Oxford Development Studies, Vol. 37, No. 1, March 2009 Teacher Incentives and Performance: An Application of Principal – Agent Theory ROSALIND LEVAČIĆ ABSTRACT The paper summarizes principal– agent (P-A) theory and applies it to the teaching profession, arguing that it provides a strong framework for analysing institutional arrangements governing the work of teachers. P-A theory proposes factors that determine whether or not paying teachers in relation to measures of performance improves teacher productivity. Teachers’ work is characterized by moral hazard, risk aversion, multiple principals and multiple objectives, which make the design of an optimal performance pay system complex, especially as it needs to be context specific. A crucial factor is the extent to which teacher motivation is altruistic or opportunistic. International evidence on teacher rewards systems and their relation to teacher performance is summarized. In many developing countries, such as India, teacher contracts fail to provide sanctions for poor performance or rewards for effective teaching. In such contexts, improved incentives for teacher performance are an essential component of reforms to raise the quality of education. 1. Introduction Teachers account for the majority of spending on education provided by schools: on average 64% in developed countries (OECD, 2007a) and often running to over 90% in developing and transition countries. Research evidence indicates that, apart from the student’s family background, the single most important factor in determining variability in student attainment within any single country is teacher quality (Hanushek, 2005). Therefore, the institutional arrangements that govern the incentives influencing teachers’ work are of prime importance for the overall quality of a nation’s education system. Contractual arrangements for teachers differ vastly both between and within countries. At one extreme, teachers are civil servants, paid a fixed salary according to criteria such as formal qualifications and experience, enjoying complete security of tenure; at the other, teachers are employed by private sector schools, subject to relatively easy dismissal and paid according to judgements about their performance. Thus, a crucial issue for education policy is what kinds of contractual arrangement will ensure the highest quality teaching force for a given expenditure. Principal – agent (P-A) theory provides great analytical clarity in addressing this key question. In this paper, I set out first to outline P-A theory, then apply it to the analysis of teacher contracts and finally consider the empirical evidence Rosalind Levačić, Department of Quantitative Social Science, Institute of Education, University of London, 20 Bedford Way, London WC1 0AL, UK. Email: [email protected] The author would like to thank Geeta Kingdon and an anonymous reviewer for helpful comments. Any errors or misinterpretations are her responsibility. ISSN 1360-0818 print/ISSN 1469-9966 online/09/010033-14 q 2009 International Development Centre, Oxford DOI: 10.1080/13600810802660844 34 R. Levačić on the relationship between teacher contracts involving performance-related pay and student attainment. 2. Principal – Agent Theory A P-A relationship arises whenever a person or organization contracts, either explicitly or implicitly, another—the agent—to perform services or supply goods. The agent is required to expend some effort in order to produce the output required by the principal, for which the principal pays the agent rewards contingent on particular circumstances occurring, as set out in the contract between them. The simplest P-A contract is between an individual principal and a single agent, as when a person engages a solicitor or—a frequently used example—when a landlord rents land to a farmer. Normative P-A theory is concerned with establishing the type of contract that optimizes returns for the principal (Douma & Schreuder, 2002)—the analysis always assumes that the objective is to maximize the principal’s welfare. Some contracts are between principals and agents who are in the same organization, for example, between management and workforce or management and shareholders, and others occur between principals and agents who are in different organizations. P-A theory was initially developed for analysing differences in behaviour between owner-managed firms and public companies, as in the latter ownership by shareholders is separated from control of the company by managers (Fama, 1980; Fama & Jensen, 1983). The central problem is how shareholders can ensure that managers run the company so as to maximize shareholder wealth. The solution is paying managers some combination of fixed salary, bonuses dependent on company performance, and share options. P-A theory is also much developed with respect to insurance and the regulation of commercial businesses that have a monopoly control of specific assets (e.g. utility companies). P-A theory relates to non-marketed services just as much as marketed ones and so has been applied to the public sector as well as the private. In public services, the over-arching P-A relationship is between the citizens-cum-taxpayers, who collectively agree to organize public services from state organizations, and the workers in state bureaucracies who are the agents. In this situation there are multiple principals and multiple agents. In the public sector, the main issue for P-A theory is what structure of rewards and sanctions will best motivate state bureau workers to produce the kinds of output desired by the principals and at least cost to the principals. P-A theory embraces issues of accountability as it examines the ways in which the principal may hold agents accountable for their performance. P-A theory demonstrates that the nature of the optimal contract between principal and agent depends on several crucial conditions. These are: . the motivation of the agents; . moral hazard, which concerns the nature of the relationship between the effort expended by the agent and their output; . the amount and distribution of information between principal and agent about the relationship between the agent’s effort and output and the costs to the principal of obtaining such information; . whether the agent performs multiple tasks or has several principals—conditions that are more likely to occur in the public sector than in the private sector. Teacher Incentives and Performance 35 2.1 Agent Motivation P-A theory assumes that agents act to maximize their utility, so what gives the agent utility from their work for the principal is crucial. Financial rewards are usually a major motivation for working. In addition, employees may gain satisfaction from the status or the perquisites of the job (e.g. nice office, expensive car). Less self-serving sources of job satisfaction, such as affiliation with colleagues, pride in quality of work and benefiting clients, can also motivate employees. There are also factors that give employees (agents) disutility, in particular the effort expended on the job, which includes effort searching for more efficient methods of production and new products, to enhance consumers’ satisfaction. In his book Motivation, Agency and Public Policy, Le Grand (2003) distinguishes two “ideal” and contrasting types of agent—knaves and knights. Knaves are solely self-interested and opportunistic, whereas knights are altruistic, motivated by intrinsic job satisfaction and the desire to promote the welfare of their clients—the ultimate principals. Each individual agent has some combination of knightish and knavish motivations, and among a group of agents providing similar services there is likely to be a distribution of different combinations of knave/knight. Two theories have been proposed to explain why agents working in non-profit organizations providing “caring services” (e.g. health, education, public administration) are more intrinsically motivated than agents in private sector organizations or non-caring non-profit organizations (Gregg et al., 2008). The first is “organizational-form” theory, according to which agents are influenced towards either knightish or knavish behaviour by the professional values and organizational cultures of their workplace settings. Hence, non-profit organizations providing caring services influence their employees to behave as knights. The second theory (the two are not mutually exclusive) is that of “mission selection”. Not-forprofit caring organizations attract agents who are naturally disposed to pro-social behaviour, which means that agents self-select into different types of organization according to the agents’ knightish or knavish predispositions. Gregg et al. tested these two theories using British Household Panel Data, and concluded that mission selection theory is strongly supported by the evidence, although organizational-form theory is not “disconfirmed”. 2.2 Moral Hazard As well as agent motivation, another crucial factor is the nature of the agent’s work and the extent to which the principal can infer agent effort from observing the agent’s output. Moral hazard exists when “the agent’s action affects the principal’s payoff, although the action is not directly observable by the principal” (Dixit, 2002). Moral hazard arises when there is some uncertainty regarding the relationship between the agent’s effort and his/her output. For example, in the case of a farmer, the quantity of harvest depends on the weather as well as the farmer’s efforts, but unless the landlord can watch the farmer continuously, the amount of output due to the farmer’s efforts is uncertain. Moral hazard poses a problem for the principal because he/she does not know to what extent the agent’s output is due to his/her effort, and so should be rewarded, or is due to forces outside the agent’s control and so should not be reflected either positively or negatively in the agent’s reward. 2.3 Information Asymmetry Moral hazard poses particular problems for the principal in devising an optimal contract when there is information asymmetry. This occurs when the principal has less knowledge 36 R. Levačić than the agent of how the agent’s effort relates to his/her output. If the principal pays the agent a fixed wage and receives the residual output, then the agent has no incentive to maximize output for the benefit of the principal because the principal does not know the extent to which low output is due to random factors outside the agent’s control or to shirking by the agent. The information asymmetry problem is compounded further for the principal if the agent’s work involves multiple kinds of output—as is the case with teaching. The only way the principal can improve his/her information about the agent’s work effort and its relationship to output is to put in place monitoring and verification procedures, and these involve varying amounts of cost depending on the nature of the production process. One of the reasons for information asymmetry is lack of competition between agents. If agents work in a highly competitive market then principals can compare the performance of agents and only contract those who work efficiently. Consequently, non-efficient or shirking agents will have to exit the market. Hence, P-A theory about the nature of optimal contracts under differing circumstances applies when competition is limited due to insufficient alternative agents or there is a lack of good, low-cost information. 2.4 Types of Optimal Contract P-A theory deduces that there are three types of optimal contract depending upon the attitude of the agent to risk.1 If the agent is risk-neutral (and so does not need any financial compensation for taking on the risk of an uncertain reward) and there is moral hazard with asymmetric information, the optimal contract for the principal is a fixed payment from the agent, which gives the agent the claim to the residual, uncertain output. Farmers’ payment of a fixed rent to landlords is an example of such a contract. This fixed fee contract gives the agent the incentive to maximize output. To maximize his/her utility the principal needs to set the fixed fee from the agent at the agent’s reservation payment—the maximum fee payment that induces the agent to enter the contract for the risky residual rewards. If agents are risk averse, then in order to compensate for risk they will insist on a lower fixed fee payment contract (other things being equal) than if they were risk neutral. In the case of risk-averse agents, one solution is for the principal to receive the residual output and pay the agent a fixed wage. This maximizes the principal’s pay-off if there is no moral hazard or information asymmetry—the principal has perfect knowledge of how much effort the agent applies and the consequent output and so can set the wage contract for a specific amount of effort. If there is moral hazard and asymmetric information and, in addition, agents are knaves, then under a fixed-wage contract they will commit minimum effort consistent with maintaining the contract. Only to the extent that the agents are not opportunistic and behave as knights will they expend effort to serve the principals’ interests. In situations of moral hazard, information asymmetry and risk-averse agents, the optimal contract is one that shares the risk between the principal and the agent. The agent receives some fixed-wage payment plus additional pay dependent on performance as specified in the contract. It is often too complex to include in a contract all possible contingencies. An explicit contract specifies exactly what the principal and agent will pay each other and receive as residual output under a range of contingent circumstances, and these agreements can be legally enforced. By contrast, an implicit contract between two or more parties in an Teacher Incentives and Performance 37 exchange relationship depends on a mutual understanding of what is expected from each party under different circumstances. Implicit contracting is dependent on shared cultural values and expectations, which contribute to and rely on trust between the parties. Implicit contracts are not usually legally enforceable and so depend on mutual trust for their fulfilment. The amount of costly monitoring and verification a principal needs to incur therefore depends on the motivations of the agent and the nature of the exchange relationship. If the agent is motivated only by financial and material rewards and the avoidance of effort and is “opportunistic” as well, then the agent will seek to obtain the maximum personal gain from the contract. This kind of behaviour is referred to as “gaming”. Any ambiguities in the contract will be exploited for the agent’s benefit and the principal’s loss. An opportunistic agent cannot be trusted by the principal, who consequently has to counter the opportunism with complex, explicit, contingent contracts and costly monitoring and verification of the agent’s work. By contrast, when the principal can trust the agent he/she can rely far more on implicit contracts and engage in little or no monitoring. 3. Principal –Agent Theory Applied to Teachers The first thing to note when applying P-A theory to the work of teachers is that a complex hierarchy of P-A relationships is involved. An example is shown in Figure 1 for a decentralized education system, such as that in England, where most state schools are maintained by elected local authorities, and each school is managed by a governing body consisting of elected parents and staff as well as local community nominees. The voters are the ultimate principals and their agents are the elected Members of Parliament and, in particular for education, the Minister of Education2 and, at local level, local councillors. The Minister of Education is the principal in relation to local authorities, as the Ministry sets national education policy and provides a proportion of school funding. Local authorities are agents with respect to both the Minster of Education and local voters. Schools are agents of the local government and held accountable by them. Parents are principals of the school governing body, as they elect representatives to them, but the governing body is also an agent of the local authority and, via it, the Minister of Education as well. The governing body is principal with respect to the head teacher, particularly in Figure 1. Multiple principal – agent relationships in education. 38 R. Levačić systems where the governing body, not the local authority or the Minister, appoints the head teacher. Finally, the head teacher, as the most senior manager of the school, acts as principal towards teachers, who are agents. In high-stakes/high-accountability education systems, such as those in Britain and many US and Australian states, for example, this hierarchy of P-A relationships is defined in terms of clear accountabilities for individual performance for teachers, school-level performance for head teachers and responsibility for school system performance by local authorities and the Minister of Education. A high-cost monitoring and verification system for assessing performance is maintained and funded by the state. Consequently, teachers are agents to several principals who may not have exactly the same objectives or give different weights to a common set of objectives. While the head teacher is the teacher’s principal as his or her immediate line manager, the actual clients—parents and students— may have different preferences for education than those determined by the Minister of Education and executed via the managerial accountability of the head teacher. The ultimate principals of the education service are voters, but they form different stakeholder groups. So parents, for example, may have different requirements from the education system than employers. In examining P-A relationships as they apply to determining optimal contracts for teachers from the point of view of the ultimate principals, teacher motivation, moral hazard and information asymmetry in relation to the work of teaching need to be considered. As noted already, the terms of an optimal contract depend on the agent’s motivations. So an important consideration for the best form of teacher contract is the mix of extrinsic financial rewards and intrinsic professional and personal job satisfaction returns teachers require in order to attract people to the profession who will be effective teachers and remain effective during their working life. Although teachers are professionals, defined by Wilson (1989) as “someone who receives important occupational rewards from a reference group whose membership is limited to people who have undergone specialized formal education and have accepted a group-defined code of proper conduct”, this does not guarantee that they always act to maximize the welfare of their clients. As Matthews (1991) pointed out, a profession can define client interest in ways other than their clients do and be able to impose their own view of client interest when there is asymmetric information and lack of competition. Professions seek to enforce barriers to entry ostensibly to protect clients from unqualified practitioners, but also to restrict supply and thereby raise incomes of certified practitioners. It is to be expected that teachers, as with other occupations, are motivated to some extent by income, and this is borne out by studies of the responsiveness supply of teachers to relative salary levels (Zymelman & DeStefano, 1993; Eide et al., 2004; Chevalier & Dolton, 2005; Glewwe & Kremer, 2006). Teachers are also motivated by the satisfaction of enabling young people to learn and by an interest in child development and in their subject. To the extent that teachers are knights and therefore intrinsically motivated, it is not necessary to relate their pay to performance. Indeed, as Dixit (2002) and Le Grand (2003) noted, introducing financial incentives to motivate workers who have intrinsic motivations can actually reduce their performance. The nature of teachers’ work is complex. For a start there are multiple outputs, some produced jointly and some that are substitutes given constraints on teaching time. Outputs include not only the more easily measurable student cognitive attainment in various areas of the curriculum, but also less tangible outcomes, such as skills in adult life that enhance employability, as well as values, behaviours and attitudes that contribute to a harmonious Teacher Incentives and Performance 39 and just society. Teachers’ work is characterized by moral hazard. The relationship between a teacher’s effort and student learning is uncertain, as student characteristics, in particular those related to their family background, are the major determinant of student learning outcomes. Furthermore, information about teachers’ work effort is asymmetric: what goes on in classrooms is largely unobserved other than by students. To observe teachers at work requires costly observation by senior teachers or inspectors. Obtaining reliable measures of student learning over a period of time with a teacher requires a system of national tests that are externally set and reliably moderated so that teachers’ outputs, after controlling for student characteristics, can be compared. In order to take account of factors beyond the control of teachers that affect student learning outcomes, additional data on students’ prior attainment and/or family background factors must be collected and analysed using sophisticated statistical methods to obtain “value added” measures of student performance that estimate the contribution of school and teacher (Goldstein, 1997; Reezigt et al., 1999). Thus, verifying teachers’ contribution to student outcomes is costly and often beyond the capacity of middle- and low-income countries. 3.1 Principal– Agent Theory Applied to Teacher Contracts The most widely practised form of teacher payment is the fixed wage, which may be differentiated by qualifications, experience and responsibility. Under a fixed-wage contract, as teachers’ salaries are not dependent on effort or output, the risk of student learning being less than teachers could achieve if they expended maximum effort is borne by the principal, i.e. by students, parents and citizens. If teachers are knights, student outcomes will be as high as possible, given the knowledge and skills possessed by teachers; if they are knaves, then the quality of student learning is very poor. Two contrasting examples of teaching professions that are oriented either to knaves or to knights are India and Finland. Kingdon & Muzammil (2003) present a powerful case of knavish teacher behaviour in Uttar Pradesh. The Indian constitution enables one-twelfth of the members of State Legislative Councils to be elected by teachers in secondary schools and higher institutions. Teachers have used their political power to advance their selfinterest, in particular to get legislation passed in the 1970s that centralized the management of teachers. Teachers in private-aided schools (privately managed but state funded) can be elected to political office and are absent from school when attending to their political duties. In 2002, 29% of students were enrolled in private-aided schools (43% of secondary students) and 38% attended state schools. In Uttar Pradesh, teacher unions have used strikes and demonstrations to raise salaries, which are about two-and-a-half times the salaries of teachers in unaided private schools. School principals are unable to enforce attendance by teachers or dismiss teachers for poor performance. Thus, salaries are totally divorced from performance, even of the most basic kind—attendance—and yet a teacher’s salary is a higher ratio of GDP per capita than in other comparable countries (3.6 for India compared with 2.6 in Asia and 2.4 in Latin America). Kingdon and Muzammil also present evidence on the very poor quality of student attainment in Uttar Pradesh. In another paper, Kingdon & Teal (2007) report that teachers in private unaided schools in India, who are paid less than teachers in private-aided and government schools, are more productive. The Indian case illustrates a system where in the state and aided school sectors, principals have few means to control teacher-agents by using incentives and where teachers have a professional culture that is highly opportunistic and knavish. 40 R. Levačić The Indian case is but one example. Many developing countries have public sector teacher contracts that do not provide any incentives for performance, combined with knavish teacher behaviour, in particular high absenteeism and poor standards of teaching (Zymelman & DeStefano, 1993; Gottlemann-Duret, 2000). Fuller & Clarke’s (1994) review of education production function studies found that the marginal output of an increase in expenditure on non-staff resources exceeded that of an equivalent increase in spending on teachers. Pritchett & Filmer (1999) explained this finding in terms of P-A theory, arguing that inefficiently excessive spending on teachers is due to the capture of the state by teacher interests. A further example is Grindle’s (2004) review of education reform in Mexico and Latin American countries, which recognizes the powerful role of teacher unions in politics and in opposing decentralizing reforms that would strengthen local-level teacher accountability. At the other extreme is Finland, whose education system receives international accolades as consistently the highest performer in PISA (OECD, 2004, 2007b). In Finland teachers are paid only slightly more than average GDP per capita (OECD, 2007a). However, teachers are respected in society and teaching is one of the most sought-after fields of study by university applicants, only 15% of whom are accepted (Laukkanen, 2008). There is a high degree of decentralization of both management and the curriculum and no high-stakes accountability. The current situation, in which teachers can be trusted to perform well without centrally imposed accountability or extrinsic incentives, has been reached after 40 years of reforms aimed at creating a fully comprehensive basic education system. As these examples show, there is no single type of optimal teacher contract. The features of an effective teacher contract depend on the characteristics of a country’s education system, including—crucially—the motivation of teachers. The criteria for differentiating teacher pay are usefully divided into three categories—inputs, processes and outputs, a shown in Table 1. An additional distinction is whether the criteria are or can Table 1. Teacher payment systems: for inputs, processes or outputs Criteria for teacher pay differentiation Individual incentives Inputs: qualifications experience hours worked grade of student taught working in socially disadvantaged area Processes: for example: undertaking professional development engaging in extracurricular activities working with parents Outputs: for example: drop-out rates completion rates student attainment student value added cognitive attainment student and parent satisfaction reports professional judgements over a range of indicators by head teachers or inspectors Source: Adapted from Mizala & Romaguera (2004). School-based incentives p p p p p p p p p p p p p p p p p p p p p p p p p Teacher Incentives and Performance 41 be applied at an individual level or at school (or department) level. As noted already, input-based criteria are the most commonly used in teacher pay schemes and almost all of these are individually based. A few, such as additional payments to give teachers incentives to work in socially deprived, isolated or poorly resourced schools, are area or school based. Some pay schemes include additional pay for engaging in specific processes that are associated with improved student outcomes, such as professional development, providing extracurricular activities or working with parents, and these can be both individual and school based (Mizala & Romaguera, 2004). Performance-related pay (PRP) schemes relate some proportion of teachers’ salaries to one or more measures of student output. Some outputs can be measured quantitatively, such as drop-out and completion rates and student attainment. However, as noted above, to avoid moral hazard problems, value added measures of student outcomes must be used in order to control for factors that affect student attainment but are not influenced by the teacher’s performance. Indicators used for PRP may be poorly specified for two reasons. First, the indicator may be a poor measure of the actual output that the principal desires. For example, the principal may desire the maximum feasible student progress, but the actual indicator used is a poor measure of the value added by the teacher. Value added measures are subject to both measurement error, if the data are not of high quality, and to some random variation. An extreme case of a poorly specified indicator would be raw examination results as these would give schools and teachers incentives to cream-off higher-ability students and neglect the rest. The second reason for poorly specified output measures arises when principals desire multiple outputs from teachers but devise PRP schemes that include only some of the outputs and exclude those that cannot be easily measured. P-A theory predicts that outputs that are excluded from PRP but are substitutes in teachers’ allocation of time will be under-produced. Hence, there is a danger that giving teachers financial incentives to produce high value added examination scores might lead them to neglect other desirable student outputs, such as cultural appreciation, physical fitness, personal development, interpersonal skills and mutual tolerance.3 Of course, these undesirable teacher responses are predicted on the assumption that teachers are knaves and consequently will produce only what they are paid for. On the other hand, if teachers are pure knights PRP is unnecessary. An advantage of using the professional judgement of head teachers or inspectors to assess the performance of teachers, rather than just an indicator of student attainment, is that a wider range of criteria can be included so as to assess all the outputs desired of teachers. The disadvantage is the element of subjective judgement and hence the potential for inconsistency or unfairness. Another issue in devising teacher PRP schemes is whether they should be related to individual or team performance, as to some extent the quality of student learning at a school is dependent on teacher collaboration. PRP linked to individual teacher performance fails to reward collaborative teacher effort and so is likely to diminish it, especially if it promotes envy and conflict between teachers. PRP schemes can be devised for teams as well as for individuals. However, team PRP schemes can suffer from “freeriding” by knavish workers, who exploit the fact that their share in a collective reward cannot be denied them and will be paid to them even if they shirk. Free-riding behaviour can be prevented by moral suasion (so that all team members behave as knights) or by peer-group monitoring and exclusion of shirkers. As teachers are generally considered to be risk averse, an optimal contract must share risk between the agent and the principals. The pay scheme must therefore consist of a fixed 42 R. Levačić amount that is not dependent on performance plus a PRP element. A further distinction in the design of PRP schemes is how pay is increased in relation to performance. One type of pay scheme consists of a fixed amount plus a portion that rises, usually linearly, with performance as measured by prescribed indicators. The second type of incentive scheme pays an additional fixed amount if performance exceeds a given threshold. In the last 15 years or so there has been increasing interest in performance pay schemes for public sector teachers4 and attempts to implement them. These attempts are mostly in developed countries because these are more likely than developing and transition states to have the capacity needed to adopt costly monitoring and verification systems for evaluating teacher performance and have relatively little corruption or teacher capture of state decision-making. There has been some development of increased incentives for teachers in Latin America. According to Mizala & Romaguera’s (2004) survey of these in Bolvia, Chile, El Salvador and Mexico, only Chile used national assessment of students to evaluate teaching quality and took into account students’ background. El Salvador’s scheme assessed school quality on a number of dimensions but did not differentiate schools by students’ social background. Consequently, the resulting rewards for teachers were biased in favour of urban rather than rural schools. The authors concluded in favour of schoollevel rewards because “evaluation schemes at the school level and collective rewards adapt best to the characteristics of education, enhancing vital elements in this process, among them teamwork and a multi-product approach. Individual evaluations pose more problems” (Mizala & Romaguera, 2004, p. 750). 4. Evidence of the Relationship between the Teacher Incentives and Student Outcomes Whether performance-related pay schemes improve teacher performance in terms of students’ learning outcomes is an empirical question, the answer to which depends on both the nature of the PRP scheme and teacher motivation. Performance-related pay could have two distinct beneficial effects. If implemented nationally and resulting in significantly higher pay levels for better-performing teachers, it could attract more people into the teaching profession who have the qualities needed to be effective teachers. Second, PRP could raise the productivity of existing teachers by inducing more effort. The second hypothesis is easier to test as it can exploit variation in PRP schemes between otherwise similar schools. Also, the first effect will take a longer time to materialize. There is a limited but growing research literature on testing the hypothesis that paying teachers in relation to their performance results in higher student attainment. Since the mid-1990s some US states and school districts have started mandatory merit pay schemes, at both school and individual level, for example Florida, South Carolina, Denver and Dallas (Atkinson et al., 2004; Figlio & Kenny, 2007). Ladd (1999) found that student test results in Dallas improved compared with other large cities in Texas after it introduced school-based merit pay. However, this improvement had also occurred in the year before the Dallas scheme was implemented. Figlio & Kenny (2007) claim that theirs is the first study on individual teacher pay incentives and attainment using nation-wide US data—the National Education Longitudinal Survey of schools, supplemented by their own survey of teacher incentive schemes. Controlling for student prior attainment and background variables as well as school-level variables, they report that student Teacher Incentives and Performance 43 attainment was slightly higher in schools with merit pay. They further conclude that merit pay was more effective in raising student attainment when restricted to a small proportion of teachers.5 A form of performance-related pay was introduced in England in 2000 and continues to the present. It is a threshold scheme: teachers who have reached the top of the six-point main pay scale can apply for promotion to upper pay band 1, which pays an additional £2500, and after some time can apply for two further upper pay bands. Teachers have to demonstrate that they meet a set of criteria, including student progress. Applications are approved by the head teacher and appeals heard by the Governing Body. Roughly 80% of eligible teachers have applied for the first upper band and around 95% has been successful. When policy changes are introduced nation-wide their impact is difficult to evaluate because there are no ready control schools that have not experienced the initiative. In an evaluation of the effects of the English PRP scheme on secondary students’ attainment using a sample of schools that provided data on teachers, Atkinson et al. (2004) concluded that teachers who were eligible for PRP and received it showed student attainment gains, comparing 2 years before and after PRP, relative to ineligible teachers (i.e. those with 6 or fewer years’ experience).6 Lavy’s (2002) study has a more robust research design as its data are from an experiment in Israel in 1995 with 65 schools where teachers in the top two-thirds of schools were awarded additional pay if school results improved. The schools in the scheme had an average point increase in student scores of 1.75 points and the percentage of students sitting national examinations increased by 2.1%. This expenditure of money was more cost-effective than spending on an increase in resources, which was another experiment tried at the same time. A study by Glewwe et al. (2003) provides an example of a PRP scheme in a developing country that induced teacher “gaming” of the system rather than long-term gains to student learning. This was a school-based incentive scheme for a sample of rural Kenyan schools in which all teachers in grades 4 – 8 in winning schools received a bonus. Lopez-Acevedo’s (2004) study of factors affecting student attainment in Mexico included the nation-wide teacher incentive scheme, the Carrera Magistral introduced in 1992. Part of the Carrera Magistral was voluntary merit pay for which classroom teachers elected to be assessed in relation to their students’ learning achievement and their professional practice, including professional development. Lopez-Acevedo’s findings include that teachers electing for the Carrera Magistral had higher value added student attainment, as did students in schools with a high degree of supervision by the school principal. Empirical evidence of the effects of teacher PRP schemes on student attainment is still relatively sparse, reflecting the fact that only a minority of public school systems reward teachers in relation to student outcomes and that even fewer such schemes are set up so as to generate data that permit robust research designs for testing the impact of such schemes on student learning. The available evidence suggests some small positive effects on student learning. Prentice et al. (2007, p. 30) concur with this, adding that “most of the improvement appears to come from previously weak students performing better”. However, as P-A theory predicts, not all teacher PRP schemes are successful, as their effectiveness depends on the design of the contract and on the cultural and institutional context in which it is operated. Furthermore, as noted by Prentice et al., it is rare for a study of teacher PRP to include data for assessing whether such schemes, even if they improve student performance, increase social welfare overall. 44 R. Levačić 5. Conclusion P-A theory provides a strong framework for analysing the institutional arrangements that govern the work of teachers. It singles out several key factors that affect how contracts for teachers should be designed to ensure that teachers in public sector schools perform efficiently. These are the motivations of teachers, in particular the balance of selfish and altruistic motives, moral hazard that characterizes teaching and asymmetric information that can be countered only by investing resources in monitoring and verification of the teacher effort – outcome relationship. When teachers are largely knights then performance monitoring and financial rewards for performance are not needed in order to elicit good performance from teachers in promoting student learning. However, in some public education systems teachers are, for various reasons, more knavish in their motivations, and the contracts they work to give no disincentives for shirking. As the Indian example indicates, such situations arise when teacher interests succeed in capturing state decision-making so that the agents rather than the principals control the process of setting contracts. If it is impossible in these contexts for parents and taxpayer interests to prevail in political decision-making so as to improve incentives for public sector teacher performance, then the expansion of private sector schools, including those run by non-governmental organizations, remains the only option for securing improvements in education quality. The private sector solution has the drawback that it may disadvantage children from poor households when fees have to be paid. Public sector subsidy of such fees may lead to the build-up over time of political pressures that once more result in knavish teachers being protected by inefficient contractual arrangements. In political systems with inefficient teacher contracts but where it is possible to summon the political will to reform them, teacher incentives can be improved without going as far as performance-related pay. In Zymelman & DeStefano (1993), the following factors are associated with poor teacher incentives and can be removed without involving PRP: . . . . guaranteed appointment on completion of training; lifetime tenure—especially as civil servants; automatic annual salary increments over a lifetime; absence of performance assessment or sanction of dismissal for poor performance; . low relative pay in poor countries, which encourages corruption (bribes for results, private tutoring); . salary unrelated to the disutility of the teaching post (e.g. in rural areas in developing countries); . salary unrelated to those that teachers with specific qualifications could earn in the private sector. Naturally, the best situation is one where teaching is a highly respected occupation in society, and so attracts high-quality entrants who are intrinsically motivated and can be trusted to perform well for society with little monitoring. It is clear that in most countries it cannot be assumed that the current teaching force has these characteristics and in some unfortunate countries teachers are very far from this ideal. In countries where teachers are poorly motivated, improved incentives for teacher performance are an essential component of reforms aimed at raising the quality of the education system. Teacher Incentives and Performance 45 Notes 1 2 3 4 5 6 P-A theory usually assumes the principals are risk-neutral because they hold a diversified portfolio of income-earning assets. This is less likely to be valid for principals in the public sector, in particular politicians who lose votes from voter perception of failure to deliver promised levels of public service. I am using Minister of Education as a generic term that is internationally applicable rather than the actual title in use in England—Secretary of State. Dixit (2002) discusses these problems more extensively, including problems of multiple principals and multiple outputs. If the outputs are complements then having indicators for only a few of the outputs will not result in sub-optimal production, but if the outputs are substitutes, there is over-production of the outputs for which teachers are rewarded and under-production of the other outputs. I limit the discussion of PRP schemes to public sector teachers. The question of whether private sector teachers and hence schools perform better than their public sector equivalents, in part because of the different incentives for teachers, has spawned a considerable literature of its own, which I do not consider here. Figlio & Kenny (2007) acknowledge that even though they control for a wide range of variables affecting student attainment, their cross-section study cannot rule out that the results are due to selection bias, i.e. that more effective schools choose to have merit pay. These results depend on assuming that all of the gains for the ineligible teachers are due to increased experience—other studies have shown that teachers improve with experience in their first few years of teaching but not later on. References Atkinson, A. et al. (2004) Evaluating the Impact of PRP for Teachers in England, Working Paper No. 04/113, Centre for Market and Public Organization. Chevalier, A. & Dolton, P. (2005) The labour market for teachers, in: S. Machin & A. Vignoles (Eds) What’s the Good of Education? (Princeton: Princeton University Press). Dixit, A. (2002) Incentives and organizations in the public sector: an interpretative review, Journal of Human Resources, 37, pp. 696– 728. Douma, S. & Schreuder, H. (2002) Economic Approaches to Organizations (London: Pearson Education). Eide, E., Goldhaber, D. & Brewer, D. (2004) The teacher labour market and teacher quality, Oxford Review of Economic Policy, 20, pp. 230 –243. Fama, E. F. & Jensen, M. C. (1983) Separation of ownership and control, Journal of Law and Economics, 26, pp. 301–326. Fama, E. G. (1980) Agency problems and the theory of the firm, Journal of Political Economy, 88, pp. 288–307. Figlio, D. N. & Kenny, L. W. (2007) Individual teacher incentives and student performance, Journal of Public Economics, 91, pp. 901–914. Fuller, B. & Clarke, P. (1994) Raising school effects while ignoring culture? Local conditions and the influence of classroom tools, rules and pedagogy, Review of Educational Research, 64, pp. 119–157. Glewwe, P. & Kremer, M. (2006) Schools, teachers and educational outcomes in developing countries, in: E. Hanushek & F. Welch (Eds) Handbook on the Economics of Education (Amsterdam: North Holland). Glewwe, P., Ilias, N. & Kremer, M. (2003) Teacher Incentives, NBER Working Paper 9671. Goldstein, H. (1997) Methods in school effectiveness research, School Effectiveness and School Improvement, 8, pp. 369–395. Gottlemann-Duret, G. (2000) The Management of Primary Teachers in S Asia: A Synthesis Report (Paris: IIEP). Gregg, P., Grout, P., Ratciffe, A., Smith, S. & Windmeijer, F. (2008) How Important is Pro-social Behaviour in the Delivery of Public Services, CMPO Working Paper Series No. 08/197, available at: www.bris. ac.uk/Depts/CMPO (accessed 16 July 2008). Grindle, M. (2004) Interests, institutions, and reformers: the politics of education decentralization in Mexico, in: R. R. Kaufman & J. M. Nelson (Eds) Crucial Needs, Weak Incentives: Social Sector Reform, Democratization, and Globalization in Latin America (Washington, DC: Woodrow Wilson Press). Hanushek, E. (2005) Economic Outcomes and School Quality, available at: www.unesco.org/iiep (Paris: International Institute of Educational Planning). Kingdon, G. G. & Muzammil, M. (2003) The Political Economy of Education in India: Teacher Politics in Uttar Pradesh (Delhi: Oxford University Press). 46 R. Levačić Kingdon, G. G. & Teal, F. (2007) Does performance related pay for teachers improve student performance? Evidence from India, Economics of Education Review, 26, pp. 473–486. Ladd, H. E. (1999) The Dallas school accountability and incentive program: an evaluation of its impacts on student outcomes, Economics of Education Review, 18, pp. 1–16. Laukkanen, R. (2008) Finnish strategy for high-level education for all, in: N. Soguel & P. Jaccard (Eds) Governance and Performance of Education Systems (Dordrecht: Springer), pp. 305–324. Lavy, V. (2002) Evaluating the effects of teachers’ performance incentives on pupil achievement, Journal of Political Economy, 110, pp. 1286–1317. Le Grand, J. (2003) Motivation, Agency and Public Policy (Oxford: Oxford University Press). Lopez-Acevedo, G. (2004) Professional Development and Incentives for Teacher Performance in Schools in Mexico, World Bank Policy Research Working Paper No. 3236. Matthews, R. (1991) The economics of professional ethics: should the professions be more like business? The Economic Journal, 101, pp. 737–750. Mizala, A. & Romaguera, P. (2004) School and teacher performance incentives: the Latin American experience, International Journal of Educational Development, 24, pp. 739 –754. OECD (2004) Learning for Tomorrow’s World: First Results from PISA 2003, available at: www.pisa.oecd.org (Paris: OECD). OECD (2007a) Education at a Glance (Paris: OECD). OECD (2007b) PISA 2006: Science Competencies for Tomorrow’s World Vol 1 (Paris: OECD). Prentice, G., Burgess, S. & Propper, C. (2007) Performance Pay in the Public Sector: A Review of the Issues and Evidence, Office of Manpower Economics, available at: www.bristol.ac.uk/cmpo/publications/publicservices.html (accessed 16 July 2008). Pritchett, L. & Filmer, D. (1999) What education production functions really show: a positive theory of education expenditures, Economics of Education Review, 18, pp. 223–239. Reezigt, G. J., Guldemond, H. & Creemers, B. (1999) Empirical validity for a comprehensive model on educational effectiveness, School Effectiveness and School Improvement, 10, pp. 193– 216. Wilson, J. Q. (1989) What Government Agencies Do and Why They Do It (New York: Basic Books). Zymelman, M. & DeStefano, J. (1993) Primary school teacher salaries in Sub-Saharan Africa, in: J. B. Oliveira & J. P. Farrell (Eds) Teachers in Developing Countries (Washington, DC: The World Bank).
© Copyright 2025 Paperzz