The Malawian Context
Background on Educational Inequalities
Malawi is a landlocked country located in the South-East of Sub-Saharan Africa. It was a former British colony before achieving independence in 1964. Leadership was then assumed by President Hastings Kamuzu Banda, who ruled as an autocrat until 1994. He left the presidency after he was defeated by Bakili Muluzi in what was Malawi’s first democratic election. Malawi consists of 3 regions; the Northern, Southern and Central. One of the first issues which the post-colonial Malawian government had to address was that of salient and persistent regional asymmetries in the educational attainment and employment outcomes. In spite of only possessing 12-13% of the country’s population, the Northern region boasted a literacy rate which doubled the national average in 1945 (Heyneman, 1972) and, in the civil service in 1969, out of the 113 highest-level Malawian civil servants, the Northern Region held over 50 per cent of the places (Vail, 1989). This imbalance was also reflected in the enrollment rates of Malawi’s sole higher education institution, the University of Malawi, more than two decades after its inauguration in 1964. By 1987, multiple sources claimed that the share of Northerners among entrants was as high as 50% (Posner, 1995; Carver, 1990). These disparities can be shown in Table (1), which evidences the North surpasses the other regions in terms of regional population shares of highest qualification and average years of schooling in 1987.
A potential source of these inequalities has been ascribed to the persistence of educational policies undertaken in the colonial era. During this era, education was predominantly supplied by church missions which established bases in various parts of the colony, and took it upon themselves to educate the surrounding population as part of a wider eﬀort to Christianize them (Heyneman, 1972). While virtually all missions were of British origin, they possessed a degree of autonomy with regards to how they would fulfill this mandate, and the divergent philosophies of these missions, sources claim, had a tremendous impact on the trajectories of educational attainment. In particular, the mission which mainly operated in the Northern region was particularly progressive, and adapted curricula which emphasized highly skilled and academic tasks intended to prepare students for participation in leadership roles within the mission and society at large (Heyneman, 1972). On the contrary, the dominant missions in the Central and Southern Region were not equally ambitious. The typical Malawian student within their remit was rarely equipped with more than basic literacy and the ability to perform vocational tasks, with the Dutch Reformed Church Mission in the central region not even teaching English (Lamba, 1984).
Of course, the reality on the ground was far more nuanced than the Northern region dominating the upper echelon of academia and the civil service with a Livingstonia-educated bourgeoisie, and being the sole region to provide key figures in the decolonisation push or post-colonial elite. The southern region also possessed a notable educated middle class which served in the civil service (Carver, 1990), and would serve as the backdrop for the infamous 1915 anti-colonial uprising led by John Chilembwe (also of southern origin). Furthermore, other factors likely influenced the trajectory of regional education and occupation outcomes, conditional on which missions which established there2. Nevertheless, the narrative which the government at the time had, and most sources on the subject have, espoused e.g. (Vail, 1989; Mkandawire, 2010; Galafa, 2019) is that of colonial legacy being necessarily deterministic.
The Structure of the Quota
In order to deal with the imbalances in higher education enrollment and employment outcomes, the government of Malawi issued a directive mandating the The University Council of Malawi to select its students from the applicant pool on the basis of aﬃrmative action rather than merit with eﬀect from the academic year beginning in 1987 (Mhango, 1993). This took form in the shape of the infamous Quota System, whereby: (i) each district was guaranteed 10 places; and quota was « discriminatory and of no solid foundation » (Mhango, 1993). The anticipated winners and losers of this are both implicit and explicit. Explicitly, this reform reallocated seats from students originating from districts which were systematically over-represented in the University student-body relative to their population share, to those from districts which were relatively under-represented. Implicitly, the reform reallocated seats from students originating from the country’s Northern districts, to those from the Central and Southern districts.
The Structure of Higher Education
At the time, Malawi only had one university: the University of Malawi. Given the rarity of higher education in and of itself, seeking higher education abroad was an economic impossibility. Therefore, virtually all of Malawi’s graduates attended the University of Malawi. Higher education in Malawi typically lasted for 4 years, and university entrants would be determined on the sole basis of their performance on the exams taken at the end of the final year of secondary school. From the perspective of the university, seats were allocated based on secondary school exam performance and the requirements of the quota. The quota was applied centrally, such that the aggregate student body was mandated to be in line with quota requirements (i.e. representative of district-level population shares) while student shares within certain faculties needn’t comply with this. For instance, in a scenario of perfect compliance, we would expect the share of students from Northern districts in the total university population to be 12%, while the share of Northern students in the Accounting course could be much higher than this.
The Political Context
In order to better attribute the potential sources of the behavioural responses driving the empirical results to come, it is important to understand the political climate in which this quota was applied. In contrast to other prominent contexts in which AA has been adopted, the pre-text for AA in Malawi was not one of compensating a sub-group for decades of discrimination which resulted in their educational outcomes lagging behind the remainder of the population. Furthermore, the political climate in which AA was applied at the time was not typified by neutrality and fairness, but blatant ethno-regional favouritism and divide. From the outset of his 30 year rule, President Banda secured and consolidated his place as Malawi’s autocrat by heightening ethno-regional fractures. He then used those fractures to attract support from his own ethnic group (the Chewa in the central region) and other interest groups, while simultaneously undermining his rivals, detractors and ordinary citizens, particularly those in the North.
When Banda was challenged by leading cabinet ministers from the Northern and Southern regions, the political fallout which ensued was regional in nature even though the initial disagreement was ideological. The ministers were summarily expelled, Chiefs from the Northern and Southern regions were dismissed, and numerous regoinal councils in the Northern and Southern regions were dissolved. The Central region (Banda’s region of origin) was unscathed and none of the dismissed cabinet ministers were Chewa (Banda’s tribe). Banda then used patronage to cement political support for his regime by issuing land reforms in 1965 which gave him the power to grant leases for estates throughout the country. This enabled him to use land to secure loyalty among the middle-class. Furthermore, he dispensed vast economic and strategic privileges upon the central region. Notably, the government moved the national capital from the southern to the central region, quadrupled the central region’s share in development expenditure from 11% to 40% between 1967 and 1972, and concentrated the majority of agricultural loans and economic gain to farmers in the central region (Vail, 1989).
Central favoritism went hand in hand with strategic discrimination against the Southern and Northern regions, with a particular emphasis being placed on education in order to undo the northern domination of the educated elite and the civil service. A highly symbolic instance of this was the illegalisation of the northern Tumbuka language from the press and radio, and the adoption of Chewa as the oﬃcial language in 1968. This had the two pronged agenda of synonymising Malawian culture with Chewa culture and frustrating northern education prospects, as it necessitated that Chewa be taken as a subject in school which, if failed, would force the failing student to resit all exams. The Malawi Examinations Board replaced the Cambridge Overseas Examination, and made grade requirements for entry into secondary school higher for northern and southern students relative to central students. Finally, between 1973-1976 the administration within the University of Malawi and the civil service were purged of non-Chewa, southern and northern individuals in favor of Chewa candidates, in order to secure a steady stream of loyal Chewa bureaucrats. These are just a handful of numerous examples which evidence Banda’s intentions to not only advantage the central region, but also disadvantage the southern and northern regions (Vail, 1989; Carver, 1990).
The Repatriation of Northern Teachers
A policy reform which arose independently to the quota, but one which likely had serious interactions with it, was the repatriation of northern teachers in 1989. Following a series of speeches in which Banda accused northerners of carrying out policies to stunt development in the south and center, and allegations that northern teachers were sabotaging the academic performance of non-northern students, Banda issued an instruction that all teachers should return to teach in their regions of birth. While this was not instituted into law, compliance to this demand was suﬃciently widespread for it to be felt and have telling ramifications. Carver (1990), which was written at the time, indicated that it was « implemented to a large degree, with many northern teachers having to return to lower paid jobs – or unemployment – in their own region ». While there is no quantitative analysis of the impact of this reform, Mkandawire (2010) provides the anecdote of Mitundu Secondary School in Lilongwe, a school in the central region which was reduced to just two teachers after the repatriation.
Given the findings in the literature, the nature of the quota and the setting in Malawi, it is possible to anticipate some of the potential discernible impacts of the quota, particularly regarding pre-university eﬀort, university entry and university completion. The first and most obvious impact pertains to university entry. If the quota was perfectly enforced, we would anticipate that the share of university students originating from districts which were historically over-represented relative to their population shares before the quota would decline after the quota, such that their university shares correspond to their population shares. We would also anticipate the analogous result for historically under-represented districts, but in the opposite direction. Implicitly, the same analysis would roughly hold at the regional level. Given the over-representation of the northern districts relative to central and southern districts, we would expect the relative university share of the northern districts to also decline. This would, of course, only apply if the districts can produce enough eligible students to fill their allocated number of seats, such that the constraint binds. Conditional on this, the aforementioned eﬀects would arise mechanically if the quota were stringently enforced.
In order to better understand the implications of the quota on pre-university eﬀort, let us consider a simplified thought experiment motivated by the theoretical model adapted by Cotton et al. (2018).
A Simple Example
Let us assume that student ability θi follows a continuous distribution and is bounded, such that there exists a spectrum of individual abilities in the entire student population. Let us further assume that there are two districts of equal population size: the northern district, N, and the southern district, S. Let the total number of students in the North and South be equal to NN and NS, respectively, and NN = N S. Let the ability distribution of students in the South, F (θi)S, be likelihood ratio dominated by the ability distribution in the North, F (θi)N . The most important implication of this is that the value of θi at each quantile of the ability distribution is higher for the North, such that the cumulative distribution function (CDF) of ability for the South is to the left of that of the North for every value of θi.
For simplicity, let us assume that university provides a positive and identical payoﬀ, P , for all students (e.g. due to securing a financially rewarding job after graduation) and that universities select students on the sole basis of the observed human capital, hi, which students produce (e.g. test scores). Furthermore, let us assume that there are fewer university seats, NC , than there are students in each district, such that NC < NN = NS only some students qualify for university. Let us further assume that the production of human capital is costly in terms of utility, its production technology depends positively and entirely on ability, and that this technology is identical for all students. It follows that the utility of student i corresponds to: Ui = A × P − hi(θi), where A = 1 if the student qualifies for university and A = 0 otherwise. Let’s assume that students are rational and maximise utility and that there is some level of human capital hi(θi) for each agent for which Ui = P − hi(θi) < 0 such that there is no incentive to go to university. It follows that the human capital exerted by each student exists in the interval hi ∈ [0, hi(θi)]. It also follows that hi(θi) for low ability students is relatively low, while that of high ability students is relatively high. Finally, let us assume that students know the entire distribution of ability, such that they know their own ability and that of every other student in the population.
Without a Quota
Let us now consider a scenario in which university considers students on the basis of free entry, such that a student competes with the entire population of students. It follows that, conditional on each realisation of θi being unique, there is some optimal level of human capital hi(θi)min which results in all university seats being perfectly filled with no shortage or excess. It also follows that hi(θi)min corresponds to the maximal human capital of the marginal student who just failed to qualify. This occurs because there is perfect knowledge of student ability, and the next subsequent student to this marginally failing student would have no incentive to produce a level of human capital more than this, as this is the minimum amount needed to qualify and the production of human capital is costly. By the same logic, all other qualifying students generate human capital equal to hi(θi)min, and all non-qualifying students produce no human capital, as there is no incentive to behave diﬀerently as far as utility is concerned.
As far as the distribution of students in university is concerned, by virtue of the dominance of F (θi)N over F (θi)S, it follows that the share of students from the South will be less than its population share, 50%, while that of the North will exceed its population share. This is because there will be a higher share of students in the North for whom hi(θi) > hi(θi)min. Hence, a no quota situation results in the North being over-represented in university and the south being under-represented.
With a Quota
Now let us consider what may happen when a quota is introduced. The university still selects students on the sole basis of human capital production, but now allocates an equal number of seats for students from the North and South. Thereafter, the university selects the best performers from each district to fill the allocated seats. In order to see how student behaviour changes according to student ability, we simply carry out the previous analysis but at the district level.
Let the prohibitive levels of human capital for the marginally failing students in the South and North be equal to hi(θi)Smin and hi(θi)Nmin, respectively. Given the dominance of the dominance of F (θi)N over F (θi)S, it follows that hi(θi)Smin < hi(θi)min < hi(θi)Nmin. This means that students in the South who had no incentive to produce human capital before the quota due to their value of θi being too low now produce human capital after the quota. This is the discouragement eﬀect at work; the quota reduces the relative competition which Southern students face such that the level of human capital required to qualify is no longer prohibitive for low ability students. However, not all Southern students increase their human capital production. There are some students who still produce no human capital as hi(θi)Smin is still prohibitive. Furthermore, there are some students, those who qualified for university without the quota, who actually reduce their human capital production as the human capital required for university entry is less demanding than before.
The converse is true for the northern district. The relative competition increases, and this forces students for whom hi(θi)min < hi(θi) < hi(θi)Nmin to reduce their human capital production to 0, while the remaining subset of qualifying students increase their human capital levels above the pre-quota level. Consequently, the district shares of students in university equals their population shares, while the pre-university eﬀort levels increase for some and decrease for others in both districts.
On aggregate, human capital (or eﬀort) can either increase or decrease depending on the pre and post-quota eﬀort levels, and the share of seats allocated to each district post-quota. In particular, pre and post-quota human capital are equal if hi(θi)min = qN × hi(θi)Nmin + (1 − qN ) × hi(θi)Smin
Where qN is the share of seats allocated to the North under a quota.
The aforementioned model is very simple and makes numerous assumptions. For instance, it assumes that schooling (or « human capital » in the model) has no inherent productive purpose beyond serving as a signal for universities. This would be in line with the Spence Model (Michael, 1973) of human capital, in which the only purpose which education serves to students is as a signal for universities to be able to discern their unobserved types. It, however, completely neglects the Beckerian perspective on the perceived role of education as a means of developing human capital (Becker, 1973). However, it still roughly captures the competitive elements of the pre-university eﬀort decisions which students face before and after the quota.
By virtue of the quota restricting a given student’s competition pool to individuals from their own district of origin, the ability distribution of the student’s competitors eﬀectively shifts, conditional on the student’s ability. For students from relatively under-represented districts, the quality of competition likely reduced, as the historical under-representation of their districts suggests that the level of ability of students within these districts was relatively low compared to the overall population. Lowering the ability of the student’s competition would increase the probability of the student entering university, conditional on the student’s own ability. This, in turn, may have increased pre-university eﬀort for some students who, in absence of the quota, may have not qualified for university without the quota but now can because of the quota. The empirical literature suggests that this may be true in terms of the intensive margin (i.e. test scores), while there has been no notable research done on extensive margin eﬀects (i.e. secondary school completion). The converse of this would be true for over-represented (northern) districts: due to greater competition, the discouragement eﬀect would reduce pre-university eﬀort and attainment at the bottom of the ability distribution, while increasing it at the top.
Regarding aggregate eﬀects, the literature suggests that the discouragement eﬀects for the losers of AA (and indeed some winners in under-represented districts) in this respect are minimal, at least relative to the encouragement eﬀects for the winners, such that a net-gain is possible. Even if there is a reduction in aggregate eﬀort, there may still be gains in terms of allocative/pareto eﬃciency if students from under-represented districts value education more than students from over-represented districts (however, I feel this to be highly debatable in this context).
With regards the impact of the quota on university completion, the debate in the literature gravitates around the mismatch hypothesis. Mismatch refers to a situation in which AA essentially sets up minority students to fail by allocating them seats in prestigious programs and institutions even though these students lack the required preparation to succeed. This can be illustrated by hearkening back to the exercise developed in previous discussion.
Suppose we are considering the students enrolled in the university when there is no quota applied. Every student enrolled has a level ability such that the cost of developing human capital in the entry phase was not prohibitive, in particular hi(θi) > hi(θi)min for each student. Let us call the distribution of ability observed among Southern entrants ΘSF , where F represents free entry. When the quota is applied, the entry requirements for Southern entrants become less stringent, such that the distribution of abilities now also consists of students that were not suﬃciently able to enter before the quota. In particular, those for whom hi(θi)Smin < hi(θi) < hi(θi)min. It follows that the lower bound for the ability distribution of Southern entrants under AA, ΘSAA, is lower than that observed before ΘSF , such that Southern university students are now, on average, less able.
Suppose there were some minimum level ability required to graduate from university, θmin which remains the same before and after the quota. Let us assume that the lower bound of abilities for the Southern students before the quota was below this value and that the upper bound was above this value, such that there was some failure already present but not all students failed. It would necessarily be the case the failure rate among Southern students would increase. If we make the analogous assumption for the Northern students, then the Northern failure rates would necessarily decrease.
With application to the Malawian context, it would be reasonable to expect that the quota increased the university failure rate in the under-represented (southern/central) districts, while potentially reducing the failure rate among over-represented (northern) districts.
Table of contents :
2 Literature Review
3 The Malawian Context
3.1 Background on Educational Inequalities
3.2 The Structure of the Quota
3.3 The Structure of Higher Education
3.4 The Political Context
3.5 The Repatriation of Northern Teachers
4 Anticipated Effects
4.1 University Entry
4.2 Pre-University Effort
4.2.1 A Simple Example
4.2.2 Without a Quota
4.2.3 With a Quota
4.3 University Completion
4.4 Importance of the Malawi an Context
4.4.1 The Repatriation : Mismatch and Discouragement
4.4.2 Student Perceptions and Discrimination
4.4.3 Limited Capacity for Enforcement
4.4.4 Annulment of the Quota
5.1 Outcomes of Interest
5.1.1 Pre-University Effort
5.1.2 University Entry
5.1.3 University Completion
5.2 Measuring Historical Over-representation
5.3 Defining Cohorts
6 Empirical Strategy
6.1 What I Would Like to Estimate
6.2 What I can Actually Estimate
6.3 Analytical Illustration
6.4 Empirical Specification
7.1 Graphical Analysis
7.1.1 Age-level Analysis
7.1.2 Generating cohorts
7.1.4 Over-representation Analysis
7.2 Regression Analysis
8 Additional Analysis
8.1 Life time Outcomes
8.2 Comparison of elites
9.2 Theoretical and Policy Implications
9.3 Potential Extensions
11 Table Appendix
12 Graph Appendix