Influence of personal characteristics on reelections in
school committees: evidence from a field experiment in
Indonesia
E.B.M. Haccou
University of Amsterdam, Amsterdam, Netherlands
December 9, 2013
Abstract: This paper describes the possible relationship between personal characteristics and the chance of a reelection in a school committee. It is found that age and working in the public sector both have a negative effect on the chance of reelection. A higher level of education has a positive relationship with reelection. Gender and position within the committee are found insignificant in relationship to a possible reelection.
Keywords: reelection, field experiment, Indonesia, school committees
Index
1 Introduction ... 3
2 Factors that influence reelections ... 6
2.1 Influence of characteristics on reelections ... 6
3 Research hypotheses ... 8
4 Data and methodology ... 9
4.1 Data ... 9 4.2 Methodology ... 10 5 Results ... 13 5.1 Estimates ... 13 5.2 Goodness of fit ... 15 5.3 Predictive quality ... 15 6 Conclusion ... 16 Appendix ... 18 Bibliography ... 19
1 Introduction
At this moment, more than 57 million children in the world do not attend primary school. Next to this, the quality of education causes that there are at least 250 million children who cannot read nor count, after attending school for four years (UK Government, 2011). Developed and developing countries work together trying to improve the quality of education in developing countries. Policy makers, among others, try to close the gap between developed and developing countries and firmly believe that improving education will solve the historical problem of inequality between developed and developing countries. An influential factor on improving the quality of education is found in participation of the community (Stiglitz, 2002). Mansuri and Rao (2012) confirmed the influence of this factor, where improvement of education can be achieved through the establishment of school committees and parent-teacher meetings. Although theoretically of great influence, in practice Banerjee and Duflo (2008) found that these institutions often do not work properly. This conclusion is confirmed by Bruns, Filmer and Patrinos (2011), who state that participation of the community can attribute to the quality of education, with the footnote that the way this should be done is explored insufficiently.
In order to investigate the influence of school committees on the quality of education Pradhan et al. (2011) conducted a research in Indonesia. In this research they examined 520 schools and tried to find which factors contribute to the improvement of quality through school committees. In 190 of the investigated schools, an election intervention took place. This intervention consisted of an election, in which the members of the school committee were democratically chosen. The reason for this intervention was that the authors suspected that school committees would contribute more to
improvement of education if the members were chosen democratically. In the school committees before intervention, the members were either self-installed or appointed, but not democratically chosen. The reason for this intervention and the suspicion of a greater contribution was found in earlier research of Ostrom and Ahn (2009). In their research, the authors find that organizing democratic elections enhances the social capital. The committee has a greater authority within the community and the communication is easier between the committee and other stakeholders like parents and teachers. Using this theory of Ostrom and Ahn (2009), Pradhan et al. (2011) found that this theory generates a hypothesis that there is a relationship between school committees and the quality of education. The quality of education did improve after the members of the school committee were chosen democratically, however elections alone did not raise learning outcomes. It was the combination of democratic elections and meetings between the school committee and the village council that raised learning outcomes. The aim of the meetings between the school committee and the village council is to improve the level of communication and collaboration between the community, parents, teachers and the school committee.
In order to make these meetings more efficient, the composition of the school committee might be of interest. Besley, Pande and Rao (2005, p. 6) found that people who work in the private sector, for example businessmen, are more likely to abuse their political position than people working in the public sector. Next to his, businessmen who want to be elected do this primarily to gain greater influence. This can decrease the quality of linkage between the school committee and the community. Transposing this research done by Besley et al. (2005) to the composition of school committees can lead to interesting outcomes of whether a member adds to the efficiency of linkage. The hypothesis is that people who work
in the private sector are less often reelected when new elections take place because voters will notice the absence of their contribution for the community.
However, the sector in which a person works might not be the only factor influencing a reelection. In their research, Pradhan et al. (2011), composed a set of variables of the members of a school committee before and after an election. This research is an extension of the research of Pradhan et al. (2011) and tries to examine which of these factors are of influence on a possible reelection. Factors used by Pradhan et al. (2011) include age, position in the committee before election, gender, level of education and working sector. The hypothesis is that these factors contribute either in a negative or in a positive way to a reelection. An hypothesis on each variable will be investigated in this paper.
This research aims at giving an elucidation on the theme of reelection and contributing factors to this reelection. Previous research provides information on several coherent topics, but research on the factors contributing to a reelection in a school committee is absent. The main question that has to be answered is whether there are factors that contribute in a positive or negative way to a reelection in a school committee.
Several factors are found to have a significant relationship with reelection in a school committee. Both age and working in the public sector have a negative effect on the chance of reelection. Having a higher education level significantly increases the chance of a reelection within the school committee. Evidence of the influence of gender and position within in the committee on a possible reelection is not found.
To investigate the influence of aforementioned factors on a reelection and whether they contribute in a positive of negative way, a literature research is of utmost importance.
In paragraph 2 related research is investigated. On the basis of this research, hypotheses are formed on each of the aforementioned factors in the third paragraph. In the fourth section the research methodology and the data used of this research will be described. The methodology is partly based on the research presented in section 2. After defining the methodology and dataset, the fifth paragraph will consist of the quantitative research that is conducted. This research is conducted on data that is also used by Pradhan et al. (2011). It will also take up an analysis about the goodness of fit and the predictive quality of the model. The last paragraph, 6, presents the answer to the question whether the aforementioned factors are of influence on the reelection of school committee members.
2 Factors that influence reelections
Previous research performed by Pradhan et al. (2011) illustrates factors that may be relevant in a reelection. Their survey uses six variables. These variables are age of the committee member, position in the committee before the election, gender of the committee member, level of education and the working sector of the committee member. The author does not give an insight into the importance of each variable. However, earlier research on every variable is available.
2.1 Influence of characteristics on reelections
The underrepresentation of women in political functions is an area that is regularly examined. Halfway 2000, women held 13.8% of the political positions globally. This percentage rose with almost 5% compared with 13 years earlier (Chattopadhyay and Duflo, 2004, p. 1409). When the representation of women is compared with other areas like
education and legal rights, it is seen that the difference between men and women is the least declined at the area of political representation (Inglehart and Norris, 2000). This has led to quotas for the number of women in political and decision-making functions (World Bank, 2001). These quotas suggest that the representation of women in political positions increases, although these quotas give no insight in the probability that a woman is reelected. The study by Chattopadhyay and Duflo (2004, p. 1427) has shown that the personality of women differs from men and that this might have influence on the reelection. The authors find that women are unlikely to be reelected at the next election.
Next to the importance of gender on the possibility of a reelection, Besley et al. (2005, p. 28) find that the level of education of politicians is also of influence on reelections. Household data of Indian villages is used to study the way political selection is established. The authors conclude that education level has a systematically positive effect on selection. These findings indicate that a higher level of politicians’ education makes them more skilled and voters will more often recognize them as such. Besley et al. (2005, p. 29) eventually emphasize the importance of focusing on characteristics of politicians as a step towards a better quality of government.
Furthermore, Avolio, Waldman and McDaniel (1990) explored the influence of age and years of experience on work performance. Although experience appears to be a better predictor of work performance, the authors found that age and work performance also have a positive relation. In addition to this research, Ferejohn (1986, p. 7) concludes in his paper that voters take the performance of incumbent candidates into account during elections. Combining these findings, it is assumed that age has a positive relation with work performance
and that better work performance, of an incumbent candidate, is of importance during elections.
Next to this and also stated in the introduction, Besley et al. (2006, p.6) find that people who work in the private sector, for example businessmen, more often abuse their political position than people working in the public sector. Gehlbach, Sonin and Zhuravskaya (2007, p. 32) have confirmed these findings. They hypothesize that the opportunity costs for a businessman of nominating himself for public office are higher than those for a professional politician. The authors conclude that businessmen may use their political position for their own interest even more when political institutions are weak. This is often the case when the democracy is immature (Gehlbach, Sonin and Zhuravskaya, 2007, p. 3).
Moreover, the position of a member may be of influence on a reelection. Farnsworth and Owen (2004) found that voters who gained information about candidates and their past used this to decide to vote for them. Supposing a chairman has more influence on what is published about him and his board than other members of a board, the assumption that influence on a reelection differs between various positions within a board is not negligible.
3 Research hypotheses
Earlier research indicates that personal characteristics as gender, level of education, age, working sector and position within the committee has a relationship with the chances of reelection in the school committee. However, no previous research is done on the degree of influence of the aforementioned characteristics. With the information available from earlier research, the following hypotheses will be examined.
Hypothesis 1: Age of the school committee member will be of significant influence on chances of reelection in the school committee.
Hypothesis 2: Members of the school committee who are
male have a higher probability of reelection.
Hypothesis 3: Members of the school committee who work in
the public sector have a higher probability of reelection than members of the school committee not working in the public sector.
Hypothesis 4: Members of the school committee with a
higher level of education have a higher probability of reelection.
Hypothesis 5: Position within the committee of the school
committee member will be of significant influence on chances of reelection in the school committee.
4 Data and methodology 4.1 Data
Data used for this research is collected for earlier research by Pradhan et al. (2011). The authors conducted surveys on members of school committees before and after elections were implemented. These elections took place in 190 school committees in Indonesia. These were chosen randomly from the sample of 520 school committees used in the research of
Pradhan et al. (2011). This sample was selected with a few restrictions1.
The estimates in this research are composed using 1795 observations of the surveys completed by members of a school committee belonging to one of the 190 schools. This number of observations is sufficient to ensure the final conclusion will be reliable.
The survey in the baseline, before elections were implemented, contains questions about personal characteristics of the members of the school committees and was conducted in 2006. Age, position within the committee, gender, education level and working sector have been indicated by each member. The survey in the endline, after elections, states whether a member is reelected or not and was conducted in 2008.
4.2 Methodology
The methodology this research is built upon comes from the binary response theory (De Boer, van Dijk, Franses, Heij and Kloek, 2004, p. 438). Binary dependent variables have only two possible outcomes. In this study the two outcomes are whether a member is reelected or not. The probability of reelection can be confined to values between zero and one by using a non-linear model:
F is a function with values ranging between zero and one and is corresponding to a logistic density function, so that this research uses the logit model.
The index-function belonging to this model gives the systematic preference and an individual-specific effect :
1 For these restrictions see “Improving Educational Quality through
Enhancing Community Participation: Results from a Randomized Field Experiment in Indonesia” by Pradhan et al. (2011).
The observed value of y, where y equals one if the member is reelected and zero if not, is related to the index by means of the equation:
For this research it is assumed that the individual effects are independent and identically distributed with symmetric density belonging to a logit model. The logistic density function of the individual effects can be stated as subjoined:
The parameters of F used for the regression which will give the maximum likelihood estimates via the binary logit model are stated in table 1. Level of education of the school committee member is divided into five levels2, with level 1 the lowest level of education and level 5 the highest. Position within the committee can be chairman, vice-chairman, secretary, treasurer or ordinary member of the school committee. To avoid perfect multicollinearity the first dummy variables of education and position are not included in the regression.
Testing on the significance of all the variables will provide answers to the hypotheses. Theoretical significance is a first criterion. However, probably of more importance is statistical significance. To investigate the statistical significance of the included coefficients, the linguistic hypotheses can be converted to graphical ones:
The test is conducted on . When this coefficient is significant within a 5% significance level, the null hypothesis will be rejected and significant influence of the concerning variable can be proved.
Symbol Description of parameter Coefficient
C Constant β0
Age Age of the member β1
GenderD Dummy variable: β2
equals 1 if member is male equals 0 if member is female
WorkD Dummy variable: β3
equals 1 if school committee member works in the public sector
equals 0 if school committee member does not work in the public sector
EducationD2 Dummy variable: β4
equals 1 if the hightest level of education of the member is level 2 equals 0 if the hightest level of education of the member is not level 2
EducationD3 Dummy variable: β5
equals 1 if the hightest level of education of the member is level 3 equals 0 if the hightest level of education of the member is not level 3
EducationD4 Dummy variable: β6
equals 1 if the hightest level of education of the member is level 4 equals 0 if the hightest level of education of the member is not level 4
EducationD5 Dummy variable: β7
equals 1 if the hightest level of education of the member is level 5 equals 0 if the hightest level of education of the member is not level 5
PositionD2 Dummy variable: β8
equals 1 if the position of the member is vice-chairman equals o if the position of the member is not vice-chairman
PositionD3 Dummy variable: β9
equals 1 if the position of the member is secretary equals o if the position of the member is not secretary
PositionD4 Dummy variable: β10
equals 1 if the position of the member is treasurer equals o if the position of the member is not treasurer
PositionD5 Dummy variable: β11
equals 1 if the position of the member is ordinary member equals o if the position of the member is not ordinary member Note: member refers to school committee member
The overall significance of the model will be tested by the Likelihood Ratio test on the null hypothesis that all coefficients, except the constant term, are zero.
Furthermore, the predictive quality of the model will be tested through a classification table of the predicted responses against the actually observed responses. The hit rate is defined as the fraction of correct predictions in the sample. The predictive quality of the model will be evaluated by comparing the hit rate with the random hit rate under the null hypothesis that the predictions are no better than random predictions.
5 Results 5.1 Estimates
Table 2 shows the maximum likelihood estimates for the binary logit regression as provided in last section. As stated in hypothesis 1, this research expects to find evidence of a significance influence of age on the possibility of reelection in the school committee. Table 2 shows that the value of the coefficient β1, corresponding with age, is negative and the p-value is 0.0028. This means that the null hypothesis will be rejected and that a higher age of a school committee member has a negative effect on the possibility of reelection.
The coefficient of β2, a dummy variable of gender, is positive but has a p-value outside the 5% critical region. Therefore the null hypothesis will not be rejected and significantly influence of gender of a school committee member on a reelection cannot be proved.
The coefficient of the dummy variable for working sector, β3, is negative and lies within the 5% critical region. The null hypotheses will therefore be rejected. Hypothesis 3 states that members of a school committee who work in the
public sector have a higher probability of reelection than members of the school committee not working in the public sector. This research finds evidence for the opposite. Working in the public sector has a negative effect on the chance of being reelected in the school committee.
Table 2. Regression output
Variable Coefficient ML estimates P-value
Constant β0 1,1637 0,0028
(0,3887)
Age β1 -0,0261 0,0000
(0,0063)
Dummy variable for gender β2 0,2149 0,1919
(0,1646)
Dummy variable for working sector β3 -0,7429 0,0000
(0,1421)
Dummy variable for education level 2 β4 -0,2483 0,1054
(0,1534)
Dummy variable for education level 3 β5 -0,0293 0,8411
(0,1459)
Dummy variable for education level 4 β6 0,6462 0,0003
(0,1770)
Dummy variable for education level 5 β7 2,0156 0,0026
(0,6693)
Dummy variable for vice-chairman β8 -0,2146 0,2955
(0,2051)
Dummy variable for secretary β9 -0,2383 0,2138
(0,1917)
Dummy variable for treasurer β10 -0,1401 0,4368
(0,1802)
Dummy variable for ordinary member β11 -0,1027 0,4634
(0,1401)
Number of observations: 1795 LR-statistic: 57,5141
McFadden R2: 0,0231 P-value LR Statistic: 0,0000
Note: Standard errors in parentheses. Hypothesis 4 argues that a higher level of education of the school committee member has a positive influence on the chance of reelection. Regarding the coefficients of the variables β4, β5, β6 and β7 this hypothesis is proved true. The
first two coefficients are not significant within the 5% significance level so those equal zero. The latter two are significant and both coefficients are positive. This means that having education level 4 or 5 significantly increases the chance of a reelection within the school committee with respect to education level 1.
The coefficients of β8, β9, β10 and β11 have a p-value outside the 5% critical region. Therefore the null hypotheses will not be rejected and significant influence of the position of a school committee member within the school committee on a reelection cannot be proved.
5.2 Goodness of fit
The overall goodness of fit of the model is tested by the Likelihood Ratio test. Although some of the coefficients are not significant, the variables are jointly significant. As stated in table 2 the LR-statistic is 57.5 with probability value 0.000. This means that the null hypothesis, that all coefficients except the constant term are zero, can be rejected. This shows that the model has explanatory power as a consequence of the jointly significance of the variables adopted in this research.
5.3 Predictive quality
Table 3 contains results on the predictive performance of the used logit model. The hit rate in this model as stated in table 3 is 0.5677. The expected hit rate can be calculated by:
This gives an expected hit rate of 0.5001. The hit rate for this model is above the expected hit rate so this implies a good predictive quality of the model.
To test whether the predictions are better than random predictions the z-value is computed:
This gives a z-value of 5.728 and the null hypothesis of no better predictions than random predictions can be rejected by the 5% significance level with critical value 1.645.
Table 3. Prediction evaluation
Estimation Equation
Dep = 0 Dep = 1 Total
P[Dep = 1] ≤ C 505 374 879
P[Dep = 1] > C 402 514 916
Total 907 888 1795
Correct 505 514 1019
% Correct 55.68 57.88 56.77
Note: C is success cutoff = 0.5
6 Conclusion
This paper examines the influence of personal characteristics on reelections in school committees after a field experiment in Indonesia. Previous research already shows the existence of significant influence of different personal characteristics on elections and reelections. This research enucleates evidence of the influence of certain personal characteristics on the chance of reelection within school committees.
Several factors are found to have a significant relationship with reelection in a school committee. Both age and working in the public sector have a negative effect on the possibility of reelection. Having education level 4 or 5 significantly increases the chance of a reelection within the school committee. Evidence of the influence of gender and
position within in the committee on a reelection within the school committee is not found.
The goodness of fit of the model is tested through a likelihood ratio test and has indicated that the variables are jointly significant. This shows that the model has explanatory power. Furthermore, the predictive quality of the model is
tested and predictions resulting from the model are found better than random predictions.
Appendix
Table A. Education levels
Indonesian level Level as in research
SD/MI 1 SMP/MTs 2 SMA/SMK/MA 3 Diploma I/II 4 Diploma III/S 4 Diploma IV/S1 4 Pascasarjana 5
Bibliography
Ahn, T.K. and E. Ostrom (2009). The Meaning of Social Capital and Its Link to Collective Action. In D. Castiglione, J. W. van Deth, & G. Wolleb (Eds.). Handbook on Social Capital, pp. 17-35. Northampton, MA: Edward Elgar.
Alisjahbana, A., R.P. Artha, A. Beatty, A. Gaduh, M. Pradhan, D. Suryadarma and M. Wong. Improving Educational Quality
through Enhancing Community Participation: Results from a Randomized Field Experiment in Indonesia. Policy Research Working Paper 5795.
Avolio, B.J., M.A. McDaniel and D.A. Waldman (1990). Age and work performance in nonmanagerial jobs: The effects of experience and occupational type. Academy of Management
Journal, 33, pp. 407–422.
Banerjee, A.V. and E. Duflo (2008). Mandated Empowerment - Handing Antipoverty Policy Back to the Poor? Annals of the New
York Academy of Science, 1136, pp. 333-341.
Besley, T., R. Pande and V. Rao (2005). Economic Growth Center Working Paper 921.
Boer, P., H.K. van Dijk, P.H. Franses, C. Heij and T. Kloek (2004). Econometric Methods with Applications in Business and
Economics. Oxford: Oxford University Press.
Bruns, B., D. Filmer and H.A. Patrinos (2011). Making Schools
Work: New Evidence on Accountability Reforms. Washington,
D.C.: The World Bank.
Chattopadhyay, R. and E. Duflo (2004). Women as policymakers: Evidence form a randomized policy experiment in India.
Econometrica, 72, (5), pp. 1409-1443.
Farnsworth, S.J. and D. Owen (2004). Internet Use and the 2000 Presidential Election. Electoral Studies, 23, (3), pp. 415-429.
Ferejohn, J. (1986).Incumbent performance and electoral control. Public Choice, 50, (1), pp. 5-25.
Gehlbach, S., K. Sonin and E. Zhuravskaya (2007). Businessman Candidates. American Journal of Political Science, 54, (3), pp. 718–736.
Inglehart, R. and P. Norris (2000). The Developmental Theory of the Gender Gap: Women’s and Men’s Voting Behavior in Global Perspective. International Political Science Review, 21, (4), pp. 441-463.
Mansuri, G. and V. Rao (2012). Localizing Development:
Does Participation Work? Washington, D.C.: The World Bank.
Stiglitz, J.E. (2002). Participation and Development: Perspectives from the Comprehensive Development Paradigm.
Review of Development Economics, 6, (2), pp. 163-182.
UK Government (2013). Making sure children in developing
countries get a good education. (https://www.gov.uk/government /policies/making-sure-children-in-developing-countries-get-a-good-education), October 11.
