) Subsidized-private schools appear to be more cost-effective than municipal schools. While test scores, controlling for socioeconomic status, are approximately equal in municipal and subsidized-private schools, unit costs are lower in private schools. Also, the model estimated here provides evidence that additional monies spent in subsidized- private schools yield slightly higher returns than those spent in public schools. Since the design and implementation of the Chile reform occurred in a setting that did not permit political opposition, they are unlikely to be easily replicated in other countries. In addition, Chile has had several institutional advantages in implementing the reform. By the standards of other developing countries, management capacity at both the central and municipal 18 level is relatively high, and the degree of public corruption is perceived to be low." As a result, it was possible to implement a financing mechanism which requires accurate counts of numbers of students in the classroom and which effectively penalizes schools for inaccumate reporting. Also, over the past two decades the Chilean public sector has become sophisticated in the use of contracting to purchase specialized expertise that would not otherwise be available (e.g., municipalities contrcting consultants to develop the detailed plans required to obtain central government financing of capital investment.) Sound financial management has also been important in implementing the decentralization component of the reform. All levels of government in Chile use the same standardized government accounts for budgeting and expenditure reporting, with municipal expenditure reports submitted on a monthly basis to the Finance Ministry. This financial information system provides the basis for periodic audits of municipal accounts by the country's Contoller General, and helps ensure that central government grants are used as intended.3" Finally, Chile is the only country in Latin America to have a national system of student testing. In principle, this can provide both municipalities and parents with information on their schools' performance. In practice, the results are not widely disseminated, and parents seldom know either their school or their child's performance.3 -" lhe point being made here is not that the Education Ministry and Chilean municipalities have adequate aagement and planning capacity; there are many deficiencies, especially among smaller municipalities and in the supervisioa capacity of the Education Ministry's provincial administrative units. 'M The penalties associated with misappropriation of govermment funds are severe, including personal liability. ' Te reasns for lack of dissemination are not clear. Te testing system was first introduced as part of a pan to measure teacher performance and to tie saaries to performance. Several difficulties, including teacher opposition, led to this plan never bei-g fully implemented. 19 REFERENCES Castaneda, Tarsicio, Combating Povery: Social Policy and Decentralizadon in Chaie During the 1980s, Centro de Estudios Publicos, Santiago, 1990. Coleman, James, Thomas Hoffer, and Sally Kilgore, High School Achievement: Public. Catholic, and Private Schools Compared, Basic Books, New York, 1982. Espinola H., Viola, Descentralzacion del Sistemc L-olar en Chile, Centro de Investigacion y Desarrollo de la Educacion (CIDE), Santiago, 1991. Harbison, Ralph W. and Eric A Hanushek, Educadonal Performance of the Poor: Lssonsfirm Rural Northeast Brazil, Oxford University Press, 1992. James, EsteUe, 7he Public/Private Division of Responsibilides for Fducadon: An Intenationel Comparison, Institute for Research on Educational Finance and Governance (IFG), January 1985. James, Estelle, Private inance an Management of ducaon in Developing Countries: Major Policy and Research Issues, Interational Institute for Educational Planning (UEP), Paris, 1991. James, Estelle, Why is There Proportionately More Enrollment in Private Schools in Some Countries?, Worldng Paper WPS 1069, The World Bank, 1993. Jimenez, Emmanuel, Marlaine Lockheed, and Nongnuch Wattanawaha, "The Readive Efficiency of Private and Public Schools: The Case of Thailand," The World Bank Economic Review, Vol. 2, No. 2, pp. 139-164. Levin, Henry M., 'The Economics of Educational Choice, Economics of Educaton Revew, Vol. 20, No. 2, pp. 137-158, 1991. Morales Frias, Joaquin, "Sistema de informacion Sobre la Calitad de la Educacion en Chile,' Santiago, April 1990. Psacharopoulos, George, 'Public Versus Private Schools in Developing Countres: Evidence from Colombia and Tanzania,I Internadonal Jounal of Educatonal Development, Vol. 7, No. 1, pp. 59-67, 1987. Reid, Gary, Civil Sernce Reform in Lain America: Lessons from Eperience, LAC Technical Depatment, The World Bank, 1992. Schiefelbein, Emesto, The Introdwction of Economic Compedtion In Clslean Eduaton, UNBESCO/OREALC, Saniago, March 1991. Human Resources Development and Operatons Policy Wortig Paper Series Contact for Tite Author Date pape HROWP1 Social Development is Nancy Birdsall March 1993 L Malca Economic Development 37720 HROWP2 Factors Affecting Achievement Eduardo Velez April 1993 B. Diallo in Primary Education: A Ernesto Schiefelbein 30887 Review of the Uterature for Jorge Valenzuela Ltn America and the Caribbean HROWP3 Social Policy and Ferility Thomas W. Merrick May 1993 0 Nadora Transitions 31091 HROWP4 Povert Social Sector Norman L Hicks May 1993 J. Abner Development and the Role of 38875 the World Bank HROWP5 Incorporating Nutrition into F. James Levinson June 1993 0. Nadora Bank-Assisted Social Funds 31091 HROWPO Global Indicators of Nutritional Rae Galloway June 1993 0. Nadora Risk Pl) 31091 HROWP7 Maling Nutrition Improvements Donald A.P. Bundy July 1993 0. Nadora at Low Cost Through Parasite Joy Miler Del Rosso 31091 Control HRO Dissomination Notes Contc for Tide Date note No. 1 Tobacco Death Toll February 19 1993 L Malca 37720 No. 2 The Benefits of Education for Women March 8. 1993 L Malca 37720 No. 3 FPverty and Income Distribution In Wn March 29, 1893 L Malca America 37720 No. 4 BIAS Is Herel Apfil 12, 1993 L Malca (Committee on Business Innovation and 37720 Simplification) No. 5 Acute Respiratory Infections April 28, 1993 L Malca 37720 No. 6 From Manpower Planning to Labor Market May 10 1993 L Malca Analysis 37720 No. 7 Ehanhidng Iestments In Education Through May 24, 1993 L Malca Better Nutition and Health 37720 No. 8 Indigenous Peoole In Ltin America June 7. 1993 L Malca 37720 No. 9 Developing Effective Employment Services June 28, 1993 L Malca 37720 No. 10 Social Secury: Promise & Pitals In July 12. 1993 L Malca Pivatimation: Experience from latin America 37720