Assessing Income Tax Perturbations

Taking a piecemeal tax reform approach to tax analysis in the spirit of Feldstein (1976), we establish a framework for assessing perturbations of the income tax schedule. It decomposes a reform into a change in tax level and a structural reform part. Focussing on the latter, the analysis singles out four effects: A social efficiency effect measured as the change in tax revenue due to behavioural changes, distributional impact due to mechanical effects, total distributional effects, and overall welfare effects conditional on inequality aversion. When applying our approach to changes in the piecewise linear income tax in Norway during 2016-2018, we identify the cut-off value for the inequality aversion for which the reform is just welfare preserving. For lower inequality aversion the decision makers have accepted a reform which enhances social efficiency at the expense of redistribution in favour of the better-off households.

1 For a historical account of these relations, see the National Human Development Report of 2010. 2 Despite the large number of coups and rebellions the level of human casualty had been relatively low in Comoros and pales in significance to the much larger number of casualties in the brutal wars in other African countries. 3 Where they stated that the likelihood of conflict escalation is higher if the majority community as a % of total population is in the range of 45 -90%. before Anjouan in 1997, though they re-integrated with the Union in 1998. While the island rights movements have been quite strong, the separatist sentiments are not nearly as strong as those in Anjouan.
There has been substantial resistance by the various actors at the national government to decentralization. With the approval of the Constitution of 2001, the government established the Autonomous Island Governments of Anjouan, Moheli and Ngazidja, providing these governments with extensive powers in a quite devolved system of government. This Constitution worked under the subsidiary principle, in which the national government held a small set of defined responsibilities, and the rest passed on to the island governments -at least in the law. It also provided for the financial autonomy of the island governments. The constitutional reforms of 2009, approved soon after the conflict in Anjuoan, transformed these power sharing arrangements, representing a general recentralization of power: it annulled the subsidiary principle by defining a set of responsibilities for island governments, and also demoted the island "Presidents" to governors. Informally, there was also substantial resistance to decentralization, with many central government officials refusing to be transferred to the island government. The recentralization of island financial decisions, concurrent with the overall movement towards recentralization, is discussed below.

Civil service recruitment and governance in Comoros
In recent years, Comoros has seen a large increase in both the overall sum of the wage bill and the number of public servants. From 2001 to 2012, the number of members of the civil service more than doubled, from approximately 6,500 to up to 13,500 in 2012. 4,5 At the same time, the wage bill is quite large, even relative to other African nations, as exhibited by Chart 3.1. With few exceptions, the wage bill in Comoros exceeds, sometimes even vastly, that of the comparison countries. The ratio of wages to different indicators confirms that the wage bill is substantial. The ratio of wages to primary domestic expenditures, where Comoros exceeds the 5 rest of the comparison countries, indicates that the wage bill is crowding out other forms of spending, especially investment.

Comoros
Furthermore, the wage bill in Comoros has been increasing substantially in recent years; from 2000 to 2011, the wage bill increased by 201%, while domestic revenues increased 174%. Such a sizeable wage bill at least partially contributes to the chronic delays in the payment of civil servants, a problem that has plagued Comoros for over a decade, but seems to have abated as of 2012 until the date of publishing this paper. Previously, the government would simply fail to pay its employees; for instance, in 1995, the government did not pay its employees approximately half of their salaries (Le Gay, 2001). Since then, delays in salary payments have become a routine phenomenon. Government employees during much of the last decade expected that they would not receive salary payments for months at a time. Civil servants seek sources of credit during the period of nonpayment, while others find alternative sources of income, contributing to chronic absenteeism.

Chart 3.2 Increases in forms of spending and revenue for the Government of Comoros
Even more importantly, salary arrears, as well as the stagnant wages of the civil service, have led to increasing levels of conflict between the government and public sector trade unions, especially the teachers union. Teachers held a strike that shut public schools for approximately six months until December 2011, during which no children attended public schools. The government only resolved the issue after threatening to fire teachers if they did not return to work, and subsequently suspended a number of teachers who refused to do so.
The growth of the public sector, and the related increases in the wage bill, crowd out other forms of spending, such as public investment, which is quite low in Comoros and driven almost exclusively by donors. In 2007, the percentage of total expenditures used for domestically financed public capital expenditure was just 3.8%, as opposed to 16.8% for foreign financed public capital expenditure. 6 In other words, the Government of Comoros finances little of its own capital investment, the vast majority coming from donors. Such a situation extends to even the most basic equipment necessary to perform government functions. For instance, one recent study estimated that there were fewer than 2,000 chairs and just over 1,000 tables for over 4,000 public sector employees who would need them (Government of Comoros, 2009). During visits to the Education Commissions in the Island Governments -a commission being the island government equivalent of a ministry, officials claimed they depended on UNICEF to provide such basic materials as computers, paper, and even pens.  As Chart 3.3 shows, the salary of the average civil servant ranges toward the low end, as measured in proportion to the GDP per capita of the country. Such a low ratio is not surprising, given the fact that the wage index in Comoros has not been systematically adjusted in over a decade, while government policy is generally to allow advancement along this scale only in exceptional cases. There is also no adjustment of salaries to inflation, meaning that salaries are reduced every year relative to the cost of living.

Chart 3.3 Ratio of Average Civil Servant Remuneration to nominal GDP per capita among African countries in recent years
The increase in the wage bill might be justified with commensurate increases in service delivery quality. Indeed, Comoros maintains high teacher to student ratios, particularly when compared to Sub-Saharan Africa. The relatively low student to teacher ratio found in Chart 3.4 should lead to an advanced quality of schooling, at least in basic subject areas, though, without figures of teacher absenteeism and double shifts, it is difficult to determine whether these ratios represent actual teachers in the classroom. However, the literature on the determinants of the quality of education emphasize that the student to teacher ratio only contributes slightly towards overall quality of education, with numerous other factors playing an important role -for instance, materials, good learning conditions, well trained teachers and parent involvement in the school.

Chart 3.4 Student to teacher ratios in Comoros and Sub-Saharan Africa as a whole, 2008.
Measures of student competence suggest that the system is not succeeding in teaching students basic subjects such as math and science. Nationally, only 18.4% of students have a minimum competence in math at the 8 th Year level, while 27.3% have such competence in science, as Chart 3.5 below demonstrates. At the high school level, only 29% of public high school exam takers in Ngazidja passed their high school degree exam, called the baccalauréat (BAC), while over 44% of private high school exam takers passed this exam. 7 Such poor performance in both sets of schools suggests that the system fails to prepare students properly for these exams, and that the public system in particular does not educate their students to a level necessary to graduate them to university.

Chart 3.5 Competence of 8th Year Students in Math and
Overall, the efforts to reform the civil service are not new. Indeed, the government, along with international institutions, has undertaken a variety of programs to address the sizeable wage bill, salary arrears and non-meritocratic recruitment of civil servants. Some recent steps are being taken towards restructuring the civil service. The Government of Comoros has launched the Economic Governance Technical Assistance Project, which works on a number of fronts, one 7 of which is civil service reform. Regarding recruitment, it supports implementation of a meritocratic and competitive hiring process by strengthening a government institution called the High Authority of the Civil Service -"Haute autorité de la fonction publique". The project also supports the Government of Comoros to streamline its civil service and to control the overall wage bill, through such measures as the civil service census and reinforcing the legal budget framework, including the position of financial controllers in the Ministry of Finance.

Section 3: Political economy and the design of the study
The importance of political economy The World Bank assessments of civil service reform (CSR) often identify political economy issues as a key impediment (Reid, undated;Public Sector Board, 2000;Independent Evaluation Group, 2008). The particular reasons for the poor performance of CSR projects are stated in Reid (undated), which identifies three main factors that political economy analyses can help to address: 1) Political economy constraints, 2) Inappropriate design, 3) Inappropriate instruments. 8

Chart 4.1 The Contribution of Political Economy Analysis to Civil Service Reform
Source: Summary of Reid (undated), with contribution by the authors.
Chart 4.1 above summarizes the three main contextual challenges presented in the World Bank literature on CSR. The first relates to the interests that are challenged by CSRs. More than anything, politicians would be wary of reforming the civil service, as they may operate in a system of political clientelism where civil service jobs are used to mobilize support. The capacity of the government to effectively implement CSR represents another key consideration. Reforms can be highly complex, in terms of institutional design and technical skills. For example, they often require strong human resources management skills, in order to implement changes in the pay structure or hiring practices. They may also face challenges in translating legislation into actual changes in bureaucratic practices.
Attempts to tailor CSR must also consider the institutional design of government, particularly, in Comoros, to adapt to both the French system and devolved island governments. Moreover, given the sparse legal framework, informal bureaucratic practices dominate 8 The other factors are 1) Insufficient implementation support and 2) Contradictory bank-supported interventions.

Considerations/ Challenges
Solutions Analyses  Institutional/ organizational analysis, including the informal rules and organizational culture recruitment processes. For instance, the recruitment process may not be clearly described in the law, but certain procedures will be followed, including approvals via signature by various officials. Similarly, the criteria for those decisions may not be clearly outlined. Such practices are often difficult to transform, particularly as changes in laws and regulations are not implemented such that the bureaucracy stops the informal practices and follows new formal ones.

Research Design
The complexity of the topic, including the number of officials and members of the public involved, as well as the fact that most decisions are made outside the public eye, necessitates a multi-faceted research approach that triangulates information. The research design therefore includes three parts, including the following:

Part One: Quantitative Data Analysis of Civil Service Database
The research begins with an analysis of trends in government employment, in which the main source of data is a questionnaire completed by practically all civil servants, with a total of nearly 11,000 observations. The peak recruitment periods are correlated with key historical events that are suggestive of explanatory factors; similarly the educational qualifications for civil servants are analyzed over time. The questionnaire gathered information on several particular topics, including years of recruitment, the home village of the civil servant, and the level and type of education, among others. This questionnaire was disseminated together with the overall civil service census, which was completed under the World Bank supported Good Economic Governance Support project via the High Authority of the Civil Service, a civil service unit that reports directly to the President of the Union of Comoros. A copy of the questionnaire appears in Appendix 2.

Part Two: Community Level Study
The research consists of interviews in a sampling of systematically selected communities on all three islands of Comoros, with over 400 individuals interviewed. Here, the research probed both the experience of local civil servants and other job seekers, as well as the nature of the patron-client relationship in communities. These communities were randomly selected from a targeted set of socio-economic groupings, so as to provide a sense of recruitment practices in different types of communities. As such, in Ngazidja, which has the largest population of the three islands, communities were divided into nine categories, based on a low, medium or high level of poverty and a small, medium or large population; a community was then randomly selected from that category, for instance, one community representing the small population size and relatively low level of poverty. In Anjouan, four communities were selected randomly, while in Moheli, two were selected randomly. Interviews in these communities focused on several issues: first, the process of hiring for those members of the community who work for the public sector; second, any support that the member communities may have offered in return for public employment; third, the social position and other characteristics (such as gender) of those who access public employment. Furthermore, because of the targeted sampling strategy, the research attempted to determine the characteristics of the communities whose members successfully access public employment. Interviews included village elites, women, public employees -particularly teachers -and youth. The teacher interviews complemented the above case study. A more complete listing of the communities visited appears in Appendix 3.

Part Three: Institutional Analysis of Recruitment, with a special focus on the Ministry/Commissions of Education
The research includes a case study of the Ministry and Commissions of Education (commission is the Island Government version of a ministry), in order to understand the recruitment practices of government from the inside, both in terms of the official and unofficial processes, which offer an insight into the drivers and constraints involved. The selection of the public education sector is based on its importance in government hiring. The Ministry and Commissions of Education employ the greatest number of civil servants, according to the data, with more than half of the total; it also manages the largest budget within the wage bill. This case study builds on the quantitative analysis described above, probing the motivation to create jobs, as well as the process of filling those jobs. To complete this case study, our research team conducted over 30 interviews with those in charge of the hiring process, primarily the Secretary Generals and Bureau Chiefs of Ministries and Commissions at both the Union and Island levels, as well as the officials involved in the Civil Service Ministries and Ministry of Finance. A semistructured questionnaire was designed and utilized in these interviews, probing the formal and informal process of hiring, the criteria involved in decisions, the interests at play, the effects of hiring on other forms of spending, and others. This perspective from bureaucrats inside the government was complemented with interviews with over 75 teachers in communities from the three islands, as part of the community case studies.
The conclusions developed from the three empirical analyses were complemented with more than 50 semi-structured interviews with a number of key informants, such as civil society, academics, trade unionists, development partners, media and government officials at both the national and island levels. These interviews explored a number of areas, such as the interest groups who would support or oppose reforming the system, and the sensitivities that derive from the separatist movements, among others.

Recruitment into the civil service
The following analysis provides different measures of recruitment into the civil service. The questionnaire that supports this analysis does not measure the actual recruitment year by year, but rather the recruitment of the staff at the time of the civil service census, during the summer and fall of 2011.

Chart 5.1 Period of Recruitment for the Present Civil Service
Source: Author database, 2011. have seen a large growth in the civil service, particularly given that the civil service census of 2001 stated that there were a total of 6,500 civil servants just a decade ago. Source: Author database, 2011. Note: the numbers inside the bars represent the number of civil servants in the questionnaire recruited in that year by either the Union or Island governments. Note: 2011 data is incomplete as the questionnaire was completed in a rolling fashion over that summer.

Chart 5.2 Recruitment of Present Civil Servants from 2000-2011, by Union and Ngazidja/Anjouan
Chart 5.2 clearly shows several trends in the recruitment of civil servants. First, in numerous years of the last decade, the annual recruitment exceeded the normal 3% annual turnover that may be expected in an administration, consisting of retirements, voluntary departures and natural growth. Second, the vast majority of recruitment takes place in the Island Governments of Anjouan and Ngazidja, which together recruited 3,073 agents from 2006-2011, versus 943 agents in the Union Government. Third, an analysis of the breakdown of recruitment by year shows that recruitment is not consistent every year, but rather comes in peaks. The data from the last decade point to three peak years of recruitment into the civil service. By far, the year of the greatest amount of recruitment is 2010, when a total of 1,263 individuals entered the civil service, of which 994 joined the island governments of Ngazidja and Anjouan, and 269 in the Union Government. Interestingly, elections were held in 2010 for both the Union President and Island Governors. Such a correlation between the election and extensive recruitment suggests a political motive, a story that is further reinforced in the following sections of the paper. Historical explanations for the other peak years, 2004 and 2008, emerge from an analysis of the island recruitment years. Despite the extensive recruitment that took place in 2010, the incumbents for the position of governor in Anjouan and Ngazidja both lost in the elections of December 2010. The governor of Anjouan ran for re-election in 2010, having won the governor race in 2008 after the removal of Bacar from power. Similarly, the governor of Ngazidja ran for re-election and lost. Some reasons for these outcomes are explored in the following section.

Chart 5.3 Recruitment of Present Civil Servants in the Government of Ngazidja by Year
Source: Author database, 2011. Note: 2011 data is incomplete as the questionnaire was completed in a rolling fashion over that summer.
As mentioned, one of the peak years for government recruitment was 2004, in which 780 civil servants joined. Nearly half of the civil servants recruited that year, 353 in total, joined the Ngazidja Government. Chart 5.3 above shows that after 2010, 2004 is the second largest year for recruitment for the Ngazidja Government. 2004 is also a peak year for the Union Government, again the second largest in the past decade with 257 civil servants recruited.
Upon the creation of autonomous island governments with the Constitution of 2001, Ngazidja presented a special case for devolution, due to the presence of the Union Government. In other words, with the new Constitution, the island territory would contain both the Union and Island Government, and not simply the Union government as before. The government of Anjouan contained staff that had already transferred authority. In Ngazidja, this had not occurred, as a number of civil servants were transferred from the Union government to the Island government only in the years after the new constitution, mainly 2003-2004. Interviews with senior officials in the Ngazidja Island government suggest that the Union Government was resistant to transfer a large number of staff to the newly established island government. These senior officials describe the scene as "chaotic" decentralization, where staff were unclear as to whether they were moving or not, with many simply refusing to shift. The Ngazidja Island government therefore used the lack of transfers as a justification for further recruitment of civil servants; at the same time, in the Union Government, observers stated that the government similarly used the loss of personnel as a reason to recruit, such that they might be replaced. Combined, these observations point to a process of decentralization that included a disorderly transfer of staff and recruitment and served to duplicate many positions, leading to enormous growth of the civil service. As such, the hypothesis that these dynamics of decentralization, including the resistance of staff to transferring and the unplanned means by which it occurred, would lead to recruitment may indeed apply, especially in 2004, a topic that deserves further scrutiny in future research.

Chart 5.4 Recruitment of Present Civil Servants into the Government of Anjouan, By Year
Source: Author database, 2011. Note: 2011 data is incomplete as the questionnaire was completed in a rolling fashion over that summer.
The data above show that, like Ngazidja, 2010 was a peak recruitment year as well, correlating with the election for island governor. 2008 was the second largest recruitment year for the entire Government of Comoros, with 859 recruits entering the civil service. The Government of Anjouan primarily hired staff in this year, when 648 civil servants were recruited. 2008 was the same year in which the African Union force removed Mohamed Bacar from power, after his illegal takeover of the executive in 2007. According to interviews with officials and numerous civil servants in communities, the Bacar regime recruited a number of volunteer and contractual workers, often as teachers. After the downfall of the Bacar regime, the newly elected president of the Island Government of Anjouan decided to normalize many of these volunteer and contractual workers into the civil service, leading to the sudden jump in 2008. The motive behind this recruitment was unclear. Various key informants suggested that it represented a gesture of reconciliation with the citizens in Anjouan, though it may have also been motivated by the desire to win their political support, either for the Union or for the executives involved. Such events suggest that devolution leads to an increase in recruitment, though, unlike the more traditional manner described above, recruitment was not the result a 'chaotic' decentralization, but rather a reaction to a period of conflict.

Chart 5.6 Recruitment in the Governments of Anjouan and Ngazidja in 2010 by Commission
Source: Author database, 2011.
Recruitment within the civil service, by Ministry or Commission, is somewhat different between the Union and Island Governments, as this analysis of 2010 suggests. In the Union Government, recruitment centered on the Ministry of Interior, which mainly employs police officers. Island governments primarily chose to recruit in the education sector in 2010, hiring 384 individuals, though significant hiring took place in the Commissions of Interior and Finance as well.
As mentioned previously, the recruitment information in this database does not quite cover all of the contractual workers who were selected as teachers. Interviews with officials in the Ngazidja Government suggest that the Education Commission chose to hire over 300 teachers as contractual workers. Meanwhile, the Government of Anjouan is reputed to have hired 500 teachers in the lead up to the election. As such, the data may underestimate the total amount of recruitment. 9

Home islands of civil servants
A common concern in Comoros is to balance the distribution of civil service jobs relative to the populations of the various islands. The perception of unequal public employment by island is sometimes used as a justification for the island governments to further recruit, claiming that either the Union Government or other Island Governments have been recruiting extensively from the population of the other islands. Chart 5.7 points to two interesting facts. First, and surprisingly, the number of Anjouanais who are civil servants, 5,629 to be precise, exceeds Ngazidjans, who number 4,890, a difference of 15%. Second, the presence of Anjouanais in the Ngazidja government, while small, is notable, with 284 out of the 3,023 in total. The difference in the number of civil servants by island of origin is even more striking in proportion to the total population. According to the 2003 Census, Ngazidjans outnumber Anjouanais by 22%. As such, as a proportion of the 2003 population, Anjouanais represent 2.31% while Ngazidjans represent 1.65%. Such a difference is not driven by a large portion of Anjouanais in the Union government, as shown above, but rather the relatively large Government of Anjouan.

Chart 5.8 Island of Origin of Civil Servants, in absolute numbers and as a proportion of population
In sum, the data present a mixed picture. On the one hand, Anjouanais are indeed underrepresented in the Union Government. However, because of the size of the Island Government of Anjouan, and particularly the high level of recruitment since 2008, the number of Anjouanais in the civil service exceeds that of Ngazidjans.

Respect for recruitment standards: the level of education among teachers
A common observation, particularly under systems driven by political clientelism, is that the government recruits unqualified candidates. While a comprehensive evaluation of the merits of various civil servants is difficult, the survey focused on the Education Ministry and Commissions, perusing their qualifications in terms of the level of education. The data do not allow an analysis targeted at teachers specifically, as it includes the staff of the Commissions of Education in the Island Governments of Anjouan and Ngazidja. However, only a small percentage of this staff does not represent teachers in the primary, middle or high schools.

Chart 5.9 Advanced degrees among present-day teachers in Anjouan and Ngazidja, by hiring period
Source: Author database, 2011.
The data present a picture that shows a surprisingly high level of qualification among teachers in terms of their level of education, with over 77% of respondents holding an advanced degree, meaning a degree above the high school level. In fact, many of these respondents are over-qualified if they are primary school teachers, who are required to have only a high school diploma. Interestingly, as Chart 5.9 suggests, this trend has been increasing over time; over 91.46% of the staff of Commissions of Education who were recruited after 2005 had an advanced diploma, a substantial increase from 64.31% of those present-day teachers who were recruited in the period up to 2000, and 72.39% from 2001 to 2005. In other words, despite potential political motives for recruitment, the merit of the recruited teachers seems to be respected.
This data points to an important observation, that even if a system of political clientelism exists -as suggested by the correlation with the 2010 election, it is not conducted such that unqualified staff are hired in the public sector; rather, the rules on qualifications continue to be respected. The reasons for this outcome are explored in a later section.

Chart 5.10 Pedagogical training among present teachers in Anjouan and Ngazidja
Source: Author database, 2011.
The level of pedagogical training among respondents is less than the level of higher education, at 68.22%. Moreover, while the proportion of staff in the Commissions of Education who possess an advanced diploma has increased over time, the proportion who received pedagogical training has dropped over time. 75.78% of teachers who were recruited in 2000 and before have undergone some type of pedagogical training, a level that declined to 70.53% of teachers recruited in 2000 to 2005, and finally 59.07% of those recruited from 2006 to 2011.
These increases in the hiring of teachers with university degrees also coincide with a large increase in the number of university graduates throughout Comoros. While updated official statistics are not available, enrollment data from the national university show a massive increase in the number of university students in Comoros, seen in Chart 5.11. They are also only reflective of students who remain in Comoros, as many Comorians go abroad to study. The lack of economic growth limits the number of skilled jobs available in Comoros, meaning that teaching jobs are often an important employment option for university graduates.

Chart 5.11 Enrollment in the University of Comoros
Source: Data from University of Comoros, 2010 As such, the analysis suggests that teachers generally hold advanced degrees, though fewer have received pedagogical training. If the previously described poor quality of education is related to teacher abilities, the focus of efforts should not be to increase the number of teachers with advanced diplomas, but rather increase the quality of education of teachers. Moreover, if pedagogical training indeed improves classroom performance of teachers, the downward trend should be of concern to policy makers, and suggests that the educational system should provide greater emphasis on this form of teacher-specific training.

Summary of Findings from the Civil Service Survey
This section presented a quantitative analysis of an original database covering nearly 11,000 civil servants. Several clear trends in the data emerge, and are correlated with key events. First, the civil service has experienced a high level of recruitment in the past decade, with the majority of years exceeding a normal growth rate of around 3%. Such levels of spending on the civil service generally even exceed other comparable African countries. Second, the peak years of recruitment correlate with several events, suggestive of certain explanations. 2010, the year of the largest amount of recruitment, was also an electoral year; 2008 was the year the separation of the island of Anjouan ended, when government suddenly recruited a large number of civil servants; and in 2004, the island government of Ngazidja formed, filling numerous key positions. Finally, the data indicate that only a small percentage of teachers who were recruited from 2005-2011 did not have an advanced degree.
The above observations support different hypotheses to explain recruitment. First, patron-client relations are likely active in Comoros, given the large amount of recruitment during elections, but events in other years do not strongly point in this direction. Meanwhile, the dynamics of devolution appear to play an important role in explaining recruitment, given the spikes in the data that are correlated with major events in the islands, such as the creation of the Ngazidha Island Government and the re-integration of Anjouan in the Union Government. Finally, the surprisingly high level of education of teachers suggests that, if political clientelism is prominent, it does not take the form of hiring just any political supporters. Rather, educational standards are respected.

Section 5: A view from the bottom up: Community case studies
The community level component of the research provides an understanding of the government recruitment practices from the ground up. It is based on qualitative research in 15 site visits on all three islands of Comoros, with over 400 individuals interviewed. The 2010 elections that occurred around the time of fieldwork provided a unique opportunity to specifically probe a potential relationship between recruitment and the elections.

A lack of employment alternatives
Interviews with youths in targeted communities suggest that there is a strong demand for public sector jobs, principally due to the lack of viable alternatives for youth in Comoros. The growth failure in the last decade, partly resulting from declining demand for vanilla and ylang ylang, the two main cash crops of Comoros, has exacerbated the problem. The near-stagnant economy is unable to generate enough opportunities for the youth who have enjoyed gradually increasing education levels. A significant part of the demand for public sector employment is fueled by this lack of alternative jobs for individuals with relatively high levels of education.
Overall, there are limited opportunities for enterprise development outside of Moroni, the capital. Economies of scale and the comparatively cheaper cost of imports act as a dampener on many initiatives. Youth who aspire to set up enterprises claim to be constrained by the relatively small size of the market and limited access to credit. In Ngazidja, most of the rural community members -notables, young men and women -are all engaged in farming to varying degrees. However, except in a few relatively large villages such as Dimadjou and Idjikounzi, agriculture is mostly pursued for subsistence purpose and not used as an income generating option. In Anjouan and Moheli, the situation was only marginally better in that there were more farmers, with a larger number of them producing small surpluses that they exchange or sell at the local market. The drastic drop in the vanilla market in the recent past, which was the main cash crop, was repeatedly mentioned in villages both in Ngazidja and Anjouan. The holdings in vanilla had been a key livelihood option and an important source of income in the villages, but the declining price has undermined production. Compared to vanilla, the cloves market showed better performance. Though there had been several lean years in the recent past, still many farmers -particularly in Anjouan -felt that it was still viable to produce cloves for the market.
Faced with these circumstances, the first option of educated youth -both male and female -seems to be to go abroad, either for further studies or for work. 10 Youngsters, particularly those with higher education do not prefer to specialize in agriculture or fishing (expansion of which is also constrained by market and access to credit issues in the current context). Seeking public sector employment comes next in order of preference. Interviews in Ngazidja indicated that an overwhelming portion of those who migrated had gone to mainland France. In Anjouan, the situation seemed to be different, desperate and precarious. The choice destination for economic migration is Mayotte. Over 80% of the families interviewed for the study had one or more of its members or close relatives who had gone to Mayotte. A majority of them had gone there 'illegally' through the dangerous sea routes using fragile boats called 'quasaquasa'. 11 The journey was reported to be perilous -hundreds had perished in the last few years 12 in the sea and thousands had been caught by the French border police, detained and 10 See Da Cruz, Fengler and Schwartzman, 2004. 11 Mayotte is less than three hours by a boat running at average speed from a point in Anjouan and until 1994 the Comorians did not need a visa to go to Mayotte. 12 During the interviews in the four villages we came across more than 10 recent deaths. deported. Some had also been shot at by the French border guards. Despite the perilous nature of the journey, thousands undertake it every year. 13

The role of Notables in social, political and economic space
Notables are the most common traditional leaders in Comoros, being selected not through a hereditary process, but rather via a series of rituals, most notably, the 'grand marriage'. 14 The presence of notables or other traditional authorities was not consistent across the islands. Instead, this authority varied, being strongly prevalent in Ngazidja, somewhat present in Moheli and quite weak in Anjouan. In Ngazidja, the key notables have an influential say in how village resources are utilized. Most of the time, these resources are not from the Government or politicians but from overseas remittances and contributions in 'grand marriages'. These resources are utilized in several ways, such as marriage ceremonies, graduation in the village social structure, access to land and traditional conflict mediation. The group of notables in a particular village represents the key decision making body.
In contrast, the situation in Anjouan was markedly different and apparently less structured. The interviewees found it difficult to explain the social structure in their villages using the framework found in Ngazidja. While the loyalties to family, village and island took on similar forms to that found in the neighboring island, the hierarchies at the village level were not as pronounced or rigid. There was no sense of traditionally sanctioned patron-client relationship to be found in villages in Anjouan. Consequently the 'grand marriage' institution was not a phenomenon there as in Ngazidja.
Though there were notables, their selection and elevation did not hinge exclusively on customary institutions like the grand marriage. Those who wielded authority in these villages derived it either from civil institutions (like the mayor's office) or through popular support. In some cases even notables and chiefs of the villages were selected based on their 'respectability' and 'listening skills' by the others in the village without a 'grand marriage' precondition. As a result, these individuals took the conflict resolution and resource allocation decisions at the village. There appeared to be a greater role for women in the public sphere in Anjouan -in some villages problems were reportedly referred to women's groups to solve before being taken to the civil authorities.
The different social structure in Anjouan -most prominently, the lack of traditional elites such as notables and the relatively smaller value placed on community cultural rituals -meant that there was less common community investment in many villages than in Ngazidja. The pooling together of funds in a systematic and regular basis, as well as the notion of the 'village fund' and institutions around managing such funds, were ad hoc and irregular (for example, raising funds when there is a need to bring electricity to the village, etc.).

Gender
There were clear variations in education levels in terms of gender. While primary enrollment was fairly even between boys and girls, numerous interviewees observed that as they progress through the academic years the tendency is for girls to drop out more than boys. They also reported that the drop-out rates of girls were high both at the completion of primary school and then again during the course of high school. This was particularly acute in small villages that did not have middle or high school degrees as the transport cost was prohibitive. Women also noted protection concerns in sending girls to schools away from village. This disadvantage in terms of education reduced the number of women who attain the qualifications that are required for the various government jobs. The uncertainty around economic returns to education also gave the parents a propensity to discontinue girls' education in the middle. But once graduated, the women are as actively involved in pursuit of a job as males.

The system of exchange: civil service jobs for electoral support
According to interviews in the communities, over the last decade, candidates for President and Governor have utilized political clientelism as a main vote winning strategy, transferring public resources -in this instance jobs -in return for electoral support. In the community visits, our research team could find little or no evidence of government projects or support that was not funded by development partners. Traditional government services in these communities were limited to public education almost exclusively at the primary level, with an occasional health center in only the larger villages. As such, given the limited role of Government in other village development work, jobs represent one of the only state resources that politicians and their representatives have to offer.

If you don't follow politics then you don't get anywhere. -Elders in Anjouan
My advice [for getting a civil service job] would be -get to know the politicians, support the ruling party.

-Newly Appointed Teacher in Ngazidja
The vast majority of interviewees claimed that the system of political clientelism is the principal means for entering the civil service, describing numerous examples of jobs provided due to support during elections in the last ten years. A common observation was that political campaigns were the main period in which civil service job aspirants could hope to be recruited. As such, these aspirants made efforts to connect with politicians, offering to work for the electoral campaigns in the hope that these services would be rewarded with employment. Quite surprisingly, traditional elites played almost no role in connecting the job aspirants with the politicians; instead, these connections were created in an informal way, through various networks, or even based on visits from politicians to the community. Politicians would generally distribute these jobs during the months just before the election, in order to mobilize support for the campaign. However, these jobs were sometimes distributed after the election as a reward for this support, though this post-election clientelism was infrequent and generally focused on a few political positions such as advisors.
Interviewees reported that in the run up to the 2010 elections for President and Governor, which included two rounds taking place on November 7 and December 26, numerous civil service jobs were given out between approximately November 2010 and January 2011. 15 As established in Section 5, teaching jobs are the most common public sector job, which was confirmed in interviews. In the recent past, these job offers were given on a contract basis -job candidates received a basic letter from the Commission asking, for example, to work in a specific village school. But their jobs remain tentative until they receive the formal documentation and a contract letter. Usually this takes at least a few months and in some instances is also contingent on the election results.

Obligations of the client in electoral campaigns
Under the arrangement of political clientelism, both youth and community leaders interviewed in every community suggested that serious most job aspirants engaged in a variety of campaign activities. Interviewees frequently claimed that the job aspirant became informally involved, going door to door in the community, encouraging neighbors to vote for the candidate in the election. At times, the job aspirant worked on the campaign as an office bearer, such as a regional secretary for the candidate, who is responsible for coordinating campaign activities in that area. Others travelled along with the election campaign teams to various parts of the Island canvassing on behalf of candidates, joining vehicles, dressing in a particular color, and taking part in large rallies, speaking on behalf of the candidate. These active campaign roles were more common in Ngazidja, relative to Anjouan and Moheli.
On each of the islands, interviewees described that family members of job aspirants would sometimes get involved in the electoral campaign in order to help their children. As a show of the extent of their support, these parents copied the informal involvement of their children, visiting neighbors and attempting to persuade them to vote for the candidate. However, our team did not find evidence of other forms of involvement, such as parents joining the larger electoral campaign and traveling to neighboring areas. Interestingly, it appeared that female relations, especially mothers and wives, would get more involved in these campaigns than fathers and brothers.
We support candidates in elections. I supported in the Governor's election for the Island as well as in the Presidential elections. I was engaged in campaigning. I did it so that my husband can get a job. Now that my candidate has won I am hoping that my husband can get a teacher job. He already got a contractual job in a high school.
-Woman in Ngazidja Not all job aspirants engage in or offer to do active campaign work, instead simply stating that the individual and his or her family would vote for the candidate. Clearly such a deal would not require much effort on the part of the job aspirant, and it was rarely successful as a strategy for accessing employment. These cases were quite infrequent, and appeared to be more present in Ngazidja.

A profile of individuals who successfully entered the civil service and their communities
Generally speaking, our team found that aspiration for joining the civil service was weak in small villages in Ngazidja, and particularly so in Anjouan, though strong in Moheli. In those villages with weak aspirations, youth generally dismissed access to civil service jobs as an unreasonable expectation. Despite the minimal likelihood, several youths in these villages made attempts to access civil service jobs by working on electoral campaigns, but none had been successful to date. In Moheli, in contrast, the villages in the sample had a number of civil servants and other public sector workers, which prompted many young people to aspire for jobs. While a clear explanation for such a practice did not present itself, a possible reason could be the limited supply of educated youth in less populated villages, or the relatively small number of votes to be won by offering civil service jobs. Interestingly, the level of poverty of the village or town did not appear as a significant explanatory variable to systems of political clientelism The profile of individuals who made an active effort to access public sector employment in Comoros included the following features: predominantly male, comparatively well educated, originally from medium and large villages or towns, having spent some time in Moroni or a regional town, relatively well-off and articulate in comparison with other youth in the village, with a capacity to call on networks who could call on networks, and willing to invest time and effort to get a job. Significant in this profile was the fact that most individuals who would be recruited had successfully completed an advanced degree. In other words, politicians would not recruit just anyone in these communities to work on their campaign.
At the same time, the value placed on educational standards may not simply derive from proper procedures being followed by island governments in terms of recruiting according to merit. Our research team could not pin down the precise reasons for such respect in the communities. In general, though, youth with a higher level of education were more respected within the village; this respect holds a political value, as community members would be more likely to listen to these youths when they urged votes for a particular candidate. The spread of higher education, described above, would mean that there would often be a supporter who both met the educational qualifications for teacher posts and was also from that village, so that the candidate would not have to hire a locally unqualified individual to gain political support in the village. Another potential mechanism could be that the community would dislike the hiring of an unqualified teacher, given that there is an abundance of qualified ones in the country. Similarly, it is possible that civil servants would refuse to accept an unqualified candidate, leading to outspoken protests.
By and large, the youth in the communities observed that those with less educational achievement (less than BAC) did not consider public sector employment as an option. They instead entered a variety of occupations that are required in the community. For men, beyond agriculture, occupations include driver, carpenter, petty shopkeeper, mason, electrician, breeder, middlemen for cloves, and others. For women, such non-agricultural occupations include seamstress, dressmaker and trader. Interestingly our research team came across only very few lower level public sector employees -like office assistants, clerical workers, security personnel and drivers in the ministries -for which individuals with a very low level of education would be considered as qualified.

A New, Transient and Unsuccessful System of Clientelism
Interviewees suggested that the elevation of government jobs as a desired resource, and the patron status conferred on whoever is able to control its distribution, is a relatively recent phenomenon. This transformation was correlated with an improvement in education levels, discussed above. Many individuals in the discussions indicated that in early days, given the increase in public sector jobs and the inability to find qualified personnel, it was easy for someone with the required education to get a position. Interviewees who had joined the various public sector jobs in the 1980s and even 1990s described a situation of a large demand and a low supply of qualified personnel, with no sense that political support was necessary in order to access civil service employment. This changed with the increasing education levels among sections of Comorian population over the last decade or more, and enterprising politicians have found it an important asset in their campaigns.
In education, many teachers observed that, up until recently, the DIFOSI degree -a higher level degree of training for primary school teachers -was a surefire way to get a job. Previously, graduates would be recruited soon after finishing this degree, in an almost automatic practice. In the early 2000s, teachers observed that the recruitment process would take a longer period of time, some having to wait a few years, before getting placed as teachers in the public primary schools. In the past several years, however, such an assurance is no longer true. 16 As I remember in 2000 there wasn't a need to do politics to get a teaching job. It was during Bakar's time that politicization became acute. At that time some teachers were not allowed to teach.
-Teacher in Anjouan Despite this transformation in recruitment practices, the system of political clientelism is also weak, given that the relationships are transient. In other countries, clientelistic relationships are developed over time, with politicians confident that their supporters will provide votes, and in exchange, receive benefits. 17 However, given the constant turnover among executives in Comoros, the voters could not be certain that the politician will follow through in promises of jobs, while the politician could not be certain that voters and supporters will actually follow through with their commitments. The community discussions indicated a growing disillusionment with the uncertainties around getting the job in exchange for political support, which undermines the relationship of political clientelism. Community members perceived a common trend of job promises by the politicians not being fulfilled after the elections, which is causing a greater degree of uncertainty that influences the ex ante investment of time and effort in the political campaigns. Clients often complained of being "betrayed", having received promises of a job that never materialized. As a result, the relationships were temporary and fragile, and the resulting losses among incumbent candidates are not surprising.
Even though we worked on the campaign, no one got a job, because we were betrayed.
-Youth in Moheli The work on electoral campaigns sometimes led to interesting means of diversifying risks among job aspirants. For instance, in Ngazidja, friends supported opposing candidates, and if the supporter of the losing candidate failed to receive a job, while the supporter of the winning candidate did receive a job, then this friend would offer to introduce the supporter of the losing candidate to a representative of the winning candidate. An even more calculated strategy emerged from Anjouan, where one interviewee described an instance in which members of the same family had strategically decided on a division of labor; the family members split their support between two candidates so as to get the benefit for the family irrespective of which of the two candidates wins.
Overall, the frustrations with this system led many individuals in communities to demand greater transparency in the hiring process. Job aspirants were unclear as to what they have to do in order to access civil service employment, as simply attaining advanced degrees is no longer a sufficient condition. They also perceived that many individuals are accessing the job market according to unevenly applied criteria, even if those criteria include working on the political campaign of a politician. More structured recruitment processes, even if not wholly meritocratic, would serve to reduce the increasing resentment of government and its leaders in these regards.

Summary of Community Case Studies
16 Not all jobs are distributed under this system of political clientelism. Interviews suggested that, in fact, there were certain jobs for which the demand was less than what is supplied by the market. Qualified nursing positions and some health department related work were mentioned as example both in Ngazidja and Anjouan (eg.mid wives, health attendants etc). Even in the last few years, the strategy to get one of these jobs was to simply obtain the relevant qualifications and submit applications to the department. 17 See for example, the empirical analyses of Schmidt et al., 1977. The main observation emerging from these case studies is that, in recent times, political clientelism appears as the dominant system of recruitment for the civil service in every case study village across the three islands. In interviews with over 75 teachers and hundreds of other citizens, a consistent story emerged of the executive, whether the president or governor, offering civil service jobs to individuals, mainly educated youth, who worked on political campaigns. Several other observations are noteworthy. First, the main form of patronage offered is teaching positions in the home villages of supporters. There were almost no infrastructure projects in the communities that were not financed from development partners, generally administered outside the main government systems. Second, the primary source of support for campaigns under political clientelistic relationships was not the average youth in the village, but rather relatively educated youth who have studied in the cities. They held other common characteristics, such as male, from a relatively wealthy background, and not from a small village in Ngazidja. Third, the form of support generally consisted of engagement with neighbors, mobilizing them on election day and attending rallies, though some play active organizing roles in the electoral campaigns. At times, the entire family of a job aspirant got involved in the campaign. Fourth, these practices appear to have become common primarily in the last decade; more senior teachers in particular claimed that recruitment during their time focused on simply selecting graduates from the teacher training schools. Finally, these exchanges were generally transitory in nature, having only recently been established, and with both sides often failing to fulfill their so that the exchange. These dynamics may partly explain the fact that incumbent candidates for governor in Ngazidja and Anjouan were not re-elected.

Section 6: Internal and external checks and balances, with a special focus on public school teachers
To scrutinize the institutional process of recruitment in the public service, the research included the public education sector as a case study, followed by a discussion of checks to recruitment external to the island governments. The selection of public school teachers is due to the fact that the education ministry and commissions are the most prominent in terms of staffing and spending. According to statistics developed from the questionnaire, approximately 54% of government staff is dedicated to the public education sector, making it by far the largest employing sector in government, with the vast majority employed in the island governments.

Recruitment of teachers in Education Commissions: weak internal checks and balances
Officials in the Education Commission described two primary means of hiring teachers for the Island Governments: the 'proper' process and the more common process that is especially applied at election time.

Diagram 7.1 Simplified process of Hiring for New Civil Servants
"Proper" Process: More Common Process, Especially During the Electoral Period: Note: at times, these candidates are simply told they are hired, without an official process of approval. Note: The Financial Controller works for the Ministry/ Commission of Finance, while the FoP is an acronym for the Civil Service Ministry/ Commissions, both at the Island level for the hiring of teachers; these are discussed in the next section.
In the 'proper' process, department directors in the Education Commission assess the need of a school for teachers, based primarily on data covering the number of hours of teaching per instructor. They then create posts to reduce the work burden of teachers. According to senior Island Education officials, there is never a shortage of interested applicants. Every year, around the months of July and August, dozens if not hundreds of qualified applicants deliver their dossiers to them. Once a demand for teachers arises in this process, officials determine which applicants meet the qualification requirements for the given level of teacher. The requirements are the following:

Type of School
Education Level for Teachers Primary School BAC (High School) Middle School PEGC 18 / BAC+2 (2 Years University) High School University Source: Interviews with education officials, 2011.
Upon identification of the qualified candidates, the hiring process for creating a post includes approval by the Civil Service Commission and the Controller of Budget in the Finance Commission of the Island Government, with no approval required from an official in the Union Government if the position is created at the island level. 19 Generally, the main purpose of the Civil Service Commission in the hiring process is to assess the appropriateness of the application; to determine whether the candidate meets the minimum qualifications; to assign a place in the salary index; and to process and register the application, creating a new post in the system for the official if one does not exist. Meanwhile, the purpose of the Financial Controller is to ensure that Candidate Hired a proper allocation of the budget is provided for recruitment. The Financial Controller may refuse to process an application based on the potential for excessive spending.
In the second hiring process in Diagram 7.1, which is detailed in Ministère de l'Education Nationale (undated), the same officials are involved, except that politicians participate, making decisions based on electoral politics and pushing recruitment during campaign season. The governor selects the individuals to hire, informing the officials in the Commission of Education, as well as the Financial Controller and officials in Civil Service Commission. Not surprisingly, those administrators responsible for the hiring process admit that they often cannot resist political pressures to hire, leading to a situation where the executive controls the hiring process. Such pressures are strongest during electoral periods, when the governor insists that recruitment take place. While some select officials claimed that the executive would not interfere with their functions, the vast majority of senior officials stated that the political influence can be quite strong. As one high-ranking official put it, "Politics is more powerful than us, the technicians." Another senior FoP official claimed that, "if I reject a request for recruitment [that comes from the governor], I will not have a job the next day." Such observations strongly suggest that the executive often dominates the decisions of the FoP and Budget Controller. This electoral hiring does not stem from even a weak system of data analysis, instead being based on political calculations.
Several observations may be made regarding both processes. Neither of them follows the law, which dictates a competitive recruitment process in which posts are advertised, and candidates are ranked by merit. As such, the government cannot claim to enforce the legally mandated procedure even when it avoids political hiring.
The same few officials who make the request to create the position also select a candidate for that position, even before the need for the position is approved. In the Commissions of Education, only a handful of officials are involved, namely the Human Resource Director and the Director General of Teachers or Director of Primary or Secondary Teachers -as applicablewith the participation of the Secretary General, Commissioner and even the Governor. Such a lack of institutional checks and balances between identifying the need for a position and the candidate clearly weakens the dis-incentives for these officials to hire teachers for their own political ends. 20 The Financial Controller, who determines whether a sufficient budget is allocated to pay for a newly created position, represents another main veto point. The main criteria for their decisions did not in fact represent a clear budget constraint. They generally depended on the documented budget for the given year, and the expenses that had already taken place. However, they did not consider actual revenues in their decision; the annual budgets generally vastly overestimate the actual revenues, meaning that the budgets are rather meaningless in terms of guiding spending.
The hiring practices are not only quite informal, but also vary from island to island. During visits to each director of FoP in the island governments, none of them could produce a manual that directed their decisions and processes of hiring. In terms of the hiring practices, in Moheli, the island government requires that civil servants generally, and teachers in particular, work without pay for between one to two years before they are hired into the civil service. In Anjouan, such a practice of working without pay was common during the time of Bacar, but has subsided since 2008. In Ngazidja, this practice is quite uncommon, with practically all teachers beginning their work with pay, though they are often deployed to schools for several months until their contract is processed.

External constraints: the devolved system of finance
The technical system for devolved government should prevent island governments from large increases in staffing, either through financial constraints or direct control by the Union over hiring. The statistics from 2010 suggest that such a system of external constraints is not successfully applied. Along with the ineffective internal checks to hiring, Union and Island governments maintain a disjointed devolved system of finance and of civil service recruitment. Such a system emerged due to a process of re-centralization that followed several years after the legal structure of devolution was established.
The Constitution of 2001 created a devolved system of finance to support the extensive responsibilities being transferred to the autonomous island governments. The system selected is a fixed quota division of revenue, to be shared by the Union and the Islands, according to the 2005 Act Fixing the Quota Parts of Shared Public Revenues between the Union and the Autonomous Islands.
Beginning in 2005, the system began to function as follows. All government revenue from customs duties, income tax, land tax and VAT passed into an account in the central bank, entitled Shared Public Revenue Account. From this sum, the Central Bank processed deductions for public debt, pensions, services and contributions to international organizations. The Central Bank would then distribute the remaining balance according to the quota division into the accounts of the Union and Island Governments at the Central Bank. 21 Chart 7.3 exhibits the fixed percentages by which the Shared Public Revenue Account would be divided. Moreover, the Constitution affords Island Governments the right to establish and collect their own taxes.

Chart 7.3 Revenue sharing breakdown by government entity
In interviews with officials in both the Island and Union governments, a commonly held view regarding spending was that the revenue gained via the quota system should be utilized by island governments to cover salary expenses, which form the majority of island expenses. Meanwhile, own source revenues, which were comparatively small, would cover operating costs. This view clearly fails to conceptualize any type of remaining revenue that would be used for capital investment, particularly by the island government.
According to the Finance Act of 2011, the following is an example of the breakdown of budgets by entity. While suggestive of spending, according to interviews, the actual spending follows quite different patterns: Chart 7.4 suggests that own revenue sources of the Island Governments are indeed small, representing under 4% of the internal shared revenue total. It moreover covers only a fraction of spending on the salary and expenses, and not even the totality of the goods and services expenses. In other words, Island Governments are almost completely reliant on the shared revenue to cover their expenses. The view that the shared quota would be used for salary expenses, combined with weak own source revenue generating capacity, means that there is little financing available not just for capital investment but also operating costs.

Chart 7.4 FY 2011 Budget lines by government entity
This system of finance changed around 2009, in a way that is indicative of the rather haphazard way in which the devolved system has been designed and maintained. The Minister of Finance, along with the Finance Commissioners of the Islands, agreed that the shared revenues would no longer enter the Central Bank account, from which the funds would be transferred to the Union and Island accounts. Instead, they would enter a "Common Salary" account at the Central Bank. The Ministry of Finance controls this account, with the understanding that, after payment of pensions, public debt and payments to international organizations, the remaining balance would pay for civil servant salaries at both the Union and Island level. 22 This adjusted system remains in place to date in end 2011, despite the fact that the 2005 Act has not been amended. Interviews with key informants are conflicting regarding the motives for these changes. Several union level officials claimed that island governments were misusing funds that should have been used for paying staff salaries for other extraneous expenses, such as missions for the governor. Meanwhile, a few island level officials suggest that the changes came, due to the desire of the central government to control revenues and expenditures, particularly as they can prioritize their own expenditures first.
Three points emerge from the analysis of the evolving financial system of the Union and Island governments. First, the fact that the Union pays for Island Government employees clearly indicates a tendency towards re-centralization of financial systems. Indeed, the Union Government collects nearly all revenue, with the exception of the relatively small amount of ownsource island revenues. Moreover, the Union Government deposits this revenue into an account that its Ministry of Finance then controls. These changes are correlated with the defeat of the separatist movement in Anjouan, and the increased political power of the Union government.
Second, the new system in which the Island Governments maintain the right to recruit employees while the Union Government pays for them represents a disjointed system between centralized and devolved government. In a devolved system, the island government uses its revenues to directly cover its expenses. In a centralized system, the Union Government employs its revenues to cover expenses for its own employees. The Government of Comoros has chosen a mixture of these systems. Several senior island officials expressed suspicion regarding the use of the central Common Salary Account, given that it is controlled by the Ministry of Finance. They believed that it was being used to cover a number of Union expenses, leaving fewer funds available for salary payments.
Third, this disjointed system provides Island Governments with little disincentive to continuously hire new staff, leading to the outcome whereby a centralization of power leads to increased recruitment of civil servants at the local level. Clearly, if the Island Government faces no clear budget constraint with spending on salaries, and the political or even social incentive to hire more employees, whether to provide more education services, provide jobs during elections or others, then it will continuously hire. As one high level official in a Finance Commission stated, "If the Union pays, of course we in the island will recruit."

Section 7: Conclusion
Through its mixed methods approach, this paper identified several trends in the recruitment of civil servants in Comoros. The first empirical section analyzed data for over 10,000 civil servants, discovering that not only did recruitment exceed the normal growth rate for a civil service in the majority of years, but also the peaks in recruitment were correlated with various historical events. The largest recruitment year by far, 2010, was furthermore an election year for the President and island Governors, suggesting political clientelism was at play. 2004 was a significant year of recruitment, driven by the Island Government of Ngazidja, where a chaotic decentralization took place, with ill-defined responsibilities and an unclear plan of transfer of personnel from the Union to the Island Government. 2008 was likewise a peak year, correlated with the end to the separation of the island of Anjouan from the Union, suggesting recruitment related to reconciliation in the island. Interestingly, however, the data showed that teachers, especially those hired recently, in fact met or exceeded the educational standards for their positions.
The second empirical section summarized the results from systematically selected community visits on all three islands. In general, there are few employment options, particularly among the increasingly educated youth. Over the last decade, representatives in the communities universally observed that there has been a rising level of political hiring, particularly during the 2010 elections. Community members described numerous instances where educated youth were in touch with political candidates, agreeing to work on the political campaign and receiving a promise of a teaching or other job in exchange. However, these relationships of political clientelism were not well developed, having been only recently instated, with many promises on both sides that were not fulfilled; these weak relationships may explain the fact that incumbent governors in both Anjouan and Ngazidja lost their re-election campaigns. Interviewees generally confirmed that the profile of candidates successfully recruited into the public service primarily included educated youth, though this may be a result not of effective checks to political hiring, but rather of the fact that their education is respected in the community -which has a political value, or of the value that the community places on having quality teachers.
The third empirical section focused on the internal and external constraints to recruitment. Interviews with high level officials in the Commissions of Education in the Island Governments, as well as the Civil Service Commissions and Financial Controllers of the Commissions of Finance, suggested heavy political influence in the hiring process, regarding when, how much, where and whom to hire. The section also described the lack of external constraints to hiring, whether in the form of veto power by the Union Government or of a firm budget limit. Nearly all revenue for Island Governments should come from a central fund divided by fixed percentages among the Union and Island Governments. Recentralization of the management of finances took place in 2007, meaning that the Union Government absorbed the salary liabilities of the Island Governments, but with no formal mandate to prevent further recruitment. Island Governments therefore faced no immediate budget constraint to increasing the size of their civil service.
Even though the analysis did not statistically test hypotheses for recruitment over the past decade, it did identify evidence to support several propositions. Most notably, there was extensive recruitment in 2004 and 2008. In 2004, the need to form a new Island Government in Ngazidja amidst a chaotic decentralization process and bureaucratic resistance to transfers contributed to the spike in recruitment; in 2008, with the end of the separatist movement, a large recruitment drive took place in Anjouan, with many consultants hired under the Bacar regime being normalized into the civil service. Also, interestingly, civil servants interviewed during the community visits -and especially teachers recruited in the 1980s -claimed that recruitment into government service was not difficult in years past, as a graduate of an institution almost automatically found employment.
This work has strong implications for policy, and especially civil service reform initiatives, but it primarily calls for a process of negotiation between the island and union governments of Comoros. The interests driving political clientelism and the dominance of the executive will undermine efforts to institute internal checks and balances. As such, a creative approach is necessary that takes such contexts into consideration. In the case of Comoros, finding a functioning fiscal framework could present such an opportunity to civil service reform, though it should be done such that they do not exacerbate inter-island relations. Structured negotiations between the executives of the island and union governments are a necessary step towards establishing a sustainable system that will result in not just a rational but politically acceptable system. The fact that the system of political clientelism, where jobs are offered in exchange for political support, is ineffective may make these executives willing to compromise this political strategy for the good of the fiscal health of government. A complementary approach is to avoid focusing attention on all aspects of the civil service, but rather where the efforts are most needed. In Comoros, qualification standards are respected even if political clientelism is dominant in recruitment, suggesting that reforms are not necessary.
Despite the contributions made by this paper to the literature, much more research is needed. Further empirical and theoretical work is required to understand the driving factors behind other states in the developing world in such areas as civil service recruitment, and to move the literature beyond the useful but overly simplistic ideas. Further comparative work within Africa would be interesting in this regard, but also outside, given that the issues identified here are not exclusively found in Africa. Moreover, a greater understanding of how patron-client relationships change over time is required, along with an assessment of the drivers of that change. Finally, while this research has broached the topic of patron-client relations in the context of decentralization, much more work is needed, especially when complemented with political economy work on decentralization.