to graduate or not:coppfs1.asu.edu/spa/abfm2008/new/nguyen.doc · web viewwhile muller (1995) find...

67
Use of Cost Function Analyses for Inter-Governmental Educational Transfers in Vietnam By Hoang-Phuong Nguyen Maxwell School of Citizenship and Public Affairs, Syracuse University (Revised Oct 2008) Abstract : No study has been conducted to address two major problems in Vietnam’s education finance system. First, the central government’s population-based formula for allocating educational transfers fails to incorporate inter-provincial efficiency and cost differentials effectively. Second, the government lacks a useful tool for estimating spending required to attain a performance standard. Using data for 2002 and 2005, the paper demonstrates how a cost-function estimation can be the reliable and valid solution to the above two problems by 1/ identifying multiple cost and efficiency factors influencing provinces’ expenditures on education such as their student-body characteristics, provinces’ geographic structures, income-related efficiency variables, and by 2/ providing an estimation of the minimum spending for a performance goal using the previous regression results. 1. INTRODUCTION Education is always one of the most important sectors in almost all countries. Vietnam is not an exception. General public education is of significant interest to policy makers and local people in the country. GVN (2005) represents a strong commitment 35

Upload: hadat

Post on 11-Apr-2018

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Use of Cost Function Analyses for Inter-Governmental Educational Transfers in Vietnam

By Hoang-Phuong NguyenMaxwell School of Citizenship and Public Affairs, Syracuse University

(Revised Oct 2008)

Abstract:

No study has been conducted to address two major problems in Vietnam’s education finance system. First, the central government’s population-based formula for allocating educational transfers fails to incorporate inter-provincial efficiency and cost differentials effectively. Second, the government lacks a useful tool for estimating spending required to attain a performance standard. Using data for 2002 and 2005, the paper demonstrates how a cost-function estimation can be the reliable and valid solution to the above two problems by 1/ identifying multiple cost and efficiency factors influencing provinces’ expenditures on education such as their student-body characteristics, provinces’ geographic structures, income-related efficiency variables, and by 2/ providing an estimation of the minimum spending for a performance goal using the previous regression results.

1. INTRODUCTION

Education is always one of the most important sectors in almost all countries. Vietnam is not an

exception. General public education is of significant interest to policy makers and local people in

the country. GVN (2005) represents a strong commitment to the achievement of Vietnam’s

Millennium Development Goals (MDG) including Goal 2 on universal primary education. The

government’s firm commitment to general education can also be seen in the promulgation and

implementation of the Education for All (EFA) Action Plan from 2003 to 2015 (GVN, 2003).

Despite the critical importance of public education on Vietnam’s public policy agenda, there has

been little research into the educational sector of Vietnam. Particularly, no scholarly study has

ever been conducted to address two major problems with the country’s education finance system.

First, the central government’s current budgeting norms for the transfer of educational funds to

35

Page 2: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

provinces are over-simplified and based primarily on population (Nguyen et al., 2001, 12). It is

because there is “no fair research and formula-based method of educational fund distribution at

the central level” (Nguyen et al., 2001, 12). The population-based norm does little to recognize

several different cost and efficiency1 factors for provinces to achieve an “adequate” education.2

For instance, a factor that is not fully accounted for in the norm is the proportion of ethnic

minority students. Provinces with large numbers of these students need to pay higher teacher

salaries to compensate for their more challenging teaching environment. The government’s

failure to fully reflect efficiency and especially cost differentials in its distribution of transfers

may lead to large inter-provincial discrepancies in access to adequate financial resources for

public education, and thus violate the most basic equity standard of educational adequacy

whereby children in every province receive an education that meets some minimum standard.3

Second, the government lacks a useful tool to estimate how much it may need to spend to

obtain a desired performance standard. Currently, the government merely expresses its intention

to increase educational funding with little attention to student performance. Specifically, the

government expected to increase the ratio of education spending to Gross Domestic Product

(GDP) from 3.7 percent in 2002 to 4.2 percent in 2015 in the 2003-2015 EFA Action Plan

(GVN, 2003, xi) without a systematic link between this spending-effort target and achievement

of performance goals. The lack of an analytical tool that can link expenditures with performance

1 As in other education cost studies, efficiency refers to the degree of resources provinces consume to produce a certain output given the current technology. It will be further discussed in later sections.2 Unlike the United States where schools and districts are held accountable to their state and federal government for student achievement usually measured by test scores (Hanushek and Raymond, 2004; Clotfelter et al., 2004), Vietnam has no explicit system of educational accountability. Therefore, what is considered to be an “adequate” or “acceptable” education is not publicly known or declared. The central government in its budgeting process might, however, have an underlying common adequate level of student performance that provinces need to achieve. 3 See Yinger (2004) for a thorough discussion of four equity standards in education aid: adequacy, access equality, fiscal neutrality and equality.

1

Page 3: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

goals makes it difficult for the government to know the level of funding necessary for provinces

to attain a certain level of student performance.

This paper’s principal contribution is to demonstrate how cost-function estimation can be

the reliable and valid answer to these two problems underlying Vietnam’s education finance

system. Employing data for 2002 and 2005, the paper shows that the educational spending per

pupil varies significantly depending on the composition of provinces’ student body, their

geographic characteristics, degree of family involvement in children’s education, and various

factors influencing provinces’ efficiency. This paper indicates that education cost analysis can

not only take into account inter-provincial efficiency and cost differentials in a more thorough

manner (via cost and efficiency indices), but can also be used to estimate total funding needed

for provinces to achieve a student-performance goal with different levels of their efficiency.

While it costs the government more to raise a desired level of performance, the total spending, all

else being equal, can be reduced with provinces becoming more efficient in their operations.

The paper proceeds as follows. The second section presents a brief overview of

Vietnam’s general public education and its financing mechanism. This section is then followed

by detailed development of a conceptual framework for the cost-function estimation. Empirical

strategy is developed in the fourth section. Next, I will describe data and measures used in the

estimation model. Discussions of the regression results are presented in the sixth section. This

section also exhibits estimation procedures and results of educational expenditures needed for the

government of Vietnam to attain a desired level of student performance under different

assumptions on provincial efficiency. Section seven concludes with suggestions for future

research.

2

Page 4: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

2. EDUCATION SECTOR AND ITS FINANCING

2.1 Public provision of education in Vietnam

This section provides background on the current system of general public education (grades 1-

12) in Vietnam.4 Public education in Vietnam is divided into three major levels: elementary (1-

5), lower secondary (6-9) and upper secondary (10-12). While public schools are totally funded

by the government, non-public schools fall into three types: semi-public, people-funded, and

private. Semi-public schools that are still owned and managed by public authorities use existing

classrooms and facilities, but self-finance primarily from parental contributions for most of their

operating expenditures such as salaries for newly-hired teachers and increasing salaries for

experienced teachers (Nguyen, 2004, 437). People-funded schools are owned, managed and

financed by non-governmental organizations; schools of the third type are privately funded

(World Bank, 1996).

<FIGURE 1 HERE>

Figure 1 shows the downward trend in absolute enrollments starting in the academic year

of 2002-2003. The decrease in total enrollments comes mostly from a decline in elementary

students that is sufficiently dramatic to more than offset rises in lower secondary and upper

secondary enrollments. The decline in enrollments is far from surprising given an aging

population. According to World Bank (2008), the proportion of the population aged between 0

and 14 declined from 43.11 percent in 1975 to 28.79 in 2006. Figure 2 indicates that the primary

enrollment in non-public educational institutions has represented less than 0.5 percent of the total

primary students since the academic year of 2001-2002. While the share of non-public students

increases to less than 3 percent at the lower-secondary level, it jumps up to 33 percent for the

upper-secondary level. Non-public upper secondary students, however, still represent less than

4 See London (2006) for more about historical developments of Vietnam’s educational systems.

3

Page 5: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

one percent of the total enrollment of all three levels. All in all, publicly provided education,

which is the focus of this study, remains the dominant form of schooling in Vietnam.

< FIGURE 2 HERE>

2.2 Financing public education

The lion’s share of expenditures on public education comes from provincial governments. They

spent from 97 to 99 percent of the total expenditures on public education in 2002 (Martinez-

Vazquez, 2005). Their expenditures on public education are financed by tax revenue and

intergovernmental transfers.

Tax revenue in general falls into three categories: taxes assigned 100 percent at the

central level, taxes assigned 100 percent at the provincial level, and shared taxes between the

central and provincial governments. A province’s tax revenue is dependent heavily on the latter

two types of tax levies.5 Taxes shared between the central and provincial governments consist

mostly of value-added taxes (VAT), personal and corporate income taxes, taxes on profit

remittances, and excise taxes;6 taxes assigned 100 percent at the sub-national government level

include land and housing taxes, natural resources taxes, taxes on transfer of land ownership, and

other user charges and fees. However, under the past and current State Budget Laws,7 provinces

in Vietnam have no autonomy over any of the above tax bases or rates imposed upon them.

Therefore, provinces with small own-source revenues cannot raise the pre-determined tax rates

or enlarge their tax bases when they need additional funds for education. The only way for these

provinces to improve their tax levies is to enhance efficiency in collection. However efficient

5 They are hereafter referred to as own-source taxes or revenue for short. Note that as later indicated, the term “own-source” does not mean that provincial governments have authority over tax rates or base.6 The sharing rate is the same across all shared taxes for each province but varies from province to province (Martinez-Vazquez, 2005).7 The current State Budget Law was enacted in 2002 and became effective in 2004.

4

Page 6: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

they are in tax collection, the total amount of tax levies are still severely limited by their narrow

tax base.

Inter-provincial disparity in tax revenue is compensated by a system of intergovernmental

transfers. The transfers in Vietnam come in two categories. The first type of transfers consists of

conditional or targeted transfers. They are made for the implementation of nationally targeted

programs in areas like poverty reduction, housing, and vaccination. The provinces also receive

intergovernmental transfers to address the consequences of unexpected national disasters or

adverse socio-economic event (Martinez-Vazquez and Gomez, 2005, 360). However, there are

no targeted, or categorical, transfers for public education in Vietnam. The second type is

unconditional “balancing” transfers intended to generate greater horizontal fiscal equity by

balancing between provinces’ own-source revenue and their expenditure needs in all sectors

including education. A few provinces with sufficient own-source revenue do not receive

balancing transfers.8 According to Martinez-Vazquez (2005, 42), the transfers have created

significant equalization but substantial disparities remain. In other words, differences in

provincial structural endowments are too large to be equally compensated by the

intergovernmental transfers.

What is of great interest is the central government’s current lack of a sophisticated

enough formula to compute the differential costs of education across provinces. The central

government now employs a very simple formula for educational aid distribution whereby the

unit cost per pupil is based on total population per province (Nguyen et al., 2001). Despite

special programs aimed at providing extra funding for mountainous and remote areas, the

adoption of the unsophisticated distribution formula leaves high-cost and poor provinces with

8 According to data provided by Vietnam’s Ministry of Finance (See the link in the data section), while all provinces received some balancing transfers in 2002, 15 out of 64 provinces did not get the transfers in 2005. However, the eligibility of these fifteen provinces for targeted transfers stays unchanged.

5

Page 7: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

insufficient funds for education, all else being equal. The insufficient funding may be

compounded at the provincial level. While targeted transfers are aimed at other programs than

public education, balancing transfers that include public education aid can in fact be used for any

municipal service. Although the current State Budget Law regulates that local governments will

need to spend eighteen percent of their total budgets on education by 2005 and 20 percent in

2010, the lack of enforcement mechanisms makes full compliance with the regulation highly

questionable (Martinez-Vazquez, 2005, 52). The lack of enforceability is highly likely to result

in even less spending on education when fiscally poor provinces have to apportion their scarce

resources between education and other public services.

3. CONCEPTUAL FRAMEWORK OF THE COST-FUNCTION APPROACH

The government of Vietnam has multiple choices to improve its education aid distribution

formula. In developed countries like the United States, calculating the resources needed for

schools and districts to achieve a particular student performance level involves one of the three

major methodological approaches: the successful-school (or “evidence-based”) approach, the

professional judgment approach, and the cost-function approach. Under the successful-school

approach, policy makers identify a set of high-performing schools and base estimates of the cost

of providing a high-quality education on the lowest level of per-pupil spending among them

(Imazeki and Reschovsky, 2005). Adopting the professional judgment approach, a government

can organize several teams of educators to identify components needed to provide an appropriate

education. Costs are then assigned to each of the agreed components (Rubenstein and Picus,

2003, 66). The cost-function approach is chosen as the focus of this paper for three principal

strengths. First, despite challenges to the legitimacy of the cost-function approach (Hanushek,

6

Page 8: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

2005, 2006; Rice, 1997), this methodology can exhibit reliability and key types of validity

(Baker, 2006; Duncombe, 2006). That is why this methodology has been widely used for both

developed countries, especially the United States,9 and developing countries (Tsang, 1988,

1997). In fact, the cost-function approach used in this paper demonstrates a high degree of

validity and reliability in all feasible tests.10 Second, it can incorporate a quite comprehensive list

of factors affecting provinces’ cost of obtaining a given education standard, thus improving the

equity of adequacy. Third, it can be used to predict the total required spending for various

performance goals under different assumptions about the efficiency of provinces. The second

and third strengths of this approach are exactly what the government of Vietnam needs to

address the two problems with its education finance system earlier identified.

The following conceptual framework for the cost-function approach is largely adapted

from the traditional education cost studies. Let C be the total cost per pupil. Owing to

inefficiencies and/or funding inadequacies, the unobserved cost is different from observed

expenditures. Following standard practice (Downes and Pogue, 1994; Duncombe and Yinger,

1998, 2005, 2007a, 2007b; Reschovsky and Imazeki, 1998, 2001, 2003) with some modifications

for Vietnam’s different circumstances,11 I have assumed that cost depends in multiplicative ways

9 Cost-function studies have been quite popular in education finance research in the United States where the studies of these types have been carried out at least in thirty states (Hoff, 2005).10 Of the three types of reliability mentioned in Duncombe (2006), I can only evaluate the internal consistency reliability. Since this is the first cost-function study for Vietnam, I cannot test my estimates’ interrater reliability that requires several studies. Nor can I assess the test-retest reliability because two years of data used in this study cannot be divided into subperiods. The limited data also make it impossible to evaluate the method’s predictive validity although all of the remaining three validity types, namely internal validity, statistical conclusion validity, and construct validity, will be herein addressed.11 While traditional cost studies may incorporate the median voter model into their modeling framework, the median voter model is irrelevant in the case of Vietnam. The entire provincial process is not subject to popular votes. Rather, the provincial budgets are internally determined (Nguyen et al., 2001). Lack of public involvement in the educational budgeting process makes it implausible to estimate the median voter-based demand regression which is often conducted together with the cost model.

7

Page 9: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

on student performance (S), and other cost factors (N) such as student-body characteristics, and

family involvement.

(1 ) C=Sα N β

where α and β are elasticities to be estimated. Provinces’ operational expenditure per pupil (E) is

a function of C and efficiency (e).

(2 ) E=Ce

Efficiency refers to the degree of resources a production process consumes to yield a

certain output given the current technology. The process is considered to be efficient if it uses the

minimum resources. Inefficient production processes take more than minimum resources. Cost-

function studies identify three important features of efficiency in education finance (Duncombe

and Yinger, 2007; Nguyen and Yinger, 2008). First, higher spending on education does not

necessarily imply inefficiency. Higher spending could be driven by factors outside the

jurisdiction’s control. Provinces with a higher concentration of low-income or ethnic minority

students tend to spend more to obtain the same level of performance. It also costs more for

structurally disadvantaged provinces to attract good teachers. Second, efficiency depends on the

analyst’s selection of outputs. While a province is considered inefficient from the perspective of

certain outputs, it may be efficient if the analyst chooses other outputs. Efficiency is thus

inevitably sensitive to, or dependent on, outputs in any production process with input sharing and

multiple outputs. Public education is such a production process. Educational outputs, such as

student performance on math, physics, English, and athletics, share the same inputs, namely

teachers, classrooms, teaching aids and so forth. Beyond use of outmoded technologies, and

other forms of waste, inefficient production of student performance in math and physics could

8

Page 10: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

simply be due to a province’s channeling more funding than other provinces into alternative

educational outputs.

Third, unobserved efficiency can be modeled. Following other cost-function studies

(Duncombe and Yinger, 2007; Nguyen and Yinger, 2008), I assume efficiency to be a function

of income-based factors influencing provincial officials’ behavior with regard to efficiency of

production, variables indicating residents’ incentives to monitor the officials (M), and other

efficiency-related variables (L). Specifically, the income-based factors consist of residents’

income (Y), balancing transfers per pupil (T), and provinces’ own-source revenue per pupil (R).

All of the efficiency variables are represented in equation (3).

(3 ) e=Y δ T fδ Rσ M φ Lω

where δ , σ , φ, and ω are elasticities to be estimated, and f is the “flypaper effect.”12 Without f,

both balancing transfers (T) and residents’ income (Y) are supposed to have the same income

effect (δ ) because they both create identical outward shifts of provinces’ budget line (Rosen,

2002, 502). However, research into intergovernmental transfers in developing countries has

found empirical evidence of the “flypaper effect” phenomenon whereby transfers to local

governments tend to lead to more local spending than an equivalent increase in local residents’

income (Lalvani, 2002; Melo, 2002). Put differently, as flies on flypaper, transfer money sticks

where it hits.

A rise in provinces’ revenue may induce officials to spend on other areas of school

activities that are not included in educational outputs measured by the researcher, thereby leading

to higher inefficiency. Greater provincial inefficiency may also result from an increase in

residents’ income that weakens their incentives to monitor provincial officials and boosts their

demand for a broader set of educational objectives. While provincial officials’ response to

12 See Hines and Thaler (1995) for arguments of the flypaper effect.

9

Page 11: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

increased revenue is expected to be more or less similar to that of educational officials in

developed countries, it is of great interest to see whether local residents in Vietnam have any

incentives to monitor provincial educational operations, and whether they in fact make any

efforts to do so for two reasons. First, unlike other more democratic countries, local people in

Vietnam have no electoral voice in the process of educational budgeting (Nguyen et al., 2001).

Second, lack of tax autonomy for provincial governments and thus of a tax-price mechanism to

reflect local residents’ willingness to pay for public services may compromise residents’ efforts

to monitor provincial authorities.

It is, however, reasonable to hypothesize that a certain degree of the resident monitoring

efforts are still in place. First, despite the lack of independent authority over tax rates and base,

provincial governments in Vietnam can autonomously set some fees and charges.13 Bird and

Vaillancourt (1998, 13) argue that sub-national governments are accountable for their residents

at the margin according to what they pay directly to the governments for local public services.

Second, a certain type of fiscal illusion could be at work. As most taxes are collected by

provincial tax authorities, residents may not be fully aware of which taxes provincial

governments collect on behalf of the central government and which taxes they collect and retain

for their own communities.

Substituting equations (1) and (3) into (2) produces

(4 ) E=Sα N β Y−δ T−fδ R−σ M−φ L−ω

Taking logs of equation (4) yields

(5 ) ln E=α ln S+β ln N−δ ln Y−fδ lnT−σ ln R−φ ln M −ω ln L

Equation (6) is then modified into equation (6) for empirical estimation purposes.

13 Nevertheless, Martinez-Vazquez and Gomez (2005, 359) note that overuse of fees and charges at the local level as a financing option may de facto deny the poorest access to consumption, contravening the country's poverty reduction and fiscal equalization objectives.

10

Page 12: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

(6 ) ln E=k+α ln S+β ln N−δ lnY −fδ ln T−σ ln R−φ ln M−ω ln L+μ+ε

where k is a constant. μ is a time dummy (=1 for 2005 and 0 otherwise) to control for temporal

macroeconomic shocks affecting all provinces, such as improvements in educational technology.

Equation (6) is estimated with robust standard errors (ɛ) to address the possibility of

heteroskedasticity and with clustering by province to account for pooling across years.

4. EMPIRICAL STRATEGY

The key variable of the student performance index in equation (6) is the equally weighted

average of the following three educational outputs: the percent of grade promoters, non-dropout

rates, and graduation rates at the three levels of public education. While the first output is derived

by taking the difference between 100 and the percent of grade repeaters, the second output

measures the percentage of students who do not drop out of school (100 – dropout rates).

Graduation rates are computed from the results of graduation exams organized uniformly across

the country at the end of each of the three education levels.14 The performance index should be

treated as endogenous for two reasons. First, the dependent variable and student performance

may be correlated with some unobserved provincial characteristics. Second, one might be

concerned about possible measurement errors in the components of the student performance

index caused by the data-processing procedures. This endogeneity must be dealt with to remove

potential biases in the cost model, thus improving the cost-function approach’s internal validity –

the most important of all validity types (Duncombe, 2006).

To address this endogeneity-validity issue, I use instrumental variable (IV) estimation

techniques with an instrument for the student performance index. Finding a valid and strong IV

14 More specifically, fifth-, ninth- and twelfth-graders have to take their respective graduation exam. These exams are the only nationally comparable ones.

11

Page 13: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

is a challenging task. It has to pass several tests. First, it should have a conceptual link with the

endogenous explanatory variable. The IV for student performance is the average share of ethnic

minority students in the same region (See Figure 3 for a map of regions in Vietnam).15 This

instrument has a conceptual link with the endogenous variable according to the “copycat” theory

(Case et al., 1993), also called yardstick competition (Besley and Case, 1995), which suggests

that student performance in a jurisdiction is closely linked to neighboring jurisdictions’ student

characteristics. Second, the IV must be uncorrelated with the error term, ε ; simply put, the IV is

not an explanatory variable in equation (6). Although this can be only tested with multiple IVs,16

student characteristics in neighboring provinces are highly unlikely to directly affect a province’s

per-pupil expenditures. Rather, their effect on spending more likely goes through student

performance as the yardstick competition theory suggests. Third, the IV must pass a test for its

strength. As documented in Table 2, the F-statistic for the first stage of the regression model is

equal to over 24.05, which is far higher than the rule-of-thumb threshold value of 10 suggested

by Staiger and Stock (1997).

<FIGURE 3 HERE>

To model vector N in equation (6), I include the share of ethnic minority students,

variables of enrollment and class size, proxies for family involvement, and two cost factors

outside provinces’ control, namely spatial area of provinces,17 and a dummy variable for

provinces receiving no balancing transfers18 (=1, and = 0 otherwise). First, the higher

concentration of minority students represents additional costs incurred by provinces to reach the

15 For similar approaches to finding instruments for student performance, see Duncombe et al. (2008), Duncombe and Yinger (2005, 2007a), and Nguyen and Yinger (2008).16 Even if there are two or more IVs, the validity of the test relies on the assumption that one of the IVs is already exogenous.17 I do not use the population density. Variables of geographical size and population are separately estimated because as later discussed, they have different implications on cost per pupil.18 See footnote 9.

12

Page 14: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

same level of student performance. Vietnam has 54 ethnic groups with Kinh being a Vietnamese-

speaking dominant majority. The other 53 ethnic minorities have their own languages. It costs

provinces more to provide additional resources, e.g., extra Vietnamese classes, for these minority

students to catch up with other Kinh students. In addition, because studies have found that

higher-qualified teachers tend to sort themselves into schools with more racially advantaged

students (Hanushek et al., 2004; Lankford et al., 2002), provinces serving large numbers of

ethnic minority students need to raise teacher salaries to be able to attract and retain qualified

teachers.

Second, education cost is a nonlinear function of enrollment. Cost per pupil depends on

how many students a province has to educate. Economies of size in the production process of

education help larger provinces incur smaller costs per pupil. At some point, expenditure per

pupil starts increasing because educating an additional student will require a larger marginal cost.

Following this logic, the model employs a quadratic specification for the enrollment variable. A

related variable that also has a quadratic specification is the student-per-class ratio. While the

enrollment variable represents economies of size at the provincial level, the pupil-per-class ratio

looks more closely into classrooms.19 An increase in the class size has two opposing effects. On

the one hand, it may reduce expenditures per pupil because having an additional student in an

existing classroom exerts little impact on the budget. On the other hand, it may cost more to help

students achieve the same performance in larger classes (Boozer and Rouse, 2001; Finn et al.,

2005; Krueger and Whitmore, 2001). However, the first effect is expected to be much larger than

the second one, which becomes dominant only after the optimal class size has been reached.

19 The threat of collinearity between enrollment and class size is unwarranted. The correlation coefficient of the two variables is only about 0.5, which is far below the 0.9 rule-of-thumb threshold for a threat of collinearity (Griffiths et al., 1993, 435).

13

Page 15: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Third, proxies for the degree of family involvement are the provincial-level proportions

of senior citizens and home-staying mothers who are mostly likely to be unemployed. The effect

of home-staying mothers on cost per pupil is unpredictable. Although literature consistently

shows that greater parental involvement can lead to better student achievement (Fan and Chen,

2001; Jeynes, 2003), and thus to smaller cost per pupil, scholars present inconclusive empirical

evidence on employed and unemployed mothers’ level of involvement in their children’s

education. While Muller (1995) find that unemployed mothers are the most involved in their

children’s education relative to part-time and full-time employed mothers, Zick et al. (2001)

provide evidence that employed mothers tend to engage more in reading/homework activities

with their children than unemployed ones. Unlike home-staying mothers, senior citizens have an

expected negative impact on cost per pupil. The inclusion of the proportion of senior citizens

who are usually grand-parents is justified from the traditional structure of Vietnamese families.

Living with the families, grand-parents are expected to spend more time with their grand-

children, to ameliorate the negative effects of economic stress on teen outcomes (Botcheva and

Feldman, 2004), and to contribute more to families and communities in education (Strom and

Strom, 1995).

Fourth, provinces’ geographic size, which lies outside the control of provinces,20 is

predicted to affect educational cost as well. Larger provinces are more likely to spend more on

students for the same level of performance. Low-income students in rural areas with inadequate

infrastructure often walk long distances to schools. The larger the provinces, the greater the

number of students who skip classes and thus have low academic performance and are

20 This variable may have efficiency implications as well. The larger the province, the farther the schools are from the monitoring radar of the provincial headquarters, and the schools may be more inefficient. However, geographical size is considered to be more of a cost factor because the cost effect is expected to be much larger than the efficiency effect.

14

Page 16: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

correspondingly more likely to drop out. All else being equal, these provinces may spend more

resources either keeping their students from dropping out of school, or helping them catch up

with their peers in other provinces. Finally, the dummy variable of fiscally capable provinces is

included because provinces receiving no balancing transfers are hypothesized to have some

unobserved characteristics that differentially affect their educational cost per pupil.

The income-related variables of efficiency are the log values of local revenue per pupil

(R), balancing transfers per pupil (T) and average monthly income (Y). As discussed earlier,

holding student performance constant, provinces with greater fiscal capacity are more inefficient

because of their relatively higher spending on extra-curricular activities not measured by the

student performance index and/or other forms of waste. Unlike an increase in the revenue of

provincial governments, a rise in residents’ monthly income affects provincial inefficiency in

multiple ways. First, greater income may induce residents to demand a wider range of

educational outcomes and thus lead to higher provincial inefficiency. Second, larger income

engenders substitution and income effects. On the one hand, higher income induces people to

substitute away from monitoring efforts and work more, which can result in greater provincial

inefficiency. On the other hand, a rise in income may create incentives for people to work less

and intensify monitoring efforts. Previous studies find that residents’ higher income is strongly

correlated with increased inefficiency (Eom and Rubenstein, 2006).

Two other variables in M that follow the above line of efficiency-related reasoning are

the proportion of college graduates and of employed people out of the province’s total

population. Employed and college-educated people who tend to have higher income have to

make trade-offs between working on the one hand and monitoring and spending more time with

their children at home on the other. They also tend to have a greater demand for other

15

Page 17: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

educational objectives than those included in the performance index. Another variable modeling

residents’ incentives to monitor in M is the share of population that owns a house. It benefits

house owners to devote more of their time to monitoring provincial officials’ efficiency because

school quality is theoretically capitalized into property values. In fact, several recent studies have

found empirical evidence on the capitalization of school quality into house values in developed

countries (Black, 1999; Clapp et al., 2008; Downes and Zabel, 2002; Figlio and Lucas, 2004;

Gibbons and Machin, 2003; Zahirovic-Herbert and Turnbull, 2008) and other emerging countries

like Singapore (Chin and Foong, 2006).21 The theory of school-quality capitalization was tested

to see if Vietnamese homeowners, out of their concern about school quality and thus their own

house values, also exert more monitoring efforts that lead to more efficiency and hence less cost

per pupil in provinces’ production of education.

The final two efficiency-related variables in vector L are provinces’ government

transparency and their population size. Transparency in government operations is likely to have

implications for efficiency. More transparency in the principal-agent relationship can make the

agent’s behavior more accountable and thus more efficient (Holmstrom, 1979). However,

transparency does not come without a price. Transparency may also have costs (Hood and Heald,

2006, 166). To test if the benefits of transparency outweigh its costs, a proxy measure of the

degree of government transparency which will be discussed at greater length in the following

section is added in vector M of equation (6). Vector L also includes the log of provincial

population.22 Larger communities are expected to be more inefficient. More people will have

21 See Yinger et al. (1988) and Ross and Yinger (1999) for reviews of earlier school quality capitalization studies.22 The model does not include teacher salaries as an explanatory variable due to the unavailability of the data. The inclusion of the population size, however, compensates partly for that missing variable. Controlling for geographical size, more densely populated areas tend to be correlated with higher teacher salaries (Chambers, 1995).

16

Page 18: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

more diverse interests in and demands for different educational outcomes uncovered in the

performance index. Additionally, larger localities require bigger and more complex governing

structures which are usually associated with declining efficiency (Afonso et al., 2005), and thus

with higher educational cost.

5. DATA AND MEASURES

Equation (6) is estimated with data in 2002 and 2005. The websites of Vietnam’s Ministry of

Finance (MOF) (http://www.mof.gov.vn) and General Statistics Office (GSO)

(http://www.gso.gov.vn) supply the majority of data for the estimation (See Table 1 for

descriptive statistics). Specifically, while the MOF provides data for provincial local revenue and

balancing transfers, the GSO is a source for data on student enrollment, the percent of ethnic

minority students, class size, average monthly income, areas and population. The 1999 Census

supplies data on the share of college graduates and the percent of households that live in

dwellings constructed from concrete. The Ministry of Labor, Invalids, and Social Affairs

(MOLISA) (2006) provides data on the percent of the employed, home-staying mothers, and

senior citizens aged 65 and above.

< TABLE 1 HERE>

The dependent variable is operational expenditures on education by provincial

governments. Data on provincial educational spending in 2002 are taken from Martinez-Vazquez

(2005), which provides the same, but complete, data as on the MOF websites. Educational

expenditures by province in 2005 are retrieved from the MOF-hosted final accounts of the

provincial budgets.23 The number of provinces changed from 61 to 64 at the beginning of 2004

23 The data can be found at http://www.mof.gov.vn/Default.aspx?tabid=5733.

17

Page 19: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

when three provinces were split up into two smaller ones each.24 The dataset used for estimation

includes only 58 provinces excluding those six reorganized provinces because the data are

missing in 2002 for three newly-formed provinces, and inconsistent between 2002 and 2005 for

three existing provinces.25

A key explanatory variable in this equation is the student performance index, S. As

previously specified, the index is derived from the average shares of grade promoters, non-

dropout rates, and graduation rates at the three educational levels. The index proves to have a

high degree of reliability and construct validity.26 We do not weight the index because every

component carries a certain degree of importance in the view of policy makers, local residents,

and parents. There have recently arisen heated public discussions about increasing rates of grade

repeaters and drop-outs (GSO, 2008). Completing universal primary and lower secondary

education is among the strategic goals explicitly stated in the national EFA Action Plan (GVN,

2003, 27). Upper secondary graduation rates are always a focus of public opinion every summer.

National printed and broadcast media run multiple articles and programs to help students prepare

for their high school graduation exams for several months every year. In addition, reports on

across-province high-school graduation rates attract substantial attention from the public.

24 Lai Chau was split into the smaller Lai Chau and Dien Bien, Dak Lak into the smaller Dak Lak and Dac Nong, and Can Tho into the municipal Can Tho and Hau Giang (Nguyen, 2008).25 Excluding missing data does not cause estimation bias when they are missing at random (Wooldridge 2003). Data for the six reorganized provinces can be considered to be randomly missing in this context because the reorganization has nothing to do with educational cost. Malesky (2005) argues that provincial divisions that have increased the number of provinces by 60 percent since 1990 come from the gerrymandering strategy adopted by reformists who wanted to free reform-oriented provinces from provinces dominated by the state-owned enterprise sector.26 I attempt to examine the construct validity and internal consistency reliability of this key variable which is the most rich in historical data. The data are quite temporally stable and internally consistent. For instance, following Duncombe’s (2006) approach, the average, mean, median and standard deviations of upper secondary graduation rates across 2002-2006 are either almost the same or differ as much as 3 percentage points. For the internal consistency of this linear composite, I compute the Cronbach’s alpha coefficient (See Greene and Carmines (1980) for a review of reliability coefficients). The value of this alpha coefficient is 0.6, which meets the minimum widely-accepted threshold (Garson 2008).

18

Page 20: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

The transparency index is derived from the Vietnam Provincial Competiveness Index

(PCI) documented in VNCI (2005, 2006). Although the index,27 one of the ten sub-indexes

comprising the PCI, is designed to measure access to public information for the private sector

(VNCI, 2005, 2006), I assume that governments with transparent mechanisms for private-sector

development are also as transparent in other public services. While the 2003 PCI had 42

provinces being surveyed, the 2005 PCI surveyed all 64 provinces in Vietnam.28 Since the PCI

was not computed until 2003, transparency data for 2002 take the values of the 2003 survey for

42 provinces and of the 2005 survey for the remaining provinces. This sensible, though

imperfect, extrapolation procedure provides a relatively accurate measure of transparency in

provincial governments. First, small changes in the transparency indexes are observed across

provinces between the 2003 and 2005 survey. Second, the years of 2003 and 2005 are still within

a single term of provincial governments. Therefore, there could be very little sudden major

change in government transparency.

6. DISCUSSIONS

6.1 Regression results

Table 2 documents how various factors have significant effects on provinces’ per pupil cost of an

adequate education. Specifically, the elasticity on the student performance index is highly

significant. Provinces are predicted to have a 5.56 percent increase in spending per pupil to

obtain a one-percent increase in student performance. Like education finance in other developed

countries, the result supports the importance and effect of money in raising student performance

27 Transparency is ranked from 1 to 10 with 10 as being the most transparent. I do not log this index in the regression for ease of interpretation. The results are almost exactly the same with either way.28 While the 2006 PCI is based on their questionnaires in the late 2005 and early 2006 (VNCI, 2006, 5), the PCI of 2005 are derived from surveys conducted in the early spring of 2003 (VNCI, 2005, 45).

19

Page 21: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

in Vietnam. The greater-than-unity elasticity of student performance indicates decreasing returns

to quality scale for student performance. Additionally, the elasticity is larger than that found in

other cost education studies in the United States which is nearly 3 (Duncombe and Yinger

2007b), or around 2 (Nguyen and Yinger, 2008; Reschosky and Imazeki, 2001). This implies that

it costs provinces more to raise an additional percent of student performance.

< TABLE 2 HERE>

Panel 2 of Table 2 presents strong empirical evidence that disadvantaged students have a

significant impact on cost per pupil. Specifically, a one percentage point increase in the share of

ethnic minority students requires a nearly 0.2 percent rise in the provinces’ per-pupil cost if the

same student performance is to be kept. The panel also indicates evidence that there is a

significant U-shaped relationship between cost per pupil and student enrollment. Because current

provincial enrollments are still very far from the minimum cost enrollment,29 provinces can still

benefit from economies of enrollment size.

Class size has a significant quadratic effect on cost per pupil. Following the approach to

compute the minimum cost enrollment, the minimum cost class size is estimated at about 40

pupils per class. Given the optimal class size, an increase in class size will lead to a reduction in

cost per pupil in 38 provinces. On the contrary, the spending per pupil of 20 provinces will go up

together with the rise in class size.30 Also, increases in the geographic size have significantly

positive impacts on spending per pupil. A one percent rise in area is associated with a 0.12

percent increase in per-pupil spending. The empirical result also shows that fiscally wealthy

provinces spend 0.1 percent more per pupil to provide an adequate education.

29 I use the coefficients on enrollment variables in Table 2 to derive the minimum cost enrollment. The log of per pupil expenditures is first differentiated with respect to logged enrollment, or d ln ( E)/d ln (enrollment )=−2.3+[ (2×0.05 )× ln (enrollment )]. The minimum cost enrollment is then equal to e [2.3 / (2×0.05 )], or around 3.2 billion.30 The minimum, average, and maximum class sizes in the dataset are 30.3, 38.4 and 44.5.

20

Page 22: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

The two variables for family involvement show negative effects on cost per pupil. While

the effect of home-staying mothers is not significant, provinces with a higher concentration of

senior citizens spend less per pupil, other things being equal. A one percentage point increase in

the share of senior citizens results in a 2.5 percent decline in spending per pupil. Seniors-cum-

grandparents in Vietnam who usually live in extended families tend to substitute for children in

doing household chores (Liu, 1999), leaving the latter with more study time and thus better

school performance.

All of the income-related variables present expected positive effects on provincial

inefficiency, leading to an increase in cost per pupil. Specifically, a one-percent increase in per

pupil provincial revenue or average monthly income is predicted to result in a 0.04 or 0.21

percent increase in per-pupil cost respectively. Higher balancing transfers also generate greater

provincial inefficiency although the effect is smaller than that caused by increased provincial

revenue or residents’ income. A one-percent rise in balancing transfers leads to an approximately

0.01 percent increase in spending per pupil. In addition to the earlier discussed channeling of

transfers to unmeasured educational outcomes, inefficiency may come from various strings

attached to the transfers (Fritzen, 2006, 21). The derived flypaper effect of 0.04 for balancing

transfers implies that VND 1000 of the transfers have the same impact on inefficiency as an

increase of VND 40 in residents’ income. This small effect and statistical insignificance indicate

that the balancing transfers do not “stick where they hit” in educational spending in Vietnam.

The result can be explained from the general-purpose nature of balancing transfers.

Panel 4 of Table 2 provides little evidence that local residents in Vietnam devote much

effort to monitoring provincial officials. Like other developed countries, for the appreciation in

property values caused by the school quality capitalization, owners of concretely structured

21

Page 23: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

housing show greater efforts to monitor provincial officials, thus making them more efficient.

However, the effect of this variable is ten times smaller in absolute values than those of the

proportions of college graduates and the employed. Provinces with a higher concentration of

these latter two demographic groups are less efficient, and thus spend more to achieve the same

level of student performance. The positive effect of the proportion of college graduates on cost

per pupil is contrary to what has been found in previous American-based cost studies where

more local residents with college education raises school district efficiency (Duncombe and

Yinger, 1997; Duncombe et al., 1997; Eom and Rubenstein 2006; Grosskopf et al., 2001), and

thus reduces expenditures per pupil (Dodson and Garrett 2004; Duncombe et al., 2008). Given

the current little room for participation in public policy-making and a lack of tax-price

mechanisms, people with relatively higher opportunity costs of time choose to substitute away

from monitoring provincial officials.

The two efficiency-related variables in Panel 5 have highly significant and expected

impacts on cost per pupil. More transparent provincial governments incur significantly less

spending per pupil than others. A one-point increase in the transparency index is associated with

a 1.7 percent decrease in cost per pupil. As expected, more populous provinces incur more cost

per pupil than others because they may have either a wider demand for educational outputs, or

more complex efficiency government structures.

Those estimates from the cost-function approach have quite strong statistical conclusion

validity which is defined as “a statistical relationship between variables in a potential cause-

effect relationship” (Duncombe, 2006). As explored in Baker (2006) and Duncombe (2006), this

validity can be tested by examining the relationship between funding gaps and student

performance. The funding gap refers to the difference between the predicted costs to meet a

22

Page 24: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

performance level31 and actual spending. The correlation coefficient of this relationship is found

to be negative and equal to -0.36, which is almost equal to -0.4 reported in Duncombe (2006).

The negative sign of the coefficient indicates that provinces with a relatively higher funding gap

perform worse, and that the cost-function estimation on average targets the right provinces.

6.2 Estimation of required spending

The cost-function estimation results reflect various determinants of the per-pupil cost of a given

education which are not included in the government’s current distribution formula. The results

can then be employed to derive cost and efficiency indices. The purpose of the indices is twofold.

First, they present in a more concrete way how significantly provinces vary in their per-pupil

cost for a given educational standard. Second, they are intended for estimating the total costs of

the government with different performance goals and efficiency levels.

Since the definition of an acceptable education is not publicly available, the cost and

efficiency indices are derived with the national average performance for illustrative purposes. To

compute a cost index,32 I multiply actual provincial and national average values by the

coefficients of all cost factors and of the other variables respectively. The antilog of the sum of

all the products is how much a province of average efficiency33 is predicted to spend for the

nation’s average performance level. The final cost index is equal to the ratio of each province’s

predicted spending to predicted cost incurred by the province with average characteristics. The

31 This validity assessment is not dependent on the choice of a performance standard which only becomes relevant in the estimation of required spending. Different standards just lead to different levels of the funding gap, but do not change the slope that is of primary interest. 32 Computations of education cost and efficiency indices are already done in other American-based cost studies. Some of the recent studies are Duncombe et al. (2003), Duncombe and Yinger (2007), Imazeki and Reschovsky (2005), Nguyen and Yinger (2008), and Taylor et al. (2002).33 The efficiency level has to be constant across all provinces so that the cost index reflects only variation in cost factors. Use of the average level of efficiency is a standard practice in other cost-function studies because it facilitates the interpretation of the cost index.

23

Page 25: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

cost index derived this way indicates the level of cost per pupil for a province of average

efficiency to obtain a national average performance compared to an average province. A higher

cost index implies greater cost. For example, a cost index of 2 means that it costs the province

twice as much to attain the average performance as an average province.

Similarly, the efficiency index is calculated by multiplying the coefficients on the

efficiency variables by actual provincial values and all other coefficients are multiplied by the

national averages. Taking the antilog of the sum of these products plus the residual from the cost

regression34 gives us predicted spending in a province of actual efficiency but with national

average performance and costs. An efficiency index is derived by dividing the predicted

spending by the minimal spending by the most efficient province. The higher the index, the more

efficient the province. The province with an index of unity is considered to be the most efficient.

<TABLE 3 HERE>

Three principal messages can be taken away from Table 3, which reports the distribution

of cost and efficiency indices across eight regions in Vietnam. First, regions vary substantially in

their cost per pupil to obtain the national average performance. More specifically, structurally

disadvantaged provinces in general have higher cost indices relative to better endowed ones. For

instance, mountainous provinces with a very high concentration of ethnic minorities in Central

Highlands and North West have the average cost indices of 1.90 or 1.63. These indices are much

higher than those of 1.02 and 0.70 constructed for relatively more structurally-endowed

provinces in South East and Red River Delta35 respectively. In other words, provinces in Central

34 I assume that all unexplained variation in spending across provinces comes from variation in efficiency, not in costs.35 Vietnam’s two biggest cities are located in these two regions with Ha Noi in Red River Delta and Ho Chi Minh City in Southeast.

24

Page 26: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Highlands and Northwest need to spend nearly twice or over time and a half more respectively

for an average level of student performance than the average province.

Second, relative to efficiency, cost factors account mostly for the variation in per-pupil

spending across regions. While the standard deviation of the cost indices is 0.51, the efficiency

indices have a much smaller deviation of 0.07. Although the smaller variation of regional or

provincial efficiency may be explained partly by the construction of the index,36 it is appropriate

to infer that provincial differentials in spending per pupil to achieve the national average

standard come to a large extent from cost factors.

Third, despite its relatively smaller variation, the efficiency indices show that more

fiscally capable provinces tend to be more inefficient. For example, the efficiency indices of 0.45

and 0.43 for relatively richer provinces in Red River Delta and South East respectively are lower

than those in other regions except Mekong River Delta. As previously discussed, these better-

endowed provinces can spend more on other educational outputs or afford to be more wasteful.

The cost and efficiency indices can be of great help for the government of Vietnam to

estimate the total minimum spending necessary to attain a particular student performance goal.

Although the government of Vietnam, as noted earlier, currently does not set performance targets

explicitly, it instead has a spending target. It aims at raising the ratio of education spending to

GDP to 4.2 percent in 2015 from 3.737 percent in 2002 (GVN 2003, xi). As an illustration, I

attempt to compute how much the Vietnamese government needs to spend if student

performance is to be brought at least up to the 50 percentile level for below-average provinces.

<TABLE 4 HERE>

36 The efficiency index has an upper bound of unity whereas the cost index can be greater than one.37 This percentage is a little higher than my calculation for 2002. It could be due to either my exclusion of capital spending, use of different datasets, or both. The implications from Table 4, however, still hold.

25

Page 27: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

The first column in Table 4 shows that actual educational expenditures in relation to GDP

increased from 3.21 percent in 2002 to 3.83 percent in 2005. However, if the student

performance of below-average provinces is to be raised to the national average, the government

is predicted to spend VND 2,608 billion more in 2005. Additional funding boosts the ratio of

educational spending to GDP to 4.14 percent nearing the target of 4.2 percent in 2015. However,

if provinces can somehow improve their efficiency,38 it costs the government much less to

achieve the same performance level. Columns 3 in Table 4 shows that when below-average

provinces enhance their efficiency to the national average, the government can save VND 1,108

or 2,268 billion in 2002 and 2005 respectively. As documented in Column 4 of Table 4, cost

savings are even much larger if the provincial efficiency can be raised up to the 75 percentile

level. All in all, while raising student performance involves greater costs, provinces’ improved

efficiency can substantially reduce the total government expenditures on education.

7. CONCLUSIONS

Vietnam’s existing education finance system is compromised with two major problems. First, the

central government’s simplified population-based formula for allocating educational aid fails to

take into account various cost and efficiency factors influencing provinces’ expenditures on

education. As the regression shows, money does matter if provinces are to raise their student

achievement, this problem with the education aid formula may, holding efficiency constant,

leave high-cost poor provinces with inadequate resources to help their students reach some

minimum target. Second, the government lacks a useful tool to estimate minimum spending to

attain a certain student performance. This problem may explain the current lack of a publicly

38 The derivation of Columns 3 and 4 assumes that provinces do not incur other costs involved in improving their efficiency.

26

Page 28: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

available “acceptable” standard earlier noted. Without a reliable estimation tool, the government

may not want to commit itself to the standard in case its budget falls short.

As the first-ever education cost-function analysis for Vietnam, the paper presents how the

cost-function estimation can be used with great reliability and validity to address these two

problems. For the first problem, the cost-function estimation shows that a multitude of

differential efficiency and especially cost factors make provinces vary substantially in their per-

pupil cost to obtain a certain performance standard. First, student-body characteristics such as the

shares of ethnic minority students and student enrollment are found to exert highly significant

effects on cost per pupil. Second, I also find evidence on the effect of provinces’ geographic

structures and family involvement, especially of the share of grand-parents, on the cost to attain a

educational target. Third, provincial officials are responsive to revenue incentives. Greater

revenue makes them more inefficient. Fourth, local residents are more likely to respond to

increased income in such a way that negatively impacts provinces’ operational efficiency. Higher

income induces them to work more and shy away from monitoring the efficiency of relevant

provincial agencies responsible for providing public education. This income effect could be due

primarily to the lack of tax-price mechanisms and of a channel for voice and active participation

in public affairs in Vietnam. These estimation results can then be employed via the cost and

efficiency indices to help the central government with the second problem. The indices can be

used for reliable estimates on the necessary expenditures to reach a performance goal.

Given the cost-function approach’s potential as a solution to the two problems, the

government of Vietnam should consider incorporating it into its process of distributing

educational aid. Once applied, the cost-function method can help the central government

generate equity in adequacy of access to education among children independent of their

27

Page 29: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

residence location, enable it to make reliable estimates of total funding needed for a performance

goal, and make its budgeting process more transparent. Although adoption of the cost-function

approach in the government’s educational aid distribution procedures will be a necessary start,

issues regarding the design and implementation of education aid are left unanswered.

Specifically, should the aid be categorical or still included in the total balancing transfers?

Should it be of a matching or lump-sum type? Since efficiency, as this paper shows, matters in

reducing educational cost, the government might want to perform cost-benefit analyses with a

view to improving provinces’ efficiency. One should note that efficiency does not come easy.

Will the cost savings in education from improved efficiency be sufficient to compensate for

spending needed to make provinces more efficient? Despite these as-yet-unanswered questions,

the paper demonstrates the strengths of the cost-function estimation that can be used to allocate

educational transfers to provinces more adequately (and thus more equitably), and to provide

reliable estimates of minimum spending for student performance goals given Vietnam’s unique

institutional characteristics.

28

Page 30: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

8. APPENDICES

Figure 1: Public education enrollment over the years

99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-070

2000000400000060000008000000

100000001200000014000000160000001800000020000000

Total ElementaryLower Secondary Upper SecondaryMillions

Source: Vietnam’s Ministry of Education and Training (http://www.moet.gov.vn/?page=11.10&view=9264)

01-02 02-03 03-04 04-05 05-06 06-070

5

10

15

20

25

30

35

40

Figure 2. Share of students in non-public schools at three levels

Primary Lower secondary Upper secondary

Source: Author’s calculations from the data at the same source as Figure 1.

29

Page 31: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Figure 3. Regions in Vietnam

Source: http://en.wikipedia.org/wiki/Image:VietnameseRegions.png

30

Page 32: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Table 1. Descriptive Statistics (2002 and 2005)

Variables Mean St. Deviation Min Max

Log of operating expenditures (in VND million) per pupil 0.194 0.318 -0.432 0.941Log of student performance index 4.471 0.109 4.155 4.603

Cost variablesPercent of ethnic minority students 19.415 27.412 0.041 96.722Log of enrollment 12.414 0.480 11.102 13.711Log of enrollment squared 154.324 12.014 123.251 187.981Log of student per class ratio 3.646 0.078 3.412 3.796Square of logged student per class ratio 13.302 0.564 11.643 14.413Percent of senior citizens 8.088 1.935 4.132 13.088Percent of home-staying mothers 2.738 2.613 0.108 11.030Log of area 8.256 0.759 6.713 9.711Fiscally capable provinces (=1) 0.241 0.430 0 1

Efficiency variablesIncome-related variablesIncome-related variablesLog of provincial revenue (in VND million) per pupil 1.134 1.331 -3.414 5.118Log of balancing transfers (in VND million) per pupil -0.717 4.435 -13.694 2.221Log of average monthly income (in VND thousand) 6.880 0.302 6.291 8.066

Resident monitoring variablesResident monitoring variablesPercent of college graduates 2.150 2.241 0.555 15.819Percent of people owning in concretely constructed dwellings 57.988 25.654 14.017 96.719

Percent of employed people 94.488 0.691 92.800 96.240

Other efficiency-related variablesOther efficiency-related variablesTransparency index 5.130 1.209 2.53 8.500Log of population 13.946 0.531 12.565 15.592

Instrumental variableAverage percent of ethnic minority students in neighboring provinces of the same region 19.261 22.882 0.720 81.813

Note: There are 116 observations. All monetary variables are in current Vietnamese Dong (VND).

31

Page 33: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Table 2. Cost Regression Resultsa (2002 and 2005)Dependent Variable: Log of operating expenditures per pupil

32

Page 34: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Variables Coefficients t-statistics

Log of student performance indexb 5.5641 3.48***

Cost variablesPercent of ethnic minority students 0.0020 2.14**Log of enrollment -2.3070 -2.90***Log of enrollment squared 0.0525 1.76*Log of student per class ratio -21.8596 -2.01**Square of logged student per class ratio 2.9492 1.97**Percent of senior citizens -0.0257 -2.80***Percent of home-staying mothers -0.0006 -0.09Log of area 0.1006 2.73***Fiscally capable provinces (=1) 0.1044 2.25**

Efficiency variablesIncome-related variablesIncome-related variablesLog of provincial revenue 0.0454 2.84***Log of average monthly income 0.2132 2.17**Log of balancing transfers 0.0079 2.35**Flypaper effect of balancing transfersc 0.0476 2.42

Resident monitoring variablesResident monitoring variablesPercent of people owning concretely constructed dwellings -0.0021 -1.83*Percent of college graduates 0.0252 3.86***Percent of employed people 0.0344 1.84*

Other efficiency-related variablesOther efficiency-related variablesTransparency index -0.0169 -1.71*Log of population 0.8032 4.30***

OthersYear (=1 if 2005 and 0 otherwise) 0.2298 3.62***Constant 19.2252 0.90Adjusted R-squared 0.867Number of observations 116F-statistics in the first stage forLog of student performance index 24.05*** p<0.01, ** p<0.05, * p<0.1a/ The cost model is estimated with the linear IV regression. Robust standard errors (controlling for clustering at provincial level) are used for hypothesis testing.b/ The log of student performance index is treated as endogenous and instrumented with

33

Page 35: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

the average percent of ethnic minority students in neighboring provinces in the same region. c/ The flypaper effect (f) modeled in equation (6) is derived after the regression and equal to the quotient of the coefficient of balancing transfers and that of average monthly income. We use the delta method to conduct hypothesis testing for this non-linear parameter. See Greene (2003, 128) or Cramer (1971, 96) for detailed discussions about the method.

Table 3. Cost and efficiency indexes by region

Regions Cost index Efficiency indexRed River Delta 0.70 0.45North East 1.97 0.60North West 1.63 0.55North Central Coast 0.80 0.47South Central Coast 0.98 0.49Central Highlands 1.90 0.59South East 1.02 0.43Mekong River Delta 0.94 0.41Standard deviation 0.51 0.07

34

Page 36: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Table 4. Actual and estimated spending in different scenariosa(in billions of current VND)

Year Actual

(1)

Scenarios with student performance (S) >= national averageScenario A

Actual efficiency(2)

Scenario BEfficiency>=average

(3)

Scenario CEfficiency>=75 percentile

(4)

2002

Total operating expendituresb on public education 17,190 17,320 16,212 15,355

GDPc 535,672 535,672 535,672 535,672Operating expenditures as a percent of GDP 3.21% 3.23% 3.03% 2.87%

2005

Total operating expenditures on public education 32,159 34,767 32,499 30,502

GDP 839,211 839,211 839,211 839,211Operating expenditures as a percent of GDP 3.83% 4.14% 3.87% 3.63%

a/ Using the cost (CI) and efficiency index (EI) derived from the cost-function estimation, I estimate the necessary total operating expenditures (TE) under three assumptions of efficiency if below-average provinces are to reach the national average performance. The estimated TEs are computed by equation (7) which is based on equations (1) and (2) with relevant modifications.

(7 ) TE=∑i=1

64 [(FA ×CI i {S }

EI i )×TN i]where the per-pupil foundation amount of funding, FA, is assumed to be the national average.39 TN is the actual total enrollment of province i. To raise below-average provinces to the national average performance, the cost index of those districts are re-calculated with the average performance. TE is then computed with this new CI using equation (7) for Scenario A. However, to get TE for Scenarios B and C, equation (7) employs the average or 75-percentile efficiency index for below-average or below-75 percentile provinces respectively.b/ There are two caveats about the data. First, since the dependent variable in the cost function is provinces’ operating expenditures per pupil, all the data presented in this table do not include capital expenditures. However, adding capital expenditures barely changes the share of education spending relative to GDP. Second, the estimated spending includes actual spending by the six provinces left out in the cost regression. Their spending accounts for only five percent of the total.c/ Data are taken from item GDP (current LCU) of World Bank (2008).

39 The per-pupil foundation amount defined as an average is a common practice (See NYSDB (2007) for an example).

34

Page 37: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

9. REFERENCES

Afonso, A., L. Schuknecht, and V. Tanzi. 2005. "Public Sector Efficiency: An International Comparison." Public Choice 123(3): 321-347.

Baker, B. D. 2006. “Evaluating the Reliability, Validity, and Usefulness of Education Cost Studies.” Journal of Education Finance 32(2): 170-201.

Besley, T., and A. Case. 1995. “Incumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstick Competition.” American Economic Review 85(1): 25-45.

Bird, R., and F. Vaillancourt. 1998. "Fiscal Decentralization in Developing Countries: An Overview." In R. Bird, and F. Vaillancourt (eds.), Fiscal Decentralization in Developing Countries. Cambridge: Cambridge University Press, pp. 1-48.

Black, S. 1999. “Do Better Schools Matter? Parental Valuation of Elementary Education.” Quarterly Journal of Economics 114(2): 577-599.

Boozer, M., and C. Rouse. 2001. "Intraschool Variation in Class Size: Patterns and Implications." Journal of Urban Economics 50(1): 163-189.

Botcheva, L., and S. Feldman. 2004. "Grandparents as Family Stabilizers during Economic Hardship in Bulgaria." International Journal of Psychology 39(3): 157-168.

Case, A., H. Rosen, and J. Hines. 1993. "Budget Spillovers and Fiscal Policy Interdependence: Evidence from the States." Journal of Public Economics 52(3): 285-307.

Chambers, J. 1995. "Public School Teacher Cost Differences Across The United States: Introduction To A Teacher Cost Index (TCI)." In W. Fowler (ed.), Developments in School Finance. Washington. DC: National Center for Education Statistics, pp. 19-32.

Chin, H., and K. Foong. 2006. "Influence of School Accessibility on Housing Values." Journal of Urban Planning and Development 132(3): 120-129.

Clapp, J., A. Nanda, and S. Ross. 2008. “Which School Attributes Matter? The Influence of School District Performance and Demographic Composition on Property Values.” Journal of Urban Economics 63(2): 451-466.

Clotfelter, C., H. Ladd, J. Vigdor, and R. Diaz. 2004. “Do School Accountability Systems Make It More Difficult for Low-Performing Schools to Attract and Retain High-Quality Teachers?” Journal of Policy Analysis and Management 23(2): 251-271.

Cramer, J. 2001. Empirical Econometrics. Amsterdam: North-Holland & Publishing Co.

Dodson, M., and T. Garrett. 2004. “Inefficient Education Spending in Public School Districts: A Case for Consolidation?” Contemporary Economic Policy 22(2): 270-280.

35

Page 38: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Downes, T., and J. Zabel. 2002. “The Impact of School Characteristics on House Prices: Chicago 1987–1991.” Journal of Urban Economics 52(1): 1-25.

Downes, T., and T. Pogue. 1994. "Adjusting School Aid Formulas for the Higher Cost of Educating Disadvantaged Students." National Tax Journal 47(1): 89-110.

Duncombe, W. 2006. "Responding to the Charge of Alchemy: Strategies for Evaluating the Reliability and Validity of Costing-Out Research." Journal of Education Finance 32(2): 137-169.

Duncombe, W., J. Miner, and J. Ruggiero. 1997. "Empirical Evaluation of Bureaucratic Models of Inefficiency." Public Choice 93(1): 1-18.

Duncombe, W., and J. Yinger. 1997. "Why is it so Hard to Help Central City Schools?" Journal of Policy Analysis and Management 16(1): 85-113.

Duncombe, W., and J. Yinger. 1998. "School Finance Reform: Aid Formulas and Equity Objectives." National Tax Journal 51(2): 2239-2262.

Duncombe, W., and J. Yinger. 2005. "How Much More does a Disadvantaged Student Cost?" Economics of Education Review 24(5): 513-532.

Duncombe, W., and J. Yinger. 2007a. "Does School District Consolidation Cut Costs?" Journal of Education Policy and Finance 4(2): 341-376.

Duncombe, W., and J. Yinger. 2007b. “Making Do: State Constraints and Local Responses in California’s Education Finance System.” Working Paper.

Duncombe, W., A. Lukemeyer, and J. Yinger. 2003. "Financing an Adequate Education: A Case Study of New York." In W. Fowler (ed.), Developments in School Finance: 2001-02. Washington. DC: National Center for Education Statistics, pp. 127-154.

Duncombe, W., A. Lukemeyer, and J. Yinger. 2008. “The No Child Left Behind Act: Have Federal Funds Been Left Behind?” Public Finance Review 36(4): 381-407.

Eom, T., and R. Rubenstein. 2006. “Do State-Funded Property Tax Exemptions Increase Local Government Inefficiency? An Analysis of New York State's STAR Program.” Public Budgeting & Finance 26(1): 66-87.

Fan, X., and M. Chen. 2001. "Parental Involvement and Students’ Academic Achievement: A Meta-Analysis." Educational Psychology Review 13(1): 1-22.

Figlio, D., and M. Lucas. 2004. “What's in a Grade? School Report Cards and the Housing Market.” American Economic Review 94(3): 591-604.

Finn, J., S. Gerber, and J. Boyd-Zaharias. 2005. "Small Classes in the Early Grades, Academic Achievement, and Graduating from High School." Journal of Educational Psychology 97(2): 214-223.

36

Page 39: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Fritzen, S. 2006. "Probing System Limits: Decentralisation and Local Political Accountability in Vietnam." Asia-Pacific Journal of Public Administration 28: 1-24.

Garson, G. David. 2008. Scales and Standard Measures. 30, 2008. Available at http://www2.chass.ncsu.edu/garson/PA765/standard.htm.

Gibbons, S., and S. Machin. 2003. “Valuing English Primary Schools.” Journal of Urban Economics 53(2): 197-219.

Greene, V., and E. Carmines. 1980. “Assessing the Reliability of Linear Composites.” Sociological Methodology 11: 160-175.

Greene, W. 2004. Econometric Analysis. New Jersey: Prentice Hall.

Griffth, W., R. Hill, and G. Judge. 1993. Learning and Practicing of Econometrics. New York: John Wiley & Sons, Inc.

Grosskopf, S., K. Hayes, L. Taylor, and W. Weber. 2001. “On the Determinants of School District Efficiency: Competition and Monitoring.” Journal of Urban Economics 49(3): 453-478.

(GSO) General Statistics Office. 2008. Drop-Outs: A Worrisome Phenomenon. Available at http://www.gso.gov.vn/default.aspx?tabid=382&idmid=2&ItemID=7036.

(GVN) Government of Vietnam. 2003. National Education for All (EFA) Action Plan: 2003-2015. Ha Noi: Government of Vietnam.

(GVN) Government of Vietnam. 2005. Viet Nam Achieving the Millenium Development Goals. Ha Noi: Government of Vietnam.

Hanushek, E., and M. Raymond. 2004. “The Effect of School Accountability Systems on the Level and Distribution of Student Achievement.” Journal of the European Economic Association 2(2-3): 406-415.

Hanushek, E., J. Kain, and S. Rivkin. 2004. "Why Public Schools Lose Teachers." Journal of Human Resources 39(2): 326-354.

Hanushek, E. 2006. "The Alchemy of “Costing Out” an Adequate Education." Paper presented at the conference "Adequacy Lawsuits: Their Growing Impact on American Education," Harvard University, Cambridge, MA, October.

Hanushek, E. 2006. "Pseudo-Science and a Sound Basic Education: Voodoo Statistics in NewYork." Education Next 5(4): 67-73.

Hines, J., and R. Thaler. 1995. “Anomalies: The Flypaper Effect.” Journal of Economic Perspectives 9(4): 217-226.

Hoff, D. 2005. "The bottom line." Education Week 24(17): 28-36.

37

Page 40: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Holmstrom, B. 1979. "Moral Hazard and Observability." Bell Journal of Economics 10(1): 74-91.

Hood, C., and D. Heald. 2006. Transparency: The key to better governance? Oxford: Oxford University Press.

Imazeki, J., and A. Reschovsky. 2005. "Assessing the Use of Econometric Analysis in Estimating the Costs of Meeting State Education Accountability Standards: Lessons from Texas." Peabody Journal of Education 80(3): 96-125.

Jeynes, W. 2003. "A Meta-Analysis: The Effects of Parental Involvement on Minority Children’s Academic Achievement." Education and Urban Society 35(2): 202-218.

Krueger, A., and D. Whitmore. 2001. "The Effect of Attending a Small Class in the Early Grades on College-Test Taking and Middle School Test Results: Evidence from Project STAR." Economic Journal 111(468): 1-28.

Lalvani, M. 2002. "The Flypaper Effect: Evidence from India." Public Budgeting & Finance 22(3): 67-88.

Lankford, H., S. Loeb, and J. Wyckoff. 2002. "Teacher Sorting and the Plight of Urban Schools: A Descriptive Analysis." Educational Evaluation and Policy Analysis 24(1): 37-62.

Liu, A. 1999. “Children's Participation behavior in Vietnam: Combined Activities.” Asia Pacific School of Economics and Management Working Paper.

Malesky, E. 2005. “Gerrymandering – Vietnamese Style: The Political Motivations behind the Creation of New Provinces in Vietnam.” Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago, IL.

Martinez-Vazquez, J. 2005. “Making Fiscal Decentralization Work in Vietnam.” Andrew Young School of Policy Studies Working Paper 04-04.

Martinez-Vazquez, J., and J. Gomez. 2005. "Effective Decentralization in Vietnam." Proceedings of the Annual Conference on Taxation, 356-361.

Melo, L. 2002. "The Flypaper Effect under Different Institutional Contexts: The Colombian Case." Public Choice 111(3-4): 317–345.

(MOLISA) Vietnam’s Ministry of Labor, Invalids, and Social Affairs. 2006. Statistical data of Employment and Unemployment in Vietnam 1996-2005. Ha Noi: Labor and Social Publishing House.

Muller, C. 1995. "Maternal Employment, Parent Involvement, and Mathematics Achievement among Adolescents." Journal of Marriage and the Family 57(1): 85-100.

Nguyen, B., V. Vu, T. Phan, and T. Nguyen. 2001. Educational Financing and Budgeting in Vietnam. Paris: International Institute for Educational Planning, UNESCO.

38

Page 41: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Nguyen, N. 2004. "Trends in the Education Sector." In P. Glewwe, A. Nisha and D. David (eds.), Economic Growth, Poverty, and Household Welfare in Vietnam. Washington, DC: World Bank, pp. 425-466.

Nguyen, P. 2008. "What is in it for the Poor? Evidence from Fiscal Decentralization in Vietnam." Journal of Public and International Affairs, 69-90.

Nguyen, P., and J. Yinger. 2008. “Education Finance Reform, Local Behavior, and Student Performance in Massachusetts.” Working Paper.

(NYSDB) New York State Division of the Budget. 2007. Description of 2007-08 New York State Executive Budget Recommendations for Elementary and Secondary Education. New York: New York State Division of the Budget.

Reschovsky, A., and J. Imazeki. 1998. "The Development of School Finance Formulas to Guarantee the provision of Adequate Education to Low-Income Students." In W. Fowler (ed.) Developments in School Finance, 1997: Does Money Matter?. Washington. DC: National Center for Education Statistics, pp. 121-148.

Reschosky, A., and J. Imazeki. 2001. "Achieving Educational Adequacy through School Finance Reform." Journal of Education Finance 26(4): 373-396.

Reschosky, A., and J. Imazeki. 2003. "Let No Child be Left Behind: Determining the Cost of Improving Student Performance." Public Finance Review 31(3): 263-290.

Rice, R. 1997. "Cost Analysis in Education: Paradox and possibility." Educational Evaluation and Policy Analysis 19(4): 309-317.

Rosen, H. 2002. Public Finance. New York: McGraw-Hill.

Ross, S., and J. Yinger. 1999. "Sorting and Voting: A Review of Literature on Urban Public Finance." In E. Mills, and P. Cheshire, Handbook of Regional and Urban Economics 3, Amsterdam: North-Holland, pp. 2003-2060.

Rubenstein, R., and L. Picus. 2003. "Politics, the Courts and the Economy: Implications for The Future of School Financing." In D. Sjoquist (ed.), State and Local Finances under Pressure. Northampton, MA: Edward Elgar Publishing, pp. 60-94.

Staiger, D, and J. Stock. 1997. "Instrumental Variables Regression with Weak Instruments." Econometrica 65(3): 557-586.

Strom, R., and S. Strom. 1995. "Intergenerational Learning: Grandparents in the Schools." Educational Gerontology 21(4): 321-335.

Taylor, L., C. Alexander, T. Gronberg, D. Jansen, and H. Keller. 2002. "Updating the Texas Cost of Education Index." Journal of Education Finance 28(2): 261-284.

39

Page 42: To Graduate or Not:coppfs1.asu.edu/spa/abfm2008/new/Nguyen.doc · Web viewWhile Muller (1995) find that unemployed mothers are the most involved in their children’s education relative

Tsang, M. 1988. "Cost Analysis for Educational Policymaking: A Review of Cost Studies in Education in Developing Countries." Review of Education Research 58(2): 181-230.

Tsang, M. 1997. "Cost Analysis for Improved Educational Policymaking and Evaluation." Educational Evaluation and Policy Analysis 19(4): 318-324.

(VNCI) Vietnam Competitiveness Initiative. 2005. “Vietnam Provincial Competitiveness Index 2005: Measuring Economic Governance for Private Sector Development.” Ha Noi: VNCI.

(VNCI) Vietnam Competitiveness Initiative. 2006. “Vietnam Provincial Competitiveness Index 2006: Measuring Economic Governance for Private Sector Development.” Ha Noi: VNCI.

Woolbridge, J. 2003. Introductory Econometrics: A Modern Approach. Cincinnati, OH: South-Western College Publishing.

World Bank. 1996. “Vietnam: Education Financing Sector Study.” World Bank Report No. 15925-VN. Washington, DC: World Bank.

World Bank. 2008. “World Development Indicators Online.” Accessed via Syracuse University Library.

Yinger, J. 2004. “State Aid and the Pursuit of Educational Equity: An Overview.” In J. Yinger (ed.), Helping Children Left Behind: State Aid and the Pursuit of Educational Equity. Cambridge, MA: The MIT Press, pp. 3-57.

Yinger, J. , H. Bloom, A. Borsch-Supan, and H. Ladd. 1998. Property Taxes and House Values. New York: Academic Press, Inc.

Zahirovic-Herbert, V., and G. Turnbull. 2008. School Quality, House Prices and Liquidity. Journal of Real Estate Finance and Economics 37(2): 113-130.

Zick, C., W. Bryant, and E. Osterbacka. 2001. "Mothers’ Employment, Parental Involvement, and the Implications for Intermediate Child Outcome." Social Science Research 30: 25-49.

40