Health Systems for Outcomes Publication 53125 Discovering the Real World: how Health Workers' Early Work Experience affects their Career Preferences Findings From the Second wave Of a Cohort Study of Young Ethiopian Doctors and Nurses Danila Serra, Pieter Serneels and Magnus Lindelow June 2008 Discovering the Real World ­ How Health Workers' Early Work Experience affects their Career Preferences FINDINGS FROM THE SECOND WAVE OF A COHORT STUDY OF YOUNG ETHIOPIAN DOCTORS AND NURSES Danila Serra Pieter Serneels Magnus Lindelow June 2008 1 ACKNOWLEDGEMENTS We would like to thank Dr Teodros Adhanom, Minister of State, and Dr Shifaraw Teklemariam, Minister of Health for their continuous support and for sharing their reflections throughout the process of this survey. We also would like to thank The Bill and Melinda Gates Foundation, the Government of France, the Government of Norway and The World Bank, for funding the study. We would also like to thank Aklilu Kidanu and his team for careful implementation of the field work, Jose Montalvo and Abigail Barr for invaluable inputs in the research and questionnaire design, and Agnes Soucat and Chris Herbst for their support throughout. Finally, the survey benefited from numerous inputs from attendants at seminars at the University of Oxford, University of California - Berkeley, University of East Anglia, London Business School, The World Bank. This survey would of course not be possible without the support of the health workers we surveyed, who remain anonymous throughout the report. We would like to thank them very much for sharing their information with us and providing us with insights into their professional experiences. 2 EXECUTIVE SUMMARY 1. While Ethiopia has made major progress in scaling up its health infrastructure, the utilization of health services remains relatively low. There is growing agreement that this has to do with the distribution and availability of human resources for health. A scarcity of accurate information on human resources makes it, however, difficult to design evidence based policies to address these issues. 2. This paper summarises the findings from the second wave of a cohort study with health workers. The 90 doctors and 219 nurses were first interviewed in 2004 where they were in their last year at school, and were re-interviewed two years later when they had entered the labor market. The data allows us to analyse (i) where they end up and how health workers are distributed; (ii) what their career preferences are and how they have changed; (iii) what is important in the choice between rural and urban areas; and (iv) what drives the likelihood to migrate abroad. We discuss each of these topics below. 3. There is limited accurate information on the distribution of doctors and nurses in Ethiopia across different dimensions like employment status, urban or rural area, public or private sector. a. We find that almost all medical and nursing students are now working in the health sector, an almost none has left the sector. b. Our data confirm that the geographical imbalance of the distribution of health professionals in Ethiopia, with only 36% of nurses and 17% of doctors working in rural locations. c. Most doctors (74%) work in public hospitals, while nurses work mostly in public hospitals (29%) and public health centres (27%). The vast majority of doctors work as General Practitioners while the nurses work mostly as general nurse. d. One fifth of doctors and a small group of the nurses (5%) have a secondary job in the health sector. The vast majority of them carries out this secondary activity in private for-profit clinics in urban areas while being full time employed in a public sector job. e. One in three health workers hold a primary or a secondary job in the private sector. Of those who require a release from the government to work in the private sector ­ 3 because they have been funded by the government, 58% does not a release from the public sector in order to work in the private sector. 4. To better understand the satisfaction and motivation of health workers we analyse the characteristics of existing jobs and how they match with health workers preferences. We find that a. Doctors in rural areas earn significantly more than those in urban areas, whether they are in the public or the private sector, while nurses earn slightly more in urban areas, but not significantly so. b. Doctors working in an urban public facility receive more training, more frequent formal evaluations, daily checks of presence and more monitoring from clients compared to their rural counterparts, while nurses in rural areas seem to work more hours and have more access to training. c. Work experience has significantly changed the health workers' preferences regarding the importance of different job characteristics. d. Health workers tend to be unsatisfied with most aspects of their job, and especially their salary, their training opportunities and their chances of promotion. In particular, about 80% of the health professionals are either unsatisfied or very unsatisfied with their currently salary. e. Health workers' satisfaction with their career choice, their economic situation and their life in general has gotten worse between 2004 and 2007. 5. How to address the geographical imbalances in the health work force distribution? We analyse the choice between rural and urban posting and find the following. a. There are significant differences in the profile of health workers who work in urban and rural posts with health workers coming from a rural background more likely to work in a rural area. b. The proportion of nurses willing to work in a rural location in the long run has declined from 34% to 18% between 2004 and 2007, whereas the proportion of doctors willing to take a rural post in the long term has increased from 9% to 11%. The latter is possibly explained by the higher earnings for doctors in rural areas. c. The majority of nurses who left their first job was working in a rural area and the most important reason to leave was that they were not happy with the location. Doctors, in contrast have as the most important reason to leave a first position that they are not happy with the salary. d. The minimum salary that makes 80% of the nurses accept a job in a rural area has gone up with 50% from close to 2,000 Birr in 2004 to 3,000 Birr in 2007. The salary that gets 80% of the doctors take up a job in a rural area has gone up with 60% from close 4 to 3,600 Birr in 2004 to 5,500 Birr in 2007. Given the current salaries, to get 80% of the nurses and doctors take up a rural position, we would need a 284% salary increase for nurses and a 245% salary increase for doctors. e. The variation in reservation wages across health workers is partially explained by gender, and by Catholic, which as argued in previous work, reflects the likelihood of having attended a catholic NGO school with a strong reputation for motivating students to help the poor. 6. Migration of health workers abroad is a major concern in Ethiopia, and there is some evidence that indicates this is a justified concern. Using the information we collected on likelihood to migrate abroad, we find that a. More than 50% of the health professionals plan to migrate abroad in the next two years; they expect to earn higher salaries abroad. b. Health workers who are more satisfied with their current job are significantly less likely to migrate abroad. c. Using data from contingent valuation questions that identify the wages for which doctors and nurses would stay in Ethiopia, we find that nurses and doctors require a salary of 6,000 Birr and 10,500 Birr respectively to be persuaded 70% of the nurses and 80% of the doctors not to leave the country. Given the current salaries in the Ethiopian public sector in urban areas, our estimates indicates a 600% salary increase of for the nurses and a 500% salary increase for the doctors. d. Doctors are more inclined to leave the country than the nurses, and would require a higher salary in order to stay in Ethiopia. We do not find any significant impact of gender, age, marital status and other individual characteristics on the reservation wage, although we find that protestant health professionals and those of Tigray ethnicity have a significantly higher reservation wage to stay in Ethiopia. 5 Contents EXECUTIVE SUMMARY 3 1. INTRODUCTION 7 2. BACKGROUND AND METHODOLOGY 8 2.1. The Health Sector and Human Resources for Health in Ethiopia 8 2.2. The Ethiopia Health Worker Cohort Study 12 3. HEALTH WORKERS' ACTIVITIES AND DISTRIBUTION Summary 15 3.1. Overview 16 3.2. Who is working 16 3.3. Working in rural or urban, public or private sector and in which region? 17 3.4. Working in which type of facility and which occupation? 19 3.5. Secondary job in the health sector 20 3.6. The release to work in the private sector 21 4. JOB CHARACTERISTICS, JOB PREFERENCES AND JOB SATISFACTION Summary 22 4.1. Overview 23 4.2. Current salaries and non-monetary benefits 23 4.3. Other important job attributes 27 4.4. Job Satisfaction 30 4.5. Satisfaction with different spheres of life 33 4.6. Attitudes towards work on the side 36 5. THE CHOICE BETWEEN URBAN AND RURAL POST Summary 37 5.1. Overview 38 5.2. The profile of health professionals working in rural and urban areas 38 5.3. The preference for a rural or urban posting 41 5.4. Reasons for quitting the first job 44 5.5. Reservation wages to work in a rural area 47 6. LIKELIHOOD TO MIGRATE ABROAD Summary 52 6.1. Overview 52 6.2. Likelihood to migrate abroad over the coming two years 53 6.3. At what wage would health workers stay in Ethiopia? 54 7. CONCLUSION 59 8. REFERENCES 61 ANNEX A: SURVEY METHODOLOGY 62 ANNEX B: JOB SEARCH 64 ANNEX C: DIFFERENCES IN FACILITY INFRASTRUCTURE 66 ANNEX D: CONTINGENT VALUATION QUESTION FOR RURAL JOB 67 ANNEX E: CONTINGENT VALUATION QUESTION FOR MIGRATION 62 6 1. INTRODUCTION The Ethiopian health sector faces a number of challenges related to human resources, like geographical imbalances, problem with job satisfaction and a high willingness to migrate abroad. In order for the Government of Ethiopia (GOE) to continue its move towards evidence based policies, it was agreed to set up The Ethiopian Health Workers Cohort Study. This study will follow health workers over time in order to better understand health workers career choices, preferences and job satisfaction. The first wave of the study was conducted in April 2007 and involved 219 nursing students and 90 medical students in their final year of study. The second wave of the survey, which took place between May and September 2007, re-interviewed 80% of the nurses and 98% of the doctors when they had entered the labor market. This paper reports the descriptive findings of the second wave as well as changes over time between the two survey rounds. The report is structured as follows. Section 1 provides an overview of the Ethiopian health sector and a brief description of the survey methodology. Section 2 presents the health professionals' current activities including the distribution across locations, sectors and facilities. Section 3 summarizes the findings of the job characteristics like salaries, non-monetary benefits and other characteristics; it also provides information about the job and life satisfaction of health professionals, and it evolution over time. Section 4 reports the results regarding health workers' willingness to work in rural areas, including an analysis of the evolution of reservation wages to work in a rural area, obtained from contingent valuation questions. Section 5, finally, focuses on the health workers' likelihood to migrate abroad in the near future, again using specifically designed questions. Apart from Section 2, each section starts with a summary of the results for that section. 7 2. BACKGROUND AND METHODOLOGY 2.1 The Health Sector and Human Resources for Health in Ethiopia There is wide agreement that human resource development represents one of the fundamental challenges for public health service delivery [see WHO (2000, 2006), and the World Bank (2002)]. Indeed, the availability of trained, qualified and motivated health workers is a necessary condition for effective health service delivery. Ethiopia, which has one the poorest health outcomes worldwide, also faces these challenges, more so than others. Between 1997 and 2005 the Government of Ethiopia designed and implemented the Ethiopian Health Sector Development Program (HSDP) which was aimed at improving all aspects of health care provision in Ethiopia, including human resource development.1 As a result of the program, health expenditure increased considerably over the last decade to reach the 5.6% of government expenditures and 2.7% of total GDP in 2004/2005. However, this still places Ethiopia behind most Sub-Saharan African countries in terms of health expenditure2, and the same holds with respect to per capita expenditure; despite a similar substantial increase from US$2.8 in 1997 to US$3.5 in 2005. Overall, Ethiopia still remains dramatically below the Sub- Saharan Africa average of US$42. A key objective of the Ethiopian HSDP was the development of infrastructures and physical resources. In this respect, the program has been highly successful and has achieved its original targets. The World Bank's "Ethiopia Health Sector Development Program. Implementation Completion and Results Report", released in 2007 registers that the number of health facilities, hospitals and health centres has more than doubled in the last decade; in particular, health posts have increased by 425%. The same report estimates that the HSDP has managed to increase the potential health service coverage from 45-50% in 1997 to 72% in 2004. However, according to the report, in 2004 the number of outpatient visits per person per year was still only 0.3. It is recognized that one of the reasons for the low utilization rate of the newly restored or newly built health facilities is to be found in the limited availability of health personnel. Although one 1 After finishing the first program (HSDP I), a second program with the same name was launched (HSDP II) 2 Malawi, Tanzania, South Africa and Zambia allocated closer to ten percent of their public spending to health, a proportion twice as large as Ethiopia's allocation (World Bank, 2003). 8 of the HSDP policy objectives was the development of human resources3, the increase in the number of health professionals registered in 2004/2005 as compared to 1996/1997 is still low to assure an effective and equitable provision of health care throughout the country. Table 1: Ethiopian health professionals by category 1996/97 2001/02 2004/05 Specialists 271 652 1,067 GP's 1,169 1,236 1,386 Health Officers 30 484 776 Pharmacists 156 118 191 Midwife Nurses 250 862 1509 Other Nurses 2,864 11,976 17,299 Pharmacy Technicians - 793 1428 Laboratory Technicians 621 1,695 2,837 Radiographers 139 247 491 Sanitarians 657 971 1,312 Health Assistants (HA) 10,625 8,149 6,363 Front Line Health Workers (FLHW) - 10,050 11,200 Health Extension Workers (HEW) - - 2,737 TOTAL 16,782 37,233 45,859 Source: Centre for National Health Development (2004) As shown in Table 1, in 2004/2005 there were 46,000 health workers in Ethiopia (against 37,000 in 1996/1997), yet, with a population of about 70 million people, Ethiopia still has one of the lowest health workers to population ratios worldwide. Well trained doctors and nurses constitute only a small percentage of the available health workers. Table 1 shows that in 2004/2005, out of the 46,000 qualified health professionals, only less than 5% were medical doctors, whereas more than 40% - the Health Assistants (HA), Front Line Health Workers (FLHW) and Health Extension Workers (HEW) ­ had received only one year of training [see Centre for National Health Development (2004)]. In addition to the limited availability of qualified and well trained health workers, there are at least three other challenges regarding human resources for health in Ethiopia; they are: (i) the potentially low satisfaction and motivation of health workers, (ii) the geographical imbalances in the distribution of health workers; and (iii) the high likelihood for health workers' to migrate abroad. 3 "Expanding the Supply and Productivity of Health Personnel" was allocated 3 percent of total base costs in the original Project development Objectives, against 27.5 per cent allocated to "Facility Expansion and Rehabilitation", and 50 per cent to "Improvements in the Technical Quality of Primary Health Care Service Provision". See World Bank (2007). 9 Satisfaction and Motivation Explorative qualitative research suggests substantial frustrations among health workers, affecting their job satisfaction and work motivation (see Lindelow and Serneels 2006). Although these findings have never been quantified, the existing findings suggests that there may be two groups of health workers: those with relatively high levels and those with relatively low levels of dissatisfaction. In Section 3 we discuss job satisfaction among the young health workers in our survey. Geographical Imbalances Health expenditures in Ethiopia have historically been distorted in favour of urban areas, and health facilities in Addis Ababa have traditionally received the largest share of public funds.4 The better quality of health and non-health infrastructures in urban areas has resulted in an uneven distribution of health workers between urban and rural areas. Highly trained professionals prefer to work in urban areas, which imply that the rural areas, where more than 80 per cent of the Ethiopian population lives, have a relative shortage of health professionals, especially of specific groups like surgeons and midwives (see Centre for National Health Development, 2004). Results from the previous round of this survey indicate that future health workers who come from a rural area themselves, are more motivated and come from less well-off households are more likely to be willing to work in a rural area (Serneels, Lindelow, Montalvo, Barr 2007). Section 2 reports the actual distribution of health workers across rural and urban areas, regions etc.. In Section 4 we analyse the analyse the choice between a rural and urban post. Migration abroad of Health Workers A third factor is the emigration of health professionals to foreign countries. Although the proportion of health workers who leave the country does not reach the extreme levels registered in other African countries (for example Mozambique, Angola or Malawi), it has steadily increased in the last few years. Clemens and Pettersson (2008) provide information on the proportion of Ethiopian (and other countries') health workers who emigrated abroad in 2000, using destination-country census and information on the in-country "stock" of health workers in Ethiopia in 2000. Table 2 shows the number and proportion of Ethiopian health professionals who left the country in 2000. 4 This is also a consequence of the historically (especially until 1993) centralized approach to health planning, resulting in centralized health care delivery more than rural, community-based health service provision. 10 Table 2: Doctors and nurses abroad by country of destination (in 2000) DOCTORS NURSES Domestic "stock" 1,310 5,342 Emigrated to: UK 65 61 USA 420 888 France 16 16 Canada 30 75 Australia 9 37 Portugal 1 0 Spain 1 0 Belgium 2 0 South Africa 9 0 Total abroad 553 1,077 Proportion of migrants 30% 17% Source: M. Clemens and G. Pettersson (2008). About 30 per cent of the doctors and 17 per cent of the nurses did emigrate abroad in 2000. Combining these figures with the figures for 2001 reported in Table 1, this suggests that about one fifth of the doctors trained in Ethiopia and still working in the health sector is working abroad.5 The results from the first wave of the cohort study suggests that this may have increased further, since 72% of the final year medical students and 62% of the final year nursing students stated they planned to go abroad within five years. In Section 5 we discuss the likelihood to migrate abroad using the data from the second wave. Another topic related to human resources that is often discussed is the growth of the private sector and the potential effect this has on health workers leaving the public sector and taking up a private job. The recent report on the HSDP implementation and results registers a 122 per cent increase in the number of private health facilities between 1997 and 2004, and documents a decline in the proportion of health professionals, especially highly skilled practitioners, employed in the public health sector from 73% in 2001/2002 to 44% in 2004/2005 (see World Bank, 2007). The change from the public to the private sector is mostly driven by higher salaries. Given that private facilities are located in urban areas, the development of the private health sector may reinforce the disparities in the distribution of health workers between urban and rural areas. On the other hand, the development of the private sector may draw more people into the health profession, and may also free up public funds. The overall effect is therefore 5 This reflects the number of emigrated doctors as a percentage of the number of Specialists and GPs in Ethiopia (652+1,236+ 553/553). 11 difficult to anticipate. In Section 3 we briefly discuss the distribution of health workers between the public and private sector. Otherwise, the current study has little to say about this topic. Until recently, the Government followed a manpower planning approach to manage human resources for health. This meant that the allocation of a health worker to a position, as well as the level of salary was heavily regulated. Until 2006, health professionals who received free public education ­ the great majority­ had to serve in the public sector for a certain number of years (usually twice the duration of their education) before receiving their license and a release certificate which would allow them to practice in the private sector, or to leave the country (if they wished so and had a visa). Additionally, health workers were allocated to their first position (in the public sector) through a public lottery, with the intention of making the allocation, especially to rural positions, impartially while at the same time providing equality of opportunity. In recent years the enforcement of the lottery and the mandatory nature of public service have progressively weakened. Following formal accusations from the government funded students that the compulsory public service was illegal since they had not been informed or required to sign a contract when they started their degree, the Ministry of Health started to issue licenses to health workers immediately after graduation (instead of withholding them until they had fulfilled their compulsory public service). Consequently, the lottery system and the mandatory allocation of health workers to public health posts have been weakened.6 Although health workers still have to serve for a number of years in the public sector in order to get a release to work in the private sector, many health workers, especially doctors, work in the private sector without release. Being in possession of the license and official degree documentation also makes it easier for health workers to leave the country. 2.2 The Ethiopia Health Worker Cohort Study The key motivation behind the Ethiopian Health Worker Cohort Study is to assist the Government of Ethiopia in its movement towards evidence based policies. The study provides unique empirical evidence on health worker career preferences, satisfaction and motivation, with a special emphasis on the choice between rural and urban posting, and the choice to 6 Currently, lottery sessions are still conducted twice a week starting in October each year, but the allocations through the lottery are no longer binding since the health workers can get their license as soon as they complete their degree and the Ministry of Health has little means to enforce the lottery results. 12 migrate abroad. This will provide a basis to design policies and interventions that effectively improve professional development, geographical distribution and retention of health workers in Ethiopia. The survey exists of two waves, which we briefly discuss below. First wave of the study: final year health students in 2004 The first wave of the study took place in April and May 2004 and conducted a survey of 219 final year nursing students and 90 final year medical students (the year before their internship), following a qualitative pre-research study administered in 2003. The 209 nursing students represent 16% of the cohort of nursing students attending their final year in 2003/2004, while the 90 medical students represent 49% of the cohort of medical students for that school year. The average age of the students in the 2004 original sample was 22 for the nursing and 23 for the medical students. The proportion of females was low (14%) among the medical students whereas it was above 50% among the nursing students. Few nursing students (10%) and even fewer (4%) medical students were married in 2004, but 35% of nurses and 25% of medical students were in a non-married relationship at the time of the survey. All Ethiopian ethnic origins were represented in our original sample.7 The first wave of the survey identified important differences among future health workers in terms of career aspirations and motivation. For instance, although the majority of the students preferred to work in an urban post, there was a significant minority who prefered to work in rural post. Moreover, although students' preferences in terms of job location and private vs. public sector appeared to be highly responsive to financial incentives, other non-monetary job attributes, such as the risk of professional isolation, access to further training, and availability of good education for children, also seemed to play a significant role.8 Second wave of the study: starting health workers in 2007 The Second wave of the Cohort Study revisited the same individuals in May - September 2007. As the health professionals have now taken up their first post, the data allows to investigate the actual level of job satisfaction as well as the career choices of our cohort, as well as how their preferences - like urban vs. rural posting­ have changed since having built experience in the sector. 7 See the fist wave descriptive report for further details on the demographics of our cohort. 8 For detailed analysis, see Serneels, Lindelow, Montalvo and Barr (2005) 13 Tracing back the individuals requested tremendous effort, and the technical details are described in Annex. We traced back 80% of the nurses and 98% of the medical doctors, giving rise to panel data for 177 nurses and 88 doctors (see Annex A for details on the follow-up process). For 4% of the nurses and 2% of the doctors in the original sample, we found out from their family that had migrated abroad. On the remaining 16% of the nurses we have no further information. To get a first idea of how different the missing nurses are, we compare the descriptive statistics. Table 3 reports the most important demographic characteristics for original sample, the final sample, the migrants, and those who are missing. Not surprisingly, the health workers who migrated abroad were all unmarried and with no children in 2004, although 50% of the nurses were in a relationship at that time. The nurses who left the country were on average older than the ones who stayed, and were mostly female (77% against 33% of males). Table 3: Characteristics of the sample, the "missing" and the migrants DOCTORS NURSES Sample Sample Missing Sample Sample Missing Migrants Migrants in 2004 in 2007 in 2007 in 2004 in 2007 in 2007 Age in 2007 (yrs) 26.3 26.3 26 25.3 25.4 26.3 24.9 Male 86% 86% 50% 49% 45% 33% 67% Married in 2004 4% 4% 0% 10% 12% 0% 0% Married in 2007 17% 11% With partner in 2004 25% 26% 0% 30% 32% 50% 17% With children in 2004 3% 3% 0% 10% 12% 0% 5% With children in 2007 3% 24% Total (Obs.) 90 88 2 0 219 177 9 30 Note: two nurses passed away and one nurse has joined a monastery. The missing health workers are more likely to be men and less likely to be married or have a partner or children, compared to the re-interviewed health workers, which may also suggest that they are more likely to have left the country. 14 3. HEALTH WORKERS' ACTIVITIES AND DISTRIBUTION Summary 1. There is little accurate information available on the distribution of doctors and nurses in Ethiopia across different dimensions like employment status, urban or rural area, public or private sector, region, etc. The random sample selection at the initial stage of our survey enables us to provide a representative picture of where starting health professionals end up. 2. About 95% of doctors and 82% of nurses are currently working in the health sector. 18% of nurses are enrolled in further education, while 4% of doctors and 2% of nurses have recently quit their job or could not find a job they liked. 3. Our data confirm that the geographical imbalance of the distribution of health professionals in Ethiopia, with only 36% of nurses and 17% of doctors working in rural locations. The majority of doctors (73%) and nurses (49%) work in the public sector in an urban area. Apart from Addis Ababa, the most centrally located regions like Amhara, Oromiya, and SNNPR attract most health workers. 4. Looking at the type of facility, most doctors (74%) work in public hospitals, while nurses work mostly in public hospitals (29%) and public health centres (27%). The vast majority of doctors work as General Practitioners while the nurses work mostly as general nurse. 5. In total 21% of the doctors and 5% of the nurses have a secondary job in the health sector. The vast majority of them (78%) carries out this secondary activity in private for- profit clinics in urban areas while being full time employed in a public sector job. Both doctors and nurses spend on average above 3 days per week in their secondary activity. The main reason for taking up a second job is to increase earnings. 6. In total 29% of health workers hold a primary or a secondary job in the private sector. Of these, 42% have not been funded by the federal or the regional government and therefore do not require a release from the government in order to work in the private sector. The remaining 58% is officially required to have a release but does not, and is strictly spoken working illegally in the private sector. This is most common among health workers who combine a primary activity in the public sector with a secondary activity in the private sector. 15 3.1. Overview There is little accurate information available on the distribution of doctors and nurses in Ethiopia across different dimensions like type of job, region, sector, etc. Existing figures are often outdated or inaccurate. Even less is known about where young health professional end up. The random sample selection in the initial stage of our survey enables us to provide a representative picture of where starting health professional end up.9 This Section describes the distribution of health workers across employment status, rural and urban areas, regions, type of facility and type of occupation. It also discusses who has a second job and who obtained a release to work in the private sector. 3.2. Who is working? 10 A unique feature about our data is that it tells us what percentage of students in health actually go on to work in the health sector. The figures in Table 4 show that 95% of the doctors and 82% of the nurses we traced currently work in the health sector in Ethiopia, while 1% of the nurses work outside the health sector.11 Among the doctors, 5% is not working because they recently quit their job (3%) or could not find a job they liked. For nurses, enrolment in further education is the most common reason for not working (15% of the nurses and 81% of those not employed), while 2% is on maternity leave or sick leave and a remaining 2% recently quit their job or could not find a job they liked. A more detailed analysis reveals that, excluding those on maternity leave, all health workers who are out of work have been in this situation for a relatively short period of time, i.e. less than 4 months.12 9 For medical students we sampled all schools, while for nurses we focused on schools in Oromiya, and SNNPR 10 Annex B contains more information on job search and job allocation. 11 Two nurses, both females, work primarily outside the health sector, , one owns a private school and the other is an employee of a private organization. Their earnings are low and it is unclear why they choose this work. The health worker who owns a school earns a monthly salary of 5000 Birr while the nurse who is employed in a private organization ­ is earning a salary of 645 Birr, which is even below the lowest salary earned by the other nurses in our sample. 12 Only one female nurse is not working because she "could not find a job she liked" and she has been in this situation for a long period of time, i.e. 30 months. 16 Table 4: Primary activities of the health workers DOCTORS NURSES Total Currently working 84 145 229 (95%) (82%) (86%) in the health sector 84 143 227 (95%) (81%) (86%) outside health sector 0 2 2 (0%) (1%) (1%) Not currently working 4 32 36 (5%) (18%) (14%) in education 26 26 (15%) (10%) on maternity leave 2 2 (1%) (1%) on sick leave 1 1 (1%) (0.4%) recently quit the job 3 2 5 (3%) (1%) (2%) Could not find a job the liked 1 1 2 (1%) (1%) (1%) 3.3. Working in a rural or urban, in the public or private sector, and in which region? What does the distribution across location and sector look like? Only 29% of the health workers who are currently employed in the health sector are working in a rural area, as shown in Table 5. The figure is substantially lower for doctors (17%) than for nurses (36%). The distribution of health workers across the public and private sector ­ where the private sector includes profit and non profit health facilities ­ shows that 73% of the nurses and 82% of the doctors work in the public sector. Table 5: Location and sector DOCTORS NURSES TOTAL Rural area 17% 36% 29% (14) (51) (65) Urban area 83% 64% 71% (70) (88) (162) Public sector 82% 73% 77% (69) (105) (174) Private sector 18% 27% 23% (15) (38) (53) Total 84 143 227 17 Since the private sector is less present in rural areas, it is interesting to consider the distribution across sectors and location simultaneously. Figure 1 shows that a majority of doctors (73%) and nearly half of the nurses work in the public sector in urban areas, whereas only 10% of doctors and 25% of nurses work in the public sector in rural areas. The private sector in urban areas employs an additional 12% of doctors and 18% of nurses work. The remaining 7% of doctors and 10% of nurses work in the private sector in rural areas. The figures underline the difficulties to get health workers in rural areas, be it in the public or private (usually not-for- profit) sector. Figure 1: Sector and location 80% 70% 60% 50% 40% 30% 20% 10% 0% public sector in public sector in private sector in private sector in rural area urban area rural area urban area DOCTORS NURSES In addition to rural-urban imbalances in the distribution of health workers, there also seem to be significant regional imbalances. Even though this distribution is affected by our initial sampling, which did not sample any nursing schools in Tigray, our sample was randomly drawn from other nursing schools and all three medical faculties. Figure 2 shows that about 26% of the doctors and more than 20% of the nurses are currently working in Addis Ababa. An additional 18% of doctors and 34% of nurses are located in the Southern Region, followed by more than 20% of doctors and nurses in the Oromiya region. The Amhara region attracts about 20% of the doctors and 7% of nurses, whereas Tigray attracts more doctors (8%) than nurses (2%) and Harari attracts more nurses (7%) than doctors (1%). The regions of Dire Dawa, Benishangul and Gambela receive only a small proportion of health professionals, and the Somali and Afar regions have attracted none of our health workers.13 13 Although it is possible that some of our missing health workers are employed in these regions. 18 Figure 2: Regional distribution snnpr addis ababa oromiya amhara tigray harari dire dawa benishangul gumuze gambela 0 5 10 15 20 25 30 35 40 % DOCTORS NURSES 3.4. Working in which type of facility and which occupation? Looking at the type of facility, we see that doctors work especially in a public hospital (74%), at the university (7%), a clinic (7%), and in a private for-profit hospital (5%), while nurses work especially in public hospital (29%), a public health centre (27%), a health bureau (9%), a private not-for-profit hospital (9%) or a not-for-profit clinic (8%). Figure 3: Type of facility public hospital public health centre private (not ngo) clinic university private (not ngo) hospital private ngo hospital private ngo clinic health bureau ngo public health post government health science college public health station public pharmacy 0 10 20 30 40 50 60 70 80 % DOCTORS NURSES 19 Regarding occupations, about 85% of the doctors are General Practitioners and 87% of the nurses are general nurses, as shown in Table 6. Only 5% of the doctors hold a specialization, and 11% of the doctors and 2% of the nurses are university lecturers. Table 6: Occupation in the health sector DOCTORS NURSES Specialist Medical Doctor 5% General Practitioner 84% Lecturer 11% 2% Head Nurse 4% Head of Health Bureau 2% Nurse 87% Medical Supervisor 3% Mothers & Children Health Care Expert 1% Coordinator 1% General Service 1% 3.5. Secondary job in the health sector A total of 19 doctors, or 21% of the doctors, and 8 nurses, or 5% of the nurses, have a secondary job in the health sector. The vast majority of them (78%) carries out this secondary activity in private for-profit clinics in urban areas. A small minority of doctors and one nurse are employed as lecturers in private colleges. Table 7: Secondary job in the health sector DOCTORS NURSES TOTAL Job in a public hospital 1 1 Job in a private facility 17 6 23 Job in a private college 2 1 3 Avg. days per week in secondary job 3.5 4 3.6 Avg. hours per day in primary job 8 9.9 8.8 Avg. hours per day in secondary job 8 9 8 Total net salary in primary job (avg.) 1971 870 1632 Total net salary in secondary job 1205 507 990 (avg.) 20 About 83% of the doctors engaged in a secondary activity have a full-time primary job in a public hospital. Only one doctor combines a part-time job in the public sector with a part-time job in the private sector. Turning to the nurses, 87% of those who have a second job hold a full- time job in the public sector, while 13% holds a full-time post in a private NGO clinic. The combination of full-time public job and part-time private job thus characterizes 85% of the health workers who carry out secondary activities in the health sector and both doctors and nurses spend on average above 3 days per week in their secondary activity. The main reason for taking up a second job is to increase earnings. 14 3.6. The release to work in the private sector In total 29% of health workers (73 in number) hold a primary (20%) or a secondary (9%) job in the private sector. Of these, 42% have not been funded by the federal or the regional government and therefore do not require a release from the government in order to work in the private sector. The remaining 58%, however, is officially required to have a release and are strictly spoken working illegally in the private sector. Working in the private sector without a release certificate is especially common among health workers who combine a primary activity in the public sector with a secondary activity in the private sector. Table 8: The release to work in the private sector Primary activity Secondary activity in the private sector in the private sector Government Government did Government Government Total paid for not pay for paid for did not pay education education education for education Release 12 1 13 55% 5% 18% No Release 10 31 21 1 60 45% 100% 95% 100% 82% 14 Five health professionals (4 nurses and 1 doctor) carry out a secondary activity outside the health sector. Of these, two are employed in public enterprises, one works in a private organization, one is a mechanic and one is employed in trade. The earnings from these activities are generally low: around 500 Birr per month. 21 4. JOB CHARACTERISTICS, JOB PREFERENCES AND SATISFACTION Summary 1. Doctors in rural areas earn significantly more than those in urban areas, whether they are in the public or the private sector, while nurses earn slightly more in urban areas, but not significantly so. In all cases, salaries in rural areas are below 2004 reservations to get 80% of the work force in rural areas. Public sector doctors working in a rural area receive significantly higher location allowance than their urban colleagues, while private sector doctors in rural areas receive more housing allowance than their urban counterparts. Public sector nurses working in rural areas receive significantly higher housing and occupational allowances than their urban colleagues. 2. Doctors working in an urban public facility receive more training, more frequent formal evaluations, daily checks of presence and more monitoring from clients compared to their rural counterparts, while nurses in rural areas seem to work more hours and have more access to training. Doctors and nurses in the private sector have more access to promotion, independent from the location, while those in the public sector have more access to training. 3. Work experience has significantly changed the health workers' preferences regarding the importance of different job characteristics. A significantly larger proportion of health professionals now considers their salary the most important job attribute compared to in 2004, while the physical conditions of the workplace also climbs in rank. About 60% of the health workers who in 2004 considered "the opportunity to help the poor" the most important attribute of a health post have changed their preference in 2007. Those who changed preferences are more likely to be females, married with children, and are more likely to now give priority to earning a good salary. 4. Health workers tend to be unsatisfied with most aspects of their job, and especially their salary, their training opportunities and their chances of promotion. In particular, about 80% of the health professionals are either unsatisfied or very unsatisfied with their currently salary. We do not find differences in the level of satisfaction with any job characteristic based on demographic characteristics. Doctors and nurses are equally (un- )satisfied with all aspects of their current job. Satisfaction with the salary seems to positively 22 depend on the current salary and negatively on the salary expectations the health workers held when they were students. Working in an urban facility increases the level of satisfaction with salary, promotion and training opportunities whereas working in the private sector in an urban location decreases the level of satisfaction with promotion opportunities. 5. Health workers' satisfaction with their career choice, their economic situation and their life in general has gotten worse between 2004 and 2007. About 80% of the health workers are either unsatisfied (20%) or completely unsatisfied (60%) with their economic situation, compared to less than 10% in 2004. About than 40% of the health workers are unsatisfied or completely unsatisfied with their life, compared to less than 20% in 2004. The doctors are more likely to be unsatisfied with all spheres of life than the nurses. The level of satisfaction with the different aspects of the current job affects satisfaction with economic situation, job and life in general, and the current wage has a positive impact on the level of satisfaction with all the spheres of life. 4.1. Overview In this section we look at both the most important job characteristics, the preferences of health workers and their satisfaction. We also look at satisfaction with different spheres of life, and attitudes towards working on the side. 4.2. Current salaries and non-pecuniary benefits In this section we look at the total earnings of doctors and nurses, and compare them across four categories of posts: public sector in a rural area, public sector in an urban area, private sector in a rural area and private sector in an urban area. In a first section we focus on the net salary, inclusive of allowances15, while a second section looks at non-pecuniary benefits. 4.2.1 Salaries The figures in Table 9 report the salaries for doctors and nurses in each of the four sectors. The figures indicate that doctors earn more in rural than in urban areas, both in the public and the private sector. In fact, working in a rural area increases a doctor's salary by a third on average. 15 We asked the health workers to give us information on the allowances they receive, if any, for: food, transport, housing, clothing, professional hazard, location in rural or remote areas, etc. 23 The average monthly wage, inclusive of allowances, earned by a doctor in a rural location is 3,383 Birr (369 USD) while that in an urban location is 2,228 Birr (240 USD). 16 Nurses, however, do not earn more in rural versus urban areas. The average monthly wage, inclusive of allowances, earned by a nurse in a rural location is 827 Birr (90 USD) whereas the average wage in an urban location is 903 Birr (99 USD). Whatever the location, working in the private sector provides significantly higher earnings than working in the public sector, both for doctors and nurses. A doctor employed in a public health facility earns on average 2050 Birr (224 USD) per month, or almost one-third of the salary in the private sector, which is 5545 Birr (608 USD). Similarly, a nurse working in the public sector earns on average 724 Birr (80 USD) per month, or nearly half the salary she would make in the private sector, which is 1280 Birr (141 USD). 17 Interesting is also that the highest doctor salary is found in a private facility in a rural area, whereas the highest nursing salary is in a private facility in an urban area, as reported in the last row of Table 9. Table 9: Total net salary by sector and location DOCTORS NURSES Public Private Private Public Private Public Public Private sector sector sector in sector in sector in sector in sector in sector in in urban in rural urban urban urban rural area rural area rural area area area area area area TOTAL NET 2494 1991 5619 3828 705 748 1126 1353 SALARY (ET Birr) p-values for equality p=0.005 p=0.03 p=0.37 p=0.21 of means test Minimum salary 1810 1206 3500 2690 576 582 375 700 Maximum salary 3165 3200 8000 7500 1181 2300 3163 5114 It is also interesting to contrast these earnings with the expectations of the health workers two years earlier, when they were still at school. In the first wave of the survey in 2004, we measured health workers "reservation wages" to work in a rural area Adjusting this 2004 figure for inflation, we find that a salary of 4765 Birr (in 2007 prices) would bring 80% of the doctors in rural areas 200km from Addis Ababa.18 Table 9 shows that the salary that a doctor earns 16 Using the current Ethiopian Birr-US Dollar exchange rate of 0.11 US dollars for 1 Ethiopian Birr. The exchange rate in the summer of 2007, when the survey was conducted was very similar, i.e. 0.1129 US dollars for 1 Ethiopian Birr.. 17 Mean equality tests of wages earned in the private and public sector by urban and rural locations give p-values of 0.0000 (rural area) and 0.0006 (urban area) for the doctors and p-values of 0.006 (rural areas) and 0.0000 (urban areas) for the nurses. 18 We asked health students at what salary they would be willing to switch from a job in Addis Ababa to a similar job in a rural area using a contingent valuation methodology (See Serneels et al. (2006) for details on methodology and results). We find that a wage of 2,562 Birr would induce 80% of the doctors to rural areas 200 Km from Addis 24 when working in the public sector in a rural location is considerably below ­ almost half of ­ the 2004 reservation wage to work in a rural area, whereas the salary that a doctor would earn in the private sector in a rural area is above the 2004 reservation wage. However, the private sector in rural areas attracts a very small fraction of doctors (7%), suggesting a shortage of positions in the private sector in rural areas. 19 For nurses, however, the average wage earned in rural areas is considerably below the nurses' 2004 reservation wage to work in a rural area, which is 2,046 Birr (in 2007 prices).20 The average salary for a nurse in a public rural facility is 705 Birr, or only 35% of the 2004 reservation wage, while the private sector rural salary is 1,126 Birr or 55% of the 2004 reservation wage. This suggests that the low payment is a major contributing factor to the low amount of nurses employed in rural areas. An interesting exercise is to analyze the different components of health workers' earnings, distinguishing between salary and allowances. Table 10 indicates that public sector doctors working in a rural area receive significantly higher location allowance than their urban colleagues, while private sector doctors in rural areas receive more housing allowance than their urban counterparts. Public sector nurses working in rural areas receive significantly higher housing and occupational allowances than their urban colleagues. 21 Table 10: Salaries and most important allowances by sector and location DOCTORS NURSES Public Public Private Public Private Private Public Private sector in sector in sector in sector in sector in sector in sector in sector in urban urban urban rural area rural area urban area rural area rural area area area area Obs. 8 61 6 9 37 67 14 25 Net salary 1332 1423 4967 3792 678 708 1051 1249 p-values p = 0.23 p= 0.09 p = 0.20 p = 0.24 Location 912 279 0 0 0 4 17.5 0 allowance Ababa. We then correct this figure for inflation by computing the change in the Ethiopian Retail Price Index between April 2004 (when the first wave survey was conducted) and June 2007 (when most of the second wave data were collected). Over this period, prices rose with 86%. 19 Since the definition of `rural' and `urban' locations depend on the respondent and enumerator's evaluation, this comparison is imperfect, but nevertheless reveals a discrepancy. 20 The 2004 reservation wage was equal to 1,100 Birr. 21 The totals in Table 10 do not correspond exactly to those in Table 9 because Table 10 only looks at the three most important allowances, namely location, housing and occupational allowances. A very small number of health workers (less than 5%) also receives food allowances, professional hazard allowance and transport allowances. 25 p-values p = 0.0001 p=/ p = 0.23 p = 0.09 Housing 187 199 75 0 19 4 18 48 allowance p-values p = 0.43 p = 0.06 p = 0.008 p = 0.32 Occupational 62 71 210 0 15 3 4 30 allowance p-values p = 0.44 p = 0.13 p= 0.06 p = 0.19 4.2.2. Non-pecuniary benefits Analysis from the first wave indicate that non-pecuniary benefits also play an important role in health workers' career choices and may for example explain why most doctors and nurses are employed in the public urban sector, as seen in Figure 1 in Section2, and which corresponds with the dominant preferences we observed in wave 1. Table 11 summarises the differences in potentially important non-pecuniary benefits associated with the different types of jobs, like the average hours of work per day, access to training, chances of promotion and free access to health care. We also look at three monitoring variables: whether there are daily checks on the health workers' presence, whether there is regular evaluation of the health worker, and the presence of complaints offices for clients to use. With the exception of the first row, which reports average hours worked, all the other rows in Table 11 report the percentages of health workers who responded affirmatively to the corresponding question. "Training" refers for example to the proportion of health workers who have received training while in their first job, while "promotion" refers to the percentage of health workers who received a promotion while in their current job, etc. The figures in Table 11 reveal that doctors working in an urban public facility receive more training and also receive more frequent formal evaluations, daily checks of presence and monitoring from clients through complaints offices than their rural counterparts. Nurses in rural areas seem to work more hours and have more access to training. A comparison between doctors and nurses in the private and the public sector, independent from the location, indicates that while the public sector gives more access to training, the private sector offers higher chances of promotion.22 22 The p-values from the corresponding proportion equality tests are equal to 0.001 for both training and promotions. It should be noted that in comparing the proportion of health workers who received training and promotions while in their current job, we are not controlling for the amount of time they have been employed in the job. The differences in training and promotions between sectors may among others reflect different rates of job turnover. 26 Table 11: Other job characteristics by sector and location DOCTORS NURSES Public Public Private Private Public Public Private Private sector in sector in sector in sector in sector in sector in sector in sector in rural area urban area rural area urban area rural area urban area rural area urban area Hours 9.5 9.01 10.1 9.3 8.9 8.4 9.7 8.9 worked (avg) p-values p = 0.34 p = 0.23 p = 0.05 p =0.12 Training 62% 66% 17% 22% 70% 55% 21% 48% p-values p = 0.43 p = 0.40 p = 0.07 p = 0.05 Promotion 12.5% 13% 17% 44% 16% 15% 29% 44% p-values p = 0.48 p =0.13 p = 0.43 p = 0.17 Free access to 88% 57% 33% 66% 67% 67% 100% 76% health care p-values p = 0.05 p = 0.10 p = 0.48 p = 0.02 Daily checks 50% 82% 50% 56% 97% 100% 71% 96% of presence p-values p = 0.02 p = 0.41 p = 0.08 p = 0.01 Evaluation on a regular 0% 10% 50% 33% 49% 39% 71% 68% basis p-values p = 0.18 p = 0.26 p =0.16 p = 0.41 Complaints 50% 54% 17% 33% 27% 61% 14% 36% offices p-values p = 0.41 p = 0.24 p = 0.000 p = 07 4.3. Other important job attributes In 2004 we asked the final year medical and nursing students in our sample to rank eight different job characteristics according to their importance to them. About 35% of the medical students and 30% of the nursing students ranked "having access to further training" as the most important job characteristic. "Earning a good salary" was ranked first by 25% of the medical students and 6% of the nursing students. Finally, about 30% of the nursing students and 15% of the medical students ranked the "opportunity to help the poor" as the most important job characteristic. The physical condition of the workplace did not seem to be an important job characteristic for most students. Results from the 2004 survey are shown in Figure 4.23 23 Besides salary, training opportunities, condition of the workplace and opportunity to help the poor, we also asked the respondents to rank the importance of job stability, good colleagues and stress at work. Since the proportion of health workers who considered these job characteristics relatively important is negligible, we do not report the corresponding results in Figure 4 and Figure 5. In 2004 we also included promotion opportunities among the job characteristics, but we omitted it in 2007. Therefore, for comparability purposes, we do not report the promotion 27 Figure 4: Importance of different job characteristics in 2004 Figure 5 shows how perceptions and preferences of doctors and nurses for different job characteristics have changed in 2007, after two years of experience. 24 The proportion of doctors and nurses who now give the greatest importance to "earning a good salary" has increased from 25% and 6% respectively to 36% and 24% respectively.25 Additionally, significantly more doctors and nurses recognize the importance of the conditions of the workplace, whereas fewer health workers give priority to "access to further training".26 To investigate the profile of the health workers who find the salary is the most important job characteristic in 2007, we use a simple probit regression and find that they are younger than average, earn more than average are more likely to have listed salary a the most important job characteristic in 2004 as well. Gender, job location, sector and marital status do not seem to play a significant role.27 opportunities results in Figure 4. Only about 10% of the doctors and the nurses considered promotion opportunities the most important job characteristic in 2004. 24 Both doctors and nurses have on average two years of work experience at the time of the second wave interview; the exact number of months of experience depends on when the program in their school finished. 25 The difference in the proportion of health professionals who give the highest importance to the salary in 2004 and 2007 is highly statistically significant. The corresponding p-values are equal respectively to 0.000 for the nurses and 0.07 for the doctors. 26 The difference in the proportion of health workers who give importance to the conditions of the workplace and in the access to further training in 2004 and 2007 is statistically significant (p-values equal to 0.0000 and 0.0001 respectively). Further analysis reveals strong differences in work place characteristics, as discussed in Annex C and further analysis is needed to understand how this affects preferences. 27 Results from these regressions are not reported here, but are available from the authors. 28 Figure 5: Importance of different job characteristics in 2007 A comparison between the figures reported in Figures 4 and 5 also reveals that the proportion of health workers who rank the "opportunity to help the poor" as the most important job characteristic has not changed significantly between 2004 and 2007. Interestingly, not all the health professionals who "want to help the poor" in 2007 are the same who "wanted to help the poor" in 2004, as, only 37% of the doctors and nurses who ranked "opportunity to help the poor" as the most important job characteristic in 2004 have retained the same preferences in 2007. On the other hand, 17% of the health workers who did not give primary importance to "the opportunity to help the poor" in 2004, did so in 2007. When analysing the profile of the health workers who ranked the opportunity to help the poor first in 2004, yet changed their preferences in 2007, we find that they are more likely to be female, married and with children. Moreover, one-third of them now consider the salary as the most important characteristic, while an additional third gives priority to good conditions of the workplace. The remaining health workers give major importance either to access to further training (an additional 22%) or having good colleagues (7%). Another exercise is to come up with a profile of the health workers who rank "the opportunity to help the poor" as the most important job attribute in 2007. Exploratory multivariate analysis show that older workers and those who expressed preferences for helping the poor in 2004 are more likely to rank "opportunity to help the poor" highest in 2007. In contrast, health professionals who have children and those who work in the public sector, especially in urban 29 areas, are less likely to consider the opportunity to help the poor as the most important job attribute in 2007.28 4.4. Job Satisfaction Whereas we asked the health professional to rank the importance of various job characteristics both in both waves, the 2007 has additional information the level of satisfaction with different job characteristics. We used a 5 point Likert scale from completely satisfied to completely unsatisfied. Figure 6 presents the findings. About 80% of the health workers are unsatisfied (20%) or completely unsatisfied (about 60%) with their salary. We obtain similar figures of dissatisfaction with chances of promotion and access to further training. A general sense of dissatisfaction also applies to the physical conditions of the workplace, especially in rural areas, where more than 20% of the respondents are completely unsatisfied with the current conditions of the health facility. Lack of mentoring and professional guidance constitutes an additional source of dissatisfaction for about 60% of the health workers. About 40% of the health workers are also unsatisfied with the opportunity to help the poor while in their current job. A comparison between the degree of (dis-)satisfaction of health workers in urban and rural locations suggests that they are in general similarly unsatisfied with most characteristics of their current job. Figure 6: Satisfaction with job characteristics 28 Results from these regressions are not reported here, but are available from the authors. 30 A different picture emerges when comparing satisfaction for those working in the private and the public sector, as reported in Figure 7. Here we can see noticeable differences with respect to the health workers' satisfaction with their salary, their chances of promotion and access to further training. While 70% of the health workers working in the private sector are unsatisfied with their job, the proportion of completely unsatisfied is below 40%, against the 70% of completely unsatisfied health professionals employed in the public sector. Figure 7: Satisfaction with job characteristics by sector Private sector Public sector In order to better understand the determinants of (dis-)satisfaction with different job characteristics, we conduct ordered probit regressions. Results for the individual levels of satisfaction with each of the eight job characteristics presented in Figures 6 and 7 are shown in different columns of Table 9. Higher values of the dependent variables indicate higher levels of satisfaction. The first column of Table 12 refers to the health workers' satisfaction with the current salary. We notice that doctors and nurses are equally dissatisfied (or satisfied) with their salaries; and that are significant differences in the levels of satisfaction based on location, with those in an urban job more satisfied. Those with higher wages are also more satisfied with their wages, and those with higher salary expectations in 2004 have lower levels of satisfaction in 2007. 29 29 Note that this funding may suffer from reverse causation as the health workers who are dissatisfied with their current salary are also more likely to believe that the salary is the most important job characteristic. 31 Table 12: The determinants of satisfaction with different job characteristics Dependent variable: Level of satisfaction with... Salary Promotions Training Conditions of Opportunities the workplace to help the poor (1) (2) (3) (4) (5) Doctor -0.01 0.07 -0.01 0.21 -0.09 [0.961] [0.801] [0.956] [0.407] [0.720] Urban job 0.60 0.63 0.57 0.37 -0.44 [0.085]* [0.049]** [0.095]* [0.239] [0.163] Public job -0.29 0.15 0.45 -0.81 -0.38 [0.442] [0.662] [0.199] [0.016]** [0.243] Urban*public -0.39 -0.98 -0.44 0.43 0.53 [0.348] [0.010]*** [0.257] [0.252] [0.146] Total wage 0.00 0.00 0.00 -0.00 0.00 [0.000]*** [0.043]** [0.508] [0.545] [0.357] Male 0.27 -0.08 0.23 -0.05 0.06 [0.158] [0.650] [0.174] [0.737] [0.705] Age 0.01 0.03 -0.01 0.01 0.02 [0.818] [0.267] [0.714] [0.641] [0.388] Born in Addis -0.16 -0.16 -0.24 0.02 0.06 [0.578] [0.509] [0.324] [0.917] [0.777] Married -0.31 -0.50 -0.28 0.46 -0.03 [0.281] [0.054]* [0.253] [0.043]** [0.899] With children 0.69 0.54 0.64 -0.13 0.09 [0.073]* [0.115] [0.050]** [0.669] [0.781] Expected wage in `04 -0.00 [0.075]* Good salary is most important -0.32 job char. in 2007 a [0.113] Access to training is most 0.02 important in 2007 a [0.913] Good work place conditions is 0.01 most important in 2007 a [0.943] Helping the poor is most 0.67 important in 2007 a [0.000]*** Constant -0.01 0.07 -0.01 0.21 -0.09 [0.961] [0.801] [0.956] [0.407] [0.720] Observations 216 226 226 226 226 R-squared 0.221 0.153 0.042 0.133 0.096 a these are dummy variables equal to 1 if the health worker in 2007 believes that the most important job characteristic is, respectively: the salary, the opportunity to undertake further training, good condition of the workplace, and the opportunity to help the por. * significant at 10%; ** significant at 5%; *** significant at 1% 32 The second column of Table 12 presents the results for the satisfaction with promotion opportunities. Those in an urban job are generally more satisfied, but those in a public sector urban job are less satisfied with their chances of promotion. Earning a relatively higher wage increases the level of satisfaction. Satisfaction with access to training is higher for those in urban jobs, and those working in the public sector are significantly less satisfied with the conditions of the workplace. Those who find the opportunity to help the poor the most important job characteristic are also more satisfied that their job offers them this opportunity. Overall, those who are better paid tend to be more satisfied with different aspects of their job. Family characteristics like being married and having children also have an effect. There do not seem to be gender or age-differences in the level of satisfaction, nor do we observe systematic differences between doctors and nurses. 4.5. Satisfaction with different spheres of life Both in 2004 and 2007, we asked the health professionals about their degree of satisfaction with different aspects of their life ­ their economic situation, their career choices, their job, and their life in general ­ using a 5 point Likert scale as before, and we can thus look at changes over time. In 2004, a very small proportion (less than 10%) of health professionals were unsatisfied or completely unsatisfied with their career choices, their economic situation (less than 10%) or their life in general (less than 20%), as shown in Figure 8. 33 Figure 8: Satisfaction with different spheres of life in 2004 Only three years later the health workers' degree of satisfaction with their economic situation and their life in general has deteriorated substantially. As Figure 9 shows, almost 60% of the health professionals are now completely unsatisfied with their economic situation while almost 30% are completely unsatisfied with life in general. Figure 9: Satisfaction with different spheres of life in 2007 Somehow surprisingly given the findings in the previous section on dissatisfaction with different aspects of their job, only 37% of the health workers are overall unsatisfied or completely unsatisfied with their job, and only 20% are unsatisfied with their career choice. 34 Table 13 reports results from multivariate analyses of the health workers' levels of satisfaction with respect to the different aspects of life presented above, where a positive estimated coefficient indicates a higher level of satisfaction. The first striking result is that doctors are more likely than nurses to be unsatisfied with their career choice, their economic situation, their life and their job. Earning more increases one's level of satisfaction with all aspects of life. The level of satisfaction with certain job characteristics also has a significant effect on the level of satisfaction with more general aspects of life, whereas we do not observe any significant differences along the lines of demographic or other individual characteristics. Table 13: The determinants of satisfaction with different aspects of life Dependent variable: level of satisfaction with.... (higher value = higher level of satisfaction) Career choices Economic Life in Current job situation general (1) (2) (3) (4) Doctor -1.04 -0.67 -0.65 -0.61 [0.000]*** [0.009]*** [0.024]** [0.018]** Urban location -0.44 -0.49 0.11 -0.04 [0.172] [0.133] [0.753] [0.900] Public sector -0.05 0.07 -0.20 -0.01 [0.877] [0.835] [0.584] [0.967] Urban*public 0.36 0.20 -0.02 0.17 [0.333] [0.594] [0.969] [0.649] Male -0.18 -0.26 0.02 -0.02 [0.267] [0.124] [0.895] [0.922] Age -0.00 0.04 -0.01 0.01 [0.926] [0.146] [0.742] [0.595] Born in Addis -0.31 0.11 0.25 -0.05 [0.183] [0.644] [0.335] [0.830] Total net wage 0.00 0.00 0.00 0.00 [0.003]*** [0.123] [0.001]*** [0.010]** Married in 07 -0.32 -0.09 0.17 0.07 [0.176] [0.692] [0.530] [0.772] With children in 07 0.20 0.00 0.07 0.28 [0.525] [0.988] [0.836] [0.368] Unsatisfied with salary -0.38 -0.23 -0.78 -0.57 [0.073]* [0.293] [0.000]*** [0.007]*** Unsatisfied with promotions -0.25 -0.09 -0.49 -0.20 [0.183] [0.644] [0.016]** [0.290] Unsatisfied with training -0.56 -0.40 -0.43 -0.28 [0.000]*** [0.013]** [0.014]** [0.079]* Unsatisfied with -0.08 -0.17 -0.23 0.00 opportunities to help the poor [0.585] [0.261] [0.199] [0.988] Unsatisfied with work place -0.60 -0.42 -0.63 -0.57 [0.000]*** [0.009]*** [0.001]*** [0.000]*** Observations 226 226 226 226 Pseudo R-squared 0.10 0.06 0.17 0.08 Note: p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% 35 4.6. Attitudes towards working on the side Given the relatively low salaries earned by health professionals in Ethiopia, we are also interested in individuals' attitudes towards working on the side. To that effect we asked health professionals to indicate their agreement or disagreement (using a 5 point Likert scale) with the statement: "doctors and nurses in the public sector have to work on the side in order to earn enough to support their families". Having asked the same question in 2004, we can compare the answers for the two years and find that an increasing proportion of doctors and nurses strongly agree or agree with the statement. In particular, nearly 80% of the doctors strongly agree in 2007, compared to about 50% in 2004, and over 50% of strongly nurses agree in 2007 compared to less than 20% in 2004. This indicates that those who disagreed with the statement in 2004 have changed position after having experience in the health sector. Figure 10: Attitudes toward working on the side 2004 2007 36 5. THE CHOICE BETWEEN RURAL AND URBAN POSTING Summary 1. There are significant differences in the profile of health workers who work in urban and rural posts. The most important findings when considering the effect of all the variables together in a multivariate analysis reveals that health workers coming from a rural background are more likely to work in a rural area. Doctors are also less likely than nurses to work in a rural area. 2. The proportion of nurses willing to work in a rural location in the long run has declined from 34% to 18% between 2004 and 2007, whereas the proportion of doctors willing to take a rural post has increased from 9% to 11%. The latter is possibly explained by the higher earnings for doctors in rural areas. Females are less likely to have long-term preferences for a rural post, while older and married health professionals, those who are highly motivated and those who expressed a preferences for rural posts in 2004, are more inclined to work in a rural area in the long-run. 3. About 60% of the health professionals are not currently working in their first job and the majority of nurses who left their first job was working in a rural area. The most important reason for nurses to leave was that they were not happy with the location. Doctors, in contrast have as the most important reason to leave a first position that they are not happy with the salary. Had these doctors and nurses kept their first job the rural-urban distribution of health professionals would be more balanced, with 44% health workers based in rural locations. 4. To identify the economic incentives that would induce doctors and nurses to take up a rural post we asked contingent valuation questions that help us to identify health workers' reservation wages to work in a rural area, which we can compare with similar data from the first round (2004). The minimum salary that makes 80% of the nurses accept a job in a rural area has gone up with 50% from close to 2,000 Birr in 2004 to 3,000 Birr in 2007. For doctors, the salary to get 80% of the doctors take up a job in a rural area has gone up with 60% from close to 3,600 Birr in 2004 to 5,500 Birr in 2007. 5. Given the current salaries of the doctors and nurses employed in rural areas, to get 80% of the nurses and doctors take up a rural position, we would need a 284% salary increase for nurses and a 245% salary increase for doctors. 6. The variation in reservation wages across health workers is partially explained by gender, with women requiring higher wages to take up a position in a rural area. Health workers who are catholic have lower reservation wages, and this most likely reflects that 37 many Catholics went to one Catholic NGO nursing school with a strong reputation for motivating its students to help the poor. 5.1. Overview To better understand the unbalanced distribution of health workers between urban and rural locations, we try to build a profile of who is working in a rural area, and contrast our findings with those obtained in 2004. We then investigate the preferences of health workers to work in a rural or urban area, and analyse why health workers quit their first job. Finally we investigate health workers' reservation wages to work in a rural area and compare this with the 2004 reservation wages. 5.2. The profile of health workers working in rural and urban areas We focus on health workers' individual characteristics that make health workers more likely to take a job in a rural rather than an urban facility. In Table 14 we compare demographic characteristics, education achievements, motivations and career preferences of the health workers located in rural and urban areas. The figures indicate that geographical distribution of health workers does not seem to be particularly gender-biased; although a relatively higher fraction of female health professionals are located in urban areas, the difference is small in magnitude and statistically insignificant. Second, married health workers are more likely to be working in rural areas, especially if they have children. Although this seems surprising at first, as health professionals with children typically prefer to work in an urban area in order to have access to better education for their children, we have to bear in mind that the children of our cohort of health workers are mostly below school going age. Third, health professionals who were born in Addis Ababa and, to a lesser extent, who took their medical education in Addis Ababa, are more likely to be currently working in an urban area, confirming that urban background may significantly decrease the likelihood to work in a rural location. Having done an internship in an urban location does not seem to affect the likelihood to work in an urban post. Table 14 also provides information on the health professionals' score in the medical knowledge test that we administered during the first wave of the survey in 2004. We find that the best qualified health workers are currently working in urban areas; although the difference is small in magnitude, it is statistically significant. Finally, the last row of Table 14 suggests that the health workers who in 2004 preferred to work in a rural (urban) area the in long run are indeed more likely to be currently working in a rural (urban) area. 38 Table 14: Characteristics of health workers by job location CURRENTLY IN A CURRENTLY IN AN URBAN AREA RURAL AREA Female 35% 43% (p-value) (p = 0.16) Married 31% 22% (p-value) (p = 0.06) With children 25% 9% (p = 0.0007) Born in Addis 6% 14% (p-value) (p = 045) School in Addis 28% 35% (p-value) (p = 0.16) Internship in an urban area 20% 18% (p-value) (p = 0.40) Test score (in percentage point) 0.47 0.50 (p-value) (p = 0.03) Would prefer to work in a rural 42% 17% area in 2004 (p-value) (p = 0.000) Note: the numbers in the table indicate the proportions of health workers in rural and urban areas belonging to the categories listed in the first column. A more advanced approach is to consider the simultaneous effect of these and other factors on the likelihood to be working in a rural area. In Table 15 we report the results of a simple probit estimation where working in a rural area is the dependent variable, where the coefficients reflect marginal effects. Columns one and two report the results for nurses only, while columns three and four report the results for the pooled data from nurses and doctors, which we expect to give more robust results because of its larger sample size. 39 Table 15: Multivariate analysis of the decision to work in a rural area Nurses Nurses and Doctors (1) (2) (1) (2) Female -0.06 -0.05 -0.01 -0.00 [0.609] [0.718] [0.891] [0.971] Age 0.00 -0.01 0.01 -0.00 [0.956] [0.498] [0.475] [0.994] Born in Addis 0.17 0.04 -0.11 -0.11 [0.446] [0.850] [0.351] [0.311] Distance from school when 6 yrs old 0.00 0.00 0.00 0.00 [0.575] [0.427] [0.144] [0.103] Test score 0.20 -0.47 0.13 -0.08 [0.755] [0.529] [0.770] [0.866] Household EXP. 2004 -0.00 -0.00 0.00 0.00 [0.772] [0.811] [0.564] [0.736] "Help the poor" 2004 0.05 -0.08 0.04 -0.04 [0.693] [0.560] [0.643] [0.618] Catholic 0.28 0.21 [0.136] [0.176] Protestant -0.29 -0.20 [0.027]** [0.018]** Tigray -0.05 -0.15 [0.875] [0.324] Married 0.20 -0.11 [0.399] [0.366] With children 0.20 0.44 [0.502] [0.038]** Wanted to work in a rural area in `04 0.00 -0.04 [0.977] [0.129] Doctor -0.22 -0.19 [0.047]** [0.678] Observations 91 89 170 167 Note: The table shows results from dprobit regressions. We report marginal effects for the continuous variables, and the effect of a change from 0 to 1 for the dichotomous variable. P-values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. Focusing on the results in columns three and four, we observe that doctors are less likely to be working in a rural area. We also find suggestive evidence that health professionals who grew up in a rural area ­ those who spent more time to reach their primary school when they were 6 years old ­ and those who have children, are more likely to be working in a rural facility, but the results are not strong enough to be significant at the 10% level. In contrast, health professionals who are Protestant seem to be less likely to be working in a rural area, it is not clear why this is, but one possible explanation is that facilities run by organizations with a protestant denomination are more likely to be in an urban area. We also find that those working in rural areas are not lower skilled compared to those working in urban areas, since the variable test score does not show up significant. The welfare of the household in which the health worker 40 grew up also has no significant effect on the probability to work in a rural area, and motivation also has no effect. 30 5.3. The preference for a rural or urban posting in 2004 and 2007 In 2004, 9% of the medical students and 34% of the nursing students expressed a preference for working in a rural location in the long term. By 2007 the proportion of nurses willing to work in a rural post in the long term declined to 18%, whereas the proportion of doctors increased to 11%, likely because of the higher salaries in rural areas. Table 16: Long-term preferences for rural or urban posts 2004 2007 Medical Nursing Total Doctors Nurses Total students students Rural area 9% 34% 26% 11% 18% 16% Urban area 91% 66% 74% 89% 82% 84% Obs. 88 175 263 87 177 264 Note: the 2004 percentages are computed only for the health workers who are not missing in 2007 Figure 11 and Figure 12 show the reasons why health workers prefer to work in a rural or an urban area in the long term, both in 2004 and 2007. Of the health professionals who prefer a rural post in the long term, a larger proportion in 2007 than in 2004 believe that the most important reason for doing so is the possibility to "provide health care where it is needed most". If we turn to the most important reasons to prefer an urban location, we see that the reason has shifted from access to promotion opportunities in 2004 to being closer to family and friends and the opportunities to find another job in 2007. More than 10% of the health workers also mention the opportunities to get a scholarship to pursue further education. 30 We do, however, find that the health worker's current welfare is lower for those living in rural areas. 41 Figure 11: Reasons for working in a rural or an urban location in 2004 Figure 12: Reasons for working in a rural or an urban location in 2007 To explore the individual characteristics that make health workers more or less likely to have long-term preferences for a rural rather than an urban post, we run a simple probit model with as the dependent variable preference to work in a rural post in the long term. Table 17 presents the results with the first and second column looking nurses only, while the third and the fourth column consider both doctors and nurses. The model in columns 1 and 3 only include exogenous variables like gender, age, rural or urban background ­proxied by the walking distance (in minutes) to the school when they were 6 years old, whether born in Addis Ababa, household welfare in 2004, knowledge test score in 2004, and willingness to help the poor in 2004. In columns 2 and 4 we add potentially endogenous or simultaneous variables like marital status, ethnicity, and religion; we also include the 2004 preferences for a rural or an urban post. 42 Table 17: The determinants of the health workers' preferences for a rural job in the long run Nurses Nurses and Doctors (1) (2) (3) (4) Female -0.14 -0.21 -0.10 -0.11 [0.069]* [0.004]*** [0.084]* [0.002]*** Age 0.01 0.00 0.01 0.00 [0.240] [0.835] [0.080]* [0.273] Born in Addis 0.01 -0.01 -0.09 -0.04 [0.954] [0.796] [0.303] [0.316] Distance from school when 6 0.00 0.00 0.00 -0.00 [0.231] [0.434] [0.440] [0.940] Test score -0.01 -0.25 0.08 -0.12 [0.986] [0.215] [0.794] [0.452] Household assets 2004 0.00 0.00 0.00 0.00 [0.914] [0.083]* [0.244] [0.029]** "Help poor" 2004 0.16 0.17 0.07 0.08 [0.055]* [0.013]** [0.246] [0.062]* Currently in a rural area -0.01 -0.00 [0.701] [0.866] Wanted a rural job in 2004 0.15 0.19 [0.023]** [0.002]*** Tigray 0.06 [0.464] Catholic 0.11 0.11 [0.103] [0.148] Protestant -0.02 -0.02 [0.650] [0.494] Married 0.31 0.07 [0.049]** [0.128] With children -0.02 0.005 [0.552] [0.926] Doctor -0.09 0.01 [0.230] [0.759] Observations 117 86 199 167 p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% The results indicate that female health professionals are less likely to have long-term preference for a rural post, whereas married health workers may be more likely to prefer a rural post in the long run. Having grown up in a rural area has some effect on the likelihood to prefer a post in a rural area in the long term, but not enough to be significant at the 10% level. Health professionals who come from a wealthier back ground tend to be more likely to prefer in to work in a rural area, a possible consequence of the fact the higher earnings in rural areas, while ranking `opportunity to help the poor' in 2004 - our proxy for motivation - has a positive effect on the long term preference for an rural posting, as do preferences for a rural post in 2004. Having children and being currently employed in a rural area tend to reduce the likelihood for nurses to have long-term preferences for a rural job, but not significantly so. Finally, doctors do not seem to be less willing to work in a rural area in the long term. 43 5.4. Reasons for quitting the first job Job turnover is typically high early in a career and this is also confirmed for our data: about 60% of the health professionals who are currently employed are not working in their first job. More interesting is to analyse who leaves their job and why. Table 18 shows that 53% of the health professionals who quit their first job were located in a rural area; 67% of them have now moved to an urban area. The initial distribution of the health workers (in their first job) was thus considerably more balanced and meant that 44% rather the current 29% of health workers were in a rural posting. (with 25% of the doctors compared to the current 17% employed in rural areas). Table 18: The first job in the health sector DOCTORS NURSES TOTAL Rural area 30% 64% 53% (17) (72) (89) Urban area 70% 36% 47% (40) (40) (80) Why did health workers leave their first job? The reasons differ depending on whether the post was in a rural or urban area, and whether the respondent is a doctor or a nurse, as depicted in Figure 13 and Figure 15. For 45% of the doctors working in a rural area and about 40% of the doctors working in an urban area, dissatisfaction with the salary was the most important reason for leaving the job. Only about 10% of the doctors working in rural area left the first job because they were not happy with the location. In contrast, about 40% of the nurses whose first job was in a rural area left the job because they "did not like the location". 44 Figure 13: Reasons why the doctors quit the first job, by location Figure 14: Reasons why nurses quit their first job, by location To better understand in how far salaries and the expectations about salaries play a role, we compare the monthly gross salaries, comprehensive of allowances, which doctors and nurses received while in their first job, either in urban or rural areas with the monthly gross salary that they expected to earn from their first job when we surveyed them in 2004.31 The 133 doctors and nurses who quit their first job were expecting to receive an average first job salary equal to 3,455 Birr and 1,751 Birr respectively. Roughly 80% of them received a first job salary which was below their expected salary, with the doctors receiving a salary that 31 Once again, we adjusted the 2004 values for inflation, as explained in footnote 13. 45 was on average 20% below their expected salary while nurses received a salary that was on average 50% of their expected salary, suggesting that expectations about salaries play an important role. Table 19: First Job Salary vs. Expectations in 2004 (Means) DOCTORS NURSES rural urban Total rural urban Total area area area area Total gross salary (first job) 2592 2546 2560 619 747 704 Minimum gross salary (first job) 1064 1200 1064 519 500 500 Maximum gross salary (first job) 7000 5000 7000 950 2380 2380 Total salary expected in 2004 3171 1391 Minimum salary expected in 2004 1860 883 Maximum salary expected in 2004 9300 3720 Has moving jobs paid off for those who did move? Distinguishing between health workers who moved (i) between rural areas; (ii) from an urban to a rural area, (iii) between urban areas; and (iv) from a rural to an urban area, as reported in Figure 15, we find that nearly all nurses who decided to quit their first job (93%) experienced a salary increase as a result. The increase was 340 Birr per month on average. Nurses who moved from an urban to a rural location also increased their earnings. In contrast, less than 50% of the doctors who changed job increased their earnings compared to their first job ­ with an average salary increase of close to 2000 Birr. Among doctors who moved from a rural to an urban location, the proportion that experienced an increase in earnings is even smaller, namely less than 30%. Doctors who saw their earnings reduce by changing job, lost on average 1800 Birr per month. Figure 15: Health workers who experienced a salary increase after quitting the first job 46 5.5. Reservation wages to work in a rural area The above results, as well as earlier quantitative and qualitative evidence, suggest that salaries in rural postings are too low to get a large proportion of health workers accept a position in a rural area.32 With this in mind, we asked health workers for the salary that would make them take up a rural posting, using careful contingent valuation questions, both in 2004 and 2007. We asked the health professionals to choose between a job in Addis Ababa, where they would earn their current salary, and a job in a rural area where they would earn their current salary plus a well defined top-up. We asked this question for different amounts of top-ups. If the health worker still preferred the job in Addis Ababa even for the highest top-up, we asked him to state the minimum salary that would make him accept the rural post. Annex D provides details about the methodology. Figure 16 shows the minimum top-ups of current salaries required by the nurses in order to move from a post in Addis Ababa to a rural post. About 6% of the nurses would require a top- up larger than the 1,800 Birr, i.e. the maximum top-up in the questions. Additionally, 10% of the nurses stated that they would never work in a remote location, no matter the salary. Figure 16: Contingent valuation Nurses' required top-ups for a rural post in 2007 In 2004 we also used contingent valuation questions in order to identify the minimum salary at which students were willing to switch from a job in Addis Ababa to a job in a rural area, either 32 See Lindelow and Serneels (2006) and Serneels, Lindelow, Montalvo and Barr (2007) 47 200 Km or 500 Km away from Addis Ababa. We can therefore compare the answers in 2004 and 2007.33 However, in 2004 the contingent valuations' base salary was fixed at 700 Birr and if the health workers were unwilling to take a rural post even when they were offered the maximum wage of 1,200 Birr we did not ask them to state their reservation wage. Therefore, the 2004 data are left-censored at 700 Birr in 2004 prices or 1116 Birr in 2007 prices, and right- censored at 1,200 Birr in 2004 prices or 2232 Birr in 2007 prices (see Annex C for details). Both in 2004 and 2007, to make more than 40% of the nurses take up a rural position, a salary of 1,500 Birr was required. But whereas in 2004 a salary of 2,046 Birr was required to make 80% of the nurses take up a post 500 Km from Addis Ababa, in 2007 the required salary had to get 80% of nurses take up a rural post had gone up to 3,000 Birr. Considering that the average monthly salaries of a nurse working in a public facility in a rural area currently equals 781 Birr, this would require an increase of 284%. Figure 17: Nurses' reservation wages to work in rural areas in 2004 and 2007 A similar analysis for the doctors, first focusing on the 2007 data only, indicates that roughly 20% of the doctors requires a top-up that exceeds the maximum top-up stated in our contingent valuation question (2,500 Birr); while one doctor would never work in a remote location, no matter the salary. 33 We adjust the 2004 values for inflation, as explained in footnote 13. 48 Figure 18: Contingent valuation Doctors' required top-ups to work in a rural area in 2007 Figure 19 compares the reservation wages for doctors in 2004 and 2007. As before, the 2004 data are left-censored (at 2232 Birr), and right-censored (at 3,906 Birr). Whereas in 2004, a salary of about 3,900 Birr was sufficient to bring 65% 34of the doctors in rural areas, in 2007, only 50% would be willing to work in a rural area at this salary. In order to make 65% of the doctors switch from a post in Addis Ababa to a post in a remote location we would now need a salary of 4300 Birr, and if we wanted to raise this percentage to 80% we would need to pay the doctors a monthly salary of 5,500 Birr (about 605 US dollar). Considering that the average monthly salary of a doctor working in a public facility in a rural area equals 1,593 Birr this would an increase of 245% for the doctors. 34 In the 2004 wave, 35% of the medical students stated that they would not be willing to work in a rural area 500 Km from Addis Ababa even if they were offered the highest possible salary (in the set of contingent valuation questions) of 3900 Birr (2100 Birr in 2004). 49 Figure 19: Doctors' reservation wages to work in rural areas in 2004 and 2007 What explains the differences in health workers reservation wages to work in a rural area? The analysis for 2004 suggested that three factors play an important role: (i) having grown up in a rural area, which is associated with lower reservation wages and thus increase the willingness to work in a rural area; (ii) household welfare with those coming from richer households having higher reservation wages (i.e. less likely to be willing to work in a rural area); and (iii) motivation to help the poor, which lowers reservation wages (and thus increases likelihood to work in a rural area). Using simple least square regressions with the reservation wage as dependent variable gives the results reported in Table 20. Columns one to four reports the results for nurses only, while columns four to six reports the results for doctors and nurses. Columns and four only include exogenous variables like gender, age, medical knowledge (test score), whether born in Addis Ababa, distance to school at age 6, welfare of the household where he health worker grew up (household expenditures) and the importance given to the job attribute `opportunity to help the poor', which is our proxy for motivation. Columns two and five add variables on ethnicity and religion, while columns three and six introduce additional variables related to job preferences add variables The results in columns three and four of Table 20 suggest that female doctors require a higher salary to work in a rural post. Having a better test score has no significant effect, countering the common concern that lower skilled health workers end up in rural areas. Coming from a 50 wealthy background tends to increase the reservation wage, but not significantly so. Whereas , the importance of helping the poor in 2004 has no effect on the reservation wages, health workers with a Catholic back ground are significantly more likely to have lower reservation wages; as argued in earlier work, this reflects that many Catholics went to one Catholic NGO nursing school that has a strong reputation for strongly motivating its student. Table 20: The determinants of the reservation wage to work in a rural area Nurses Nurses and Doctors (1) (2) (3) (4) Female 0.14 0.07 0.34 0.34 [0.421] [0.692] [0.008]*** [0.008]*** Age -0.01 0.00 -0.00 -0.00 [0.607] [0.987] [0.910] [0.979] Test score -0.54 -0.34 -0.48 -0.22 [0.575] [0.725] [0.475] [0.746] Born in Addis 0.24 0.38 -0.04 0.05 [0.465] [0.254] [0.805] [0.760] Distance school at 6 -0.00 -0.00 0.00 0.00 [0.953] [0.716] [0.665] [0.778] HH EXP 2004 -0.00 -0.00 -0.00 -0.00 [0.388] [0.630] [0.279] [0.385] Help poor 2004 0.01 0.18 -0.11 -0.02 [0.958] [0.320] [0.373] [0.863] Tigray 0.51 0.13 [0.205] [0.553] Catholic -0.54 -0.39 [0.034]** [0.057]* Protestant 0.13 0.28 [0.490] [0.052]* Married 1.03 [0.000]*** With children 0.13 [0.433] Doctor 1.06 0.07 [0.000]*** [0.781] Constant 8.51 7.85 8.04 7.69 [0.000]*** [0.000]*** [0.000]*** [0.000]*** Observations 92 92 169 169 R-squared 0.049 0.145 0.334 0.382 p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% 51 6. LIKELIHOOD TO MIGRATE ABROAD Summary 1. More than 50% of the health professionals plan to emigrate abroad in the next two years. Health workers expect to earn higher salaries abroad. 2. Health workers who are more satisfied with their current job are significantly less likely to migrate abroad. Protestant health workers are more likely to migrate, possibly because they have been more exposed to international work. 3. Using data from contingent valuation questions that identify the wages for which doctors and nurses would stay in Ethiopia, we find that we would need to pay nurses and doctors a salary of 6,000 Birr and 10,500 Birr respectively if we want to persuade 70% of the nurses and 80% of the doctors not to leave the country. Given the current salaries in the Ethiopian public sector in urban areas, our estimates indicates a 600% salary increase of for the nurses and a 500% salary increase for the doctors. 4. The doctors are more inclined to leave the country than the nurses, and would require a higher salary in order to stay in Ethiopia if they had the chance to move abroad. We do not find any significant impact of gender, age, marital status and other individual characteristics on the reservation wage, although we find that those coming from a wealthier back ground tend to be less likely to move abroad. Protestant health professionals and those of Tigray ethnicity have a significantly higher reservation wage to stay in Ethiopia. 6.1. Overview A major concern among policymakers in Ethiopia is that health workers are likely to migrate abroad. Although some evidence exists (see Section 2), there is limited hard evidence that illustrates the scope and nature of the problem. To address this we asked a number of questions related to health workers' likelihood to migrate. In this analysis we focus on the data generated from two questions, one on the health workers plans to migrate abroad over the next two years, and one using contingent valuation questions similar to the one used for the rural ­ urban choice. 52 6.2. Likelihood to migrate abroad in the coming two years About 52% of the nurses and 60% of the doctors plan to migrate abroad in the next two years. That health workers are serious about their intention to migrate is clear from the fact that more than 80% of them have applied for a lottery visa, or DV, which would allow them to leave the country. Nearly 90% of the health professionals who are planning to emigrate would like to go to the United States, and almost all of them (97%) would like to work in the health sector if they move abroad; only 25% think that it may be difficult to find work in the health sector abroad. When asked about expectations regarding earnings abroad, the average doctor expects to earn 11,280 USD per month while the average nurse expects to earn 10,198 USD per month.35 67% of health workers are more inclined to leave the country now compared to two years ago, possibly because they have become more aware of the earnings and working conditions in the health sector in Ethiopia. To identify which health workers have strong preferences to migrate abroad, we estimate a probit model with planning to migrate abroad in the next two years as the dependent variable and individual characteristics like gender, age, medical knowledge (test score), rural back ground and motivation to help the poor as explanatory variables. The results are presented in Table 21. As before, we estimate the model separately on nurses (columns one to three) and nurses and doctors (columns four to six). In columns one and three respectively, we only include exogenous variables, and then add ethnic and religious variables (columns two and four), and finally add other potentially relevant variables like married, having children and satisfaction with the job. The results indicate that protestants and those who less satisfied with their job are more likely to have migration plans. It is unclear how to interpret the protestant result, and further analysis is needed here, but the increased activity of US NGOs may play an important role., and those who are protestant are more exposed to international work and are therefore more likely to plan accordingly; it may also be possible that health workers become protestant in order to increase their chances to migrate to the US. A more intuitive result is that those who are satisfied with their current job are less likely to plan to migrate in the near future. 35 About half of the nurses and 60% of the doctors expect to earn equal or above 10,000 USD per month. 53 Table 21: Probit estimation of the willingness to emigrate abroad Nurses Nurses and Doctors (1) (2) (3) (4) Female -0.10 -0.07 -0.11 -0.10 [0.346] [0.507] [0.185] [0.259] Age -0.01 -0.00 -0.01 -0.01 [0.732] [0.831] [0.334] [0.521] Test score -0.47 0.09 -0.46 0.07 [0.411] [0.887] [0.312] [0.880] Born in Addis -0.23 -0.22 -0.04 -0.09 [0.258] [0.295] [0.746] [0.467] Distance to school at age six 0.00 0.00 0.00 0.00 [0.390] [0.322] [0.901] [0.691] HH EXP 2004 -0.00 -0.00 -0.00 -0.00 [0.965] [0.701] [0.792] [0.719] "Help the poor" 2004 0.10 0.11 0.05 0.06 [0.324] [0.346] [0.516] [0.536] Tigray 0.04 0.05 [0.889] [0.771] Catholic 0.02 0.01 [0.908] [0.934] Protestant 0.22 0.22 [0.068]* [0.027]** Married -0.06 -0.14 [0.766] [0.248] Children -0.08 -0.05 [0.732] [0.778] Satisfaction with the job -0.03 -0.04 [0.201] [0.037]** Doctor 0.15 0.04 [0.204] [0.733] Observations 117 117 199 199 p values in brackets * significant at 10%; ** significant at 5%; *** significant at 1% 6.3. At what wages would health workers stay in Ethiopia? To find out how earnings can affect health workers' decision to migrate abroad, we ask contingent valuation questions that identify the salary that would induce them to stay in Ethiopia rather than migrate abroad. Annex E gives more details about the exact question. Similarly to before, Figure 21 shows the salary top-ups required by nurses in order to stay in Ethiopia, when given the choice to migrate to the US and shows that about 30% of the nurses would choose to stay in Ethiopia at their current wage, while close to 30% state that they would leave Ethiopia if they had the possibility, whatever salary they are paid in Ethiopia. 54 Figure 20: Top-ups required by the nurses to stay in Ethiopia Figure 21 shows the top-ups required by doctors to stay in Ethiopia. While about 30% would stay in Ethiopia at their current wage, about 10% would not remain in Ethiopia if they were offered to opportunity to migrate, whatever salary they would be earning in Ethiopia. Figure 21: Top-ups required by the doctors to stay in Ethiopia In Figure 22 we plot the reservation wages ­ or the minimum salaries required to stay in Ethiopia (rather than the top-ups) - for both nurses and doctors. Figure 22 tells us that to keep 55 70% of the nurses in Ethiopia we would need a salary of 5,358 Birr per month or above, which is close to seven times the current average salary of a nurse working in the public sector in an urban location, (which is about 800 Birr). For doctors we would need more than 10,500 Birr per month to persuade 80% of them to stay in Ethiopia. An additional 5,000 Birr would be required to persuade a further 10% not to leave the country if they had the chance. Considering that the average salary of a doctor working in the public sector in an urban location is about 1,700 Birr, this means that we would require a salary increase of about 500% to make 80% of the doctors stay in Ethiopia. Figure 22: Reservation wages to stay in Ethiopia What drives the differences in these reservation wages? In Table 22 we show the results of a OLS estimation with the logarithm of the reservation wage as a dependent variables and the familiar individual characteristics as explanatory variables. We find that neither gender, nor age, medical knowledge, rural background have any effect. Those coming from a wealthier back ground may be more likely to want to migrate abroad. Motivation to help the poor also has no effect, but Protestant and Tigray nurses require a higher wage to stay in Ethiopia, whereas nurses who have children are less likely to leave the country. Further work is needed to analyse the role of these factors in more depth. We also observe, not surprisingly, that doctors require significantly higher salaries than nurses in order to stay in Ethiopia if they had the chance to migrate. 56 Table 22: Least squares estimation of the log of reservation wage to stay in Ethiopia Nurses Nurses and Doctors (1) (2) (3) (4) Female -0.14 -0.19 -0.26 -0.23 [0.492] [0.363] [0.347] [0.426] Age -0.02 0.01 -0.06 -0.04 [0.541] [0.837] [0.187] [0.385] Test score 0.95 1.15 0.73 0.90 [0.393] [0.320] [0.613] [0.559] Born in Addis -0.35 -0.27 -0.02 -0.03 [0.376] [0.481] [0.944] [0.942] Distance to school 0.00 0.00 -0.00 0.00 [0.825] [0.584] [0.931] [0.875] HH EXP 2004 0.00 -0.00 -0.00 -0.00 [0.940] [0.840] [0.084]* [0.101] "Help the poor" `04 -0.20 0.07 -0.00 0.13 [0.321] [0.743] [0.989] [0.663] Tigray 1.12 0.37 [0.035]** [0.453] Catholic 0.03 0.08 [0.912] [0.862] Protestant 0.45 0.57 [0.048]** [0.083]* Currently in rural area -0.21 -0.22 [0.284] [0.432] Wanted a rural post in `04 -0.11 0.01 [0.621] [0.982] Married -0.19 -0.22 [0.608] [0.571] With children -0.39 -0.15 [0.389] [0.798] Doctor 0.87 1.00 [0.020]** [0.019]** Constant 8.11 7.57 10.94 10.28 [0.000]* [0.000]*** [0.000]*** [0.000]*** ** Observations 95 89 172 166 R-squared 0.054 0.222 0.158 0.199 p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% 57 58 7. CONCLUSION There is a growing consensus that the human resources for health have a determining role to play regarding the utilization of health services in Ethiopia. To design effective and evidence based policies, more empirical evidence is needed. In this paper we summarize the findings of the second wave of a cohort study with health professionals. Revisiting health professional who were in their last year of school in 2004 and had entered the labor market at the second visit, in 2007, allows us to analyze three central concerns regarding human resources for health in Ethiopia: their satisfaction and motivation, the geographical imbalances in the distribution of health workers, and their migration abroad. An important finding is that Health workers tend to be unsatisfied with most aspects of their job, and especially their salary, their training opportunities and their chances of promotion. We also find that work experience has significantly altered health worker preferences regarding the importance of different job characteristics. Although to some extent this seems to reflect the process of `growing up' and being confronted with the sometimes harsh realities of life (their satisfaction with their career choice, their economic situation and their life in general have all deteriorated between 2004 and 2007), there is also heterogeneity among health workers. The data also reveal substantial heterogeneity in job attributes across the different types of jobs (with doctors in rural areas, for example, earning significantly more than those in urban areas, while doctors working in an urban public facility receiving more training), and may indicate that mismatches in the labor market play a potentially important role. Regarding the geographical distribution our data confirms that there are substantial geographical imbalances about one third of nurses and less than one fifth of doctors working in rural post. In line with earlier work we find health workers coming from a rural background are more likely to want to work in a rural area. This provides a strong argument for policymakers to take into account where a health worker grew up when allocating health workers to positions. Incentives also seem to play a role as the proportion of doctors willing to take a rural post in the long term has increased, presumable because the earnings for doctors in rural areas have also increased. We find that, given the current salaries, to get 80% of the nurses and doctors take up a rural position, we would need a 284% salary increase for nurses and a 245% salary increase for doctors. International migration of health workers abroad is a major concern in Ethiopia, and there is evidence that indicates this is a justified concern. Our data confirms that more than 50% of the 59 health professionals plan to migrate abroad in the next two years, where they expect to earn higher salaries. A key finding is that health workers who are more satisfied with their current job are significantly less likely to migrate abroad. But again incentives also play a role. And to persuade 70% of the nurses and 80% of the doctors not to leave the country, our estimates indicate that salaries in the urban public sector have to increase with 600% while those for doctors need to increase with 500%. Although this paper reports some interesting first results, further analysis is needed to gain more insights on the above topics. 60 8. REFERENCES Centre for National Health Development in Ethiopia, "Health and Health development Indicators, 2004-2005" (HHRI), The Earth Institute at Columbia University. Clemens M and G. Pettersson (2008), "New data on African health professionals abroad", Human Resources for Health, 6:1. Lindelow, M and P. Serneels (2006) "The performance of health workers in Ethiopia: Results from qualitative research". Soc Sci Med, 62(9), 2225-2235. PAHO (Pan American Health Organization). 1998. "Human Resources: a Critical Factor in Health Sector Reform." Report of a meeting in San José, Costa Rica, 3-5 December 1997. Washington, D.C Serneels P., M. Lindelow, (2005) "An honourable calling? Findings from the first wave of a cohort study with final year nursing and medical students in Ethiopia", research report for The World Bank and the Ministry of Health, Ethiopia, forthcoming as Africa Research Working Paper World Bank. Serneels P., M. Lindelow, J. Montalvo and A. Barr (2007) "For public service or money: Understanding geographical imbalances in the health workforce", Health Policy and Planning, Volume 22, Number 3, Pp. 128-138 Serra, D. (2006) "Report on mission to Addis Ababa, 31 July ­ 5 August", World Bank. World Bank (2002) "Public health and World bank Operations", The Human Development Network, Health Nutrition and Population Series, World Bank. World Bank (2003). Public Expenditure Review: Public Spending in the Social Sectors (PER). Washington, D.C.: The World Bank World Bank (2007) "Ethiopia Health Sector Development Program. Implementation Completion and Results Report", Report No: ICR0000383. WHO (2000) "What resources are needed?", Chapter 4, World Health Report 2000. Geneva: World Health Organization. WHO (2006), World Health Report: Working together for health. World Health Organization. 61 ANNEX A SURVEY METHODOLOGY The first wave of the health Worker Cohort Study was conducted in April 2007, involving 219 medical students and 90 medical students in their final year of study. Table A1 shows the eight nursing schools and the three universities which were selected to participate in the study36. Public schools funded by the federal Government, public schools funded by the regional government, private NGO schools and private for profit schools are all represented in our 2004 sample. Table A1: The original cohort study sample 2003/4 medical Surveyed Surveyed 2003/4 final students Nursing Medical School Ownership Region nursing (before starting students students students the internship) Addis Ababa Addis 32 30 Public-Federal Government 136 67 University Ababa Gonder College 30 30 Public-Federal Government Amhara 55 53 Jimma University 20 30 Public-Federal Government Oromiya 45 72 Asella Nursing Public-Regional Health 30 Oromiya 101 School Bureau Yirgalem School of Public-Regional Health 30 SNNPR 90 Nursing Bureau Selihom Nursing 27 Private SNNPR 30 School St. Lukes Catholic 20 NGO Oromiya 27 College Medco BioMedical Addis 30 Private 107 College Ababa Total 219 90 551 192 Source: Serneels P., M. Lindelow, J. Montalvo and A. Barr (2005) "An honourable calling? Findings from the first wave of a cohort study with final year nursing and medical students in Ethiopia". Follow-up efforts in 2005, 2006 and 2007 The last section of the first wave questionnaire registered several contact details of the students, i.e. their personal address and phone number, the contact details of a living relative, the contact details of a close friend, and the contact details of a friend or relative based in Addis Ababa. Unfortunately, 11 nursing students did not provide us with any contact detail, making it impossible for us to include them in future waves of the survey. 36 Please the 2004 descriptive report for details about the sampling procedure employed. 62 The doctors in our sample completed their degree between June 2005 and August 2006. The first follow-up attempts were made during the summer of 2005. We managed to contact all the 90 doctors and 172 of the original 219 nurses. During the summer of 2006 we tried to re-contact the 262 health professionals for which we had contact details from the follow-up phone interviews conducted in 2005. We were able to contact 83 doctors and 168 nurses. In addition, we were informed that 5 nurses had moved abroad and one had died in 2006. Excluding the 11 nurses that gave us no contact details in 2004, the reason we were not able to contact 36 additional nurses in 2005 and 2006, and 7 doctors in 2006 was that the contact details they gave us in 2004 proved to be already outdated in 2005 and/or in 2006. In particular, either: 1) the phone numbers did not work, or 2) nobody answered, 3) the friends and/or relatives who we were able to contact did not seem to know the health workers or 4) the friends and/or relatives who we were able to contact refused to give us their contact details. The second wave of the cohort study started in May 2007 and ended in September 2007.37 By using all available contact details from both the first wave survey and the follow-up phone calls of 2005 and 2006, we managed to contact and interview a total of 88 doctors and 177 nurses in their job location, as shown in Table A2. A total of 9 health workers ­ 2 doctors and 8 nurses ­ emigrated abroad between 2005 and 2007; additionally, 2 nurses have passed away and one nurse has joined a monastery. We are currently trying to obtain contact details of the health workers who left the country from their relatives and/or friends using all available information from the 2004 questionnaire and the follow-up phone calls conducted in 2005 and 2006. Table A2: The original cohort study sample (2004) and the second wave sample (2007) DOCTORS NURSES First wave (2004) 90 219 First follow-up by phone (2005) 90 172 Second follow-up by phone (2006) 83 168 Second wave survey (2007) 88 177 Migrants 2 9 Passed away or joined a monastery 0 3 Follow-up rate 100% 86% We plan to interview the migrants over the phone or by email in the near future. If we include the health workers who left the country and those who passed away and joined the monastery, we get a follow-up rate of 100% for the doctors and of 86% per cent for the nurses. 37 Further efforts to contact missing health workers were made until December 2007. 63 ANNEX B JOB SEARCH The second wave survey contains information about the job search process the health workers went through after graduation. Only 14 doctors (16%) and 79 (45%) nurses participated in the national lottery after graduating. The most frequently reported reasons for not participating in the lottery is the will to personally find a job, independently from the government (52%), and the unwillingness to work in a region that they didn't like (14%). Half of the nurses stated that they were assigned to posts through regional quotas, suggesting that the mandatory allocation system administered by the regional governments in exchange for free education was still working in 2004.38 As further evidence that the allocation to posts through the lottery was already not binding in 2004 and 2005, we find that only half of the participants in the lottery decided to go to the health bureau of the region to which they had been allocated by the lottery, in order to be assigned to a specific post within that region. Moreover, one fourth of these health workers claimed that if they didn't like the post allocated to them by the regional health bureau, they could decline the offer. The majority of the doctors found their current job through advertisement in the media (37%), or through personal contacts in the facility (24%). An additional 12% personally applied to the competent regional health bureau and about 8% used personal contacts through relatives of the manager of the facility. 38 Doctors are funded by the federal government. Under the old allocation system, only students whose education was funded by the federal government were required to participate in the national lottery. The students whose education was funded by a regional government were allocated to a regional post by the health bureau of the region who funded their education. 64 Figure A1: How the health workers found their job advertisement in the media contacts through workers in the facility through the national lottery personally applied to the regional health bureau contacts through relatives of manager advertisement in public places transfer by assignment contacts through community, church etc. it is my clinic request of zonal heads the establishment sponsored by education 0 5 10 15 20 25 30 35 40 % NURSES DOCTORS The job search process followed by the nurses seems more diverse. Those who did not go through the national lottery and did not find their job through the media, found their job through advertisements in public places (7%), by applying to the regional health bureau (5%), or through personal contacts of various kind (about 13%). A small percentage of nurses were simply transferred (6%) or assigned (4%) to the specific facility where they currently work. 65 ANNEX C DIFFERENCES IN FACILITY INFRASTRUCTURE BETWEEN URBAN AND RURAL LOCATIONS The physical conditions of the facility where the health professionals work and the availability of resources for health care provision also differ significantly depending on whether the facility is located in urban or rural areas, and whether it is publicly or privately owned. Interviews with officials at the Ministry of Health during the design phase of the Second Wave survey suggested that one of the reasons why doctors do not want to work in rural areas is the impossibility for them to do the job they have studied and trained for. The lack of adequate infrastructures might impede a doctor working in a rural area from performing basic surgeries. This may generate frustration and dissatisfaction, especially among young doctors. Table A4 summarizes our findings with respect to a few key characteristics of public and private facilities located in urban and rural areas, as reported by the health professionals in our cohort. Table A4: Some characteristics of the health facilities Number of black-out with equipment for With with days per month (avg.) major surgeries laparology blood bank public sector in 6 22% 7% 7% rural area public sector in 3 66% 39% 33% urban area private sector in 8 21% 5% 5% rural area private sector in 3 61% 24% 12% urban area The poor quality of the rural health facilities is striking. Besides the higher number of days in which these facilities experience electrical blackout as compared to urban facilities, the rural facilities are also considerably less likely to be furnished with the necessary equipment for major surgeries, blood banks and laparology. 66 ANNEX D CONTINGENT VALUATION QUESTION FOR RURAL VERSUS URBAN JOB We designed the contingent valuation questions with the aim to generate a set of specific and realistic decision choices for the health professionals. We set the reference salary equal to the health professionals' current salary rather than a hypothetical base salary because, after running different versions of the questions during the pilot, we realized that referring to the current salary rendered the contingent valuation questions more realistic, with clear advantages in terms of the quality of the generated data. The question, which aims at revealing the reservation wage that would make a health worker accept a job in a rural area, was phrased as follows for the nurses: Imagine that you are offered two jobs as a health worker in the public sector, one in Addis Ababa and one in a rural area 500 km from Addis Ababa. Both contracts are for at least 3 years. Your monthly salary for the job in Addis Ababa would be equal to your current salary. Which job would you choose if... Your monthly salary for the rural job would be (current salary) Birr? Your monthly salary for the rural job would be (current+300) Birr? Your monthly salary for the rural job would be (current+900) Birr? Your monthly salary for the rural job would be (current+1200) Birr? Your monthly salary for the rural job would be (current+1500) Birr? Your monthly salary for the rural job would be (current+1800) Birr? The equivalent question for doctors was similar with as only difference that the salary increments were 500 Birr rather than 300 Birr.39 When a health worker expressed a preference for the job in Addis Ababa even at the highest rural salary ("current salary plus 1800 Birr" for nurses and "current salary plus 2,500 Birr" for doctors) the health worker was asked to write down the salary that make him accept a rural post. 39 In the pilot phase of the second wave of the survey we realized that increments of 300 Birr were informative for the nurses but not for the doctors. Therefore we decided to adopt slightly different contingent valuation framings for nurses and doctors. 67 In 2004 we also asked contingent valuation questions in order to identify the minimum salary at which the students were willing to switch from a job in Addis Ababa to a job in a rural area, either 200 Km or 500 Km away from Addis Ababa. The reference salary was then set equal to 700 Birr for the nurses and 1,400 Birr for the doctors; the salary increments were set equal to 100 Birr for both nursing and medical students, and the maximum possible salary was set to 1,200 Birr for the nurses and 2,100 Birr for the doctors. In 2004, if the health workers were unwilling to take a rural post even when they were offered maximum wages, we did not ask them to state the minimum wage they would require in order to accept the rural post, therefore the 2004 data are truncated at 1,200 Birr for nurses and 2,100 Birr for doctors. (for details see Serneels et al 2007 and Serneels et al 2004) 68 ANNEX E CONTINGENT VALUATION QUESTION FOR DOMESTIC JOB VERSUS MIGRATION We designed the contingent valuation questions with the aim to generate a set of specific and realistic decision choices for the health professionals. We set the reference salary equal to the health professionals' current salary rather than a hypothetical base salary because, after running different versions of the questions during the pilot, we realized that referring to the current salary rendered the contingent valuation questions more realistic, with clear advantages in terms of the quality of the generated data. In order to identify the reservation wages that would induce a nurse to stay in Ethiopia if he or she had a chance to move abroad, we presented the nurses with questions of the following form: Imagine that you are offered a job as a health worker in the public sector in Addis Ababa for life. Imagine that you are also given the possibility to emigrate to the US. However, you do not know whether you will be able to find a job abroad, what type of job, and how much you will earn. What would you choose if... Your monthly salary for the job in Addis would be (current salary +300) Birr? Your monthly salary for the job in Addis would be (current salary +900) Birr? Your monthly salary for the job in Addis would be (current salary +1200) Birr? Your monthly salary for the job in Addis would be (current salary +1500) Birr? Your monthly salary for the job in Addis would be (current salary +1800) Birr? Your monthly salary for the job in Addis would be (current salary +2100) Birr? We asked the same questions to the doctors, changing the salary increments to 500 Birr. We chose a public job in Addis Ababa as the reference job because we aim to investigate the health professionals' preference for migration when the opportunity cost of migrating is very high. While designing the questions we assumed that holding a public post in Addis Ababa represents the preferred job situation for most health professionals. Ex post survey data suggest that more than 50% of the doctors and 36% of the nurses would prefer to work in Addis Ababa in the long term. 69 This paper summarizes the findings from the second wave of a cohort study with health workers. The 90 doctors and 219 nurses were first interviewed in 2004 where they were in their last year at school, and were re-interviewed two years later when they had entered the labor market. The data allowed the authors to analyze (i) where they end up and how health workers are distributed; (ii) what their career preferences are and how they have changed; (iii) what is important in the choice between rural and urban areas; (iv) what drives the likelihood to migrate abroad. This paper was produced by the World Bank's Africa Region Human Resources for Health team, with funding from the Government of Norway and the Gates Foundation. 2009 © All Rights Reserved. Health Systems for Outcomes Publication THE WORLD BANK