Level and determinants of incentives for village midwives in Indonesia

10
Level and determinants of incentives for village midwives in Indonesia Tim Ensor, 1 * Zahid Quayyum, 2 Mardiati Nadjib 3 and Purwa Sucahya 4 Accepted 9 September 2008 Since the early 1990s Indonesia has attempted to increase the level of skilled attendance at birth by placing rural midwives in every village in an effort to reduce persistently high levels of maternal mortality. Yet evidence suggests that there remains insufficient incentive to ensure an equal distribution across areas while the poor in all areas continue to access skilled attendance much less than those in richer groups. We report on a survey that was conducted as part of a complex evaluation of the rural midwife programme in Banten Province, to better understand the effect of financial incentives on the distribution of midwives and use of services. Midwives obtain almost two-thirds of their income from private clinical practice. Private income is strongly associated with competence and experience. Multivariate analysis suggests that midwives are well able to earn a substantial private income even in remoter areas. Yet the study also found a high level of unwillingness to move posts to a more remote area for a variety of non- financial reasons. The results suggest that the access to skilled attendance of those unable to afford fees may be impaired by the dependence on fee income, a result supported by companion household studies. In addition, ensuring that staff live and work in remoter areas is only likely to be financially sustainable if midwives can be attracted to live in these areas early in their careers. Finally, the overall strategy of basing skilled attendance mainly on village services throughout the country may need to be re-visited, with alternative models offered in areas where it continues to be impractical even with a change in the incentive framework. Keywords Village midwives, incentives, Indonesia, maternal care, poverty, health work- force, human resources for health KEY MESSAGES Rural midwives in Indonesia earn substantial income from private practice. This needs to be replaced by the public sector if they are to properly serve those who cannot afford to pay for services. Ensuring that midwives are available in rural areas requires both additional funding for guaranteed salaries and also development of career pathways that attract staff to these areas at an early stage in their careers. Women are well able to differentiate between good and bad midwives, perhaps using knowledge gained from informal community networks. * Corresponding author. Immpact, University of Aberdeen, Health Sciences Building, Foresterhill, Aberdeen, AB25 2ZD, UK. Tel: þ 44 1904 633280. E-mail: [email protected] 1 Immpact, University of Aberdeen, Health Sciences Building, Foresterhill, Aberdeen, AB25 2ZD, UK. 2 Immpact, University of Aberdeen, Health Sciences Building, Foresterhill, Aberdeen, AB25 2ZD, UK. Tel: þ44 (0)122 455 1844. E-mail: [email protected] 3 Department of Public Health, University of Indonesia, Indonesia. E-mail: [email protected] 4 Department of Public Health, University of Indonesia, Indonesia. E-mail: [email protected] Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine ß The Author 2008; all rights reserved. Health Policy and Planning 2008;1–10 doi:10.1093/heapol/czn040 1 Health Policy and Planning Advance Access published November 20, 2008

Transcript of Level and determinants of incentives for village midwives in Indonesia

Level and determinants of incentives for villagemidwives in IndonesiaTim Ensor,1* Zahid Quayyum,2 Mardiati Nadjib3 and Purwa Sucahya4

Accepted 9 September 2008

Since the early 1990s Indonesia has attempted to increase the level of skilled

attendance at birth by placing rural midwives in every village in an effort to

reduce persistently high levels of maternal mortality. Yet evidence suggests that

there remains insufficient incentive to ensure an equal distribution across areas

while the poor in all areas continue to access skilled attendance much less than

those in richer groups. We report on a survey that was conducted as part of a

complex evaluation of the rural midwife programme in Banten Province, to better

understand the effect of financial incentives on the distribution of midwives and

use of services. Midwives obtain almost two-thirds of their income from private

clinical practice. Private income is strongly associated with competence and

experience. Multivariate analysis suggests that midwives are well able to earn a

substantial private income even in remoter areas. Yet the study also found a high

level of unwillingness to move posts to a more remote area for a variety of non-

financial reasons. The results suggest that the access to skilled attendance of those

unable to afford fees may be impaired by the dependence on fee income, a result

supported by companion household studies. In addition, ensuring that staff live

and work in remoter areas is only likely to be financially sustainable if midwives

can be attracted to live in these areas early in their careers. Finally, the overall

strategy of basing skilled attendance mainly on village services throughout the

country may need to be re-visited, with alternative models offered in areas where

it continues to be impractical even with a change in the incentive framework.

Keywords Village midwives, incentives, Indonesia, maternal care, poverty, health work-

force, human resources for health

KEY MESSAGES

� Rural midwives in Indonesia earn substantial income from private practice. This needs to be replaced by the public

sector if they are to properly serve those who cannot afford to pay for services.

� Ensuring that midwives are available in rural areas requires both additional funding for guaranteed salaries and also

development of career pathways that attract staff to these areas at an early stage in their careers.

� Women are well able to differentiate between good and bad midwives, perhaps using knowledge gained from informal

community networks.

* Corresponding author. Immpact, University of Aberdeen, Health SciencesBuilding, Foresterhill, Aberdeen, AB25 2ZD, UK. Tel: þ 44 1904 633280.E-mail: [email protected]

1 Immpact, University of Aberdeen, Health Sciences Building, Foresterhill,Aberdeen, AB25 2ZD, UK.

2 Immpact, University of Aberdeen, Health Sciences Building,Foresterhill, Aberdeen, AB25 2ZD, UK. Tel: þ44 (0)122 455 1844.E-mail: [email protected]

3 Department of Public Health, University of Indonesia, Indonesia.E-mail: [email protected]

4 Department of Public Health, University of Indonesia, Indonesia.E-mail: [email protected]

Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine

� The Author 2008; all rights reserved.

Health Policy and Planning 2008;1–10

doi:10.1093/heapol/czn040

1

Health Policy and Planning Advance Access published November 20, 2008

IntroductionEver since the early 1990s Indonesia has attempted to increase

the level of skilled attendance at time of delivery through a

policy of placing rural midwives (Bidan di Desa) in every village

in an effort to reduce persistently high levels of maternal

mortality (Shiffman 2003). Rural midwives are given a 1 year

diploma training (compared with the longer 3 year training

common for hospital midwives) and operate as multi-purpose

providers of health services with a particular focus on

pregnancy, delivery and post-partum care. Midwives largely

work in the community, but are each attached to a health

centre (at sub-district level) where they work for a few hours a

week, largely providing care for pregnant women and children.

They serve one or more villages, a total of 3500 people, and

perform perhaps 36 deliveries per year (Makowiecka et al.

2007). According to one study, maternity services make up

more than 50% of the workload of a midwife, but this varies

from as low as 19% up to 80% (Ensor et al., forthcoming). The

policy has been successful in training large numbers of mid-

wives, and deliveries with a professional midwife or doctor have

risen across the country to around 50% (from around 35% in

the late 1980s) (Hatt et al. 2007). Yet questions continue to be

raised about whether the policy is having a substantial impact

on use of skilled delivery and on maternal health outcomes

in the most remote, rural communities, which was the main

original focus of the programme. A particular concern is

whether the incentives to work in remote areas are sufficient

to encourage midwives to locate there and provide services to

the most vulnerable groups.

Providing adequate incentive for public workers to deliver

health services in the public sector, particularly in rural areas

and to vulnerable groups, has policy significance in many low-

and middle-income countries. Much of the literature focuses on

the incentives to medical doctors, focusing on whether it is

beneficial to the public sector that doctors earn incomes from

dual clinical practice, i.e. private practice undertaken by

workers employed in the public sector. Research has examined

the benefits and problems involved in permitting dual clinical

practice, worker absenteeism and unofficial payments (Hammer

and Jack 2002; Berman and Cuizon 2004; Chaudhury and

Hammer 2004; Ensor and Thompson 2006). Another strand of

research looks at the balance of financial and non-financial

factors that motivate practitioners (Stilwell 2001; Dieleman

et al. 2003). In Indonesia private dual practice is ubiquitous and

accepted by government as a way of ensuring that practitioners

can derive a satisfactory income. The main issue is whether

midwives are provided with adequate incentive—financial or

non-financial—for them to work in the poorest areas and serve

the poorest women. Despite the original intention of the policy

to service rural areas, it is increasingly apparent that midwives,

particularly those that are more experienced, prefer to move to

urban areas to practice. A companion study found, for example,

that while in urban areas around 10% of communities had no

midwife resident, in remote areas this figure rose to more than

60% (Makowiecka et al. 2007).

As part of a complex evaluation of the Village Midwife

programme undertaken by Immpact,1 we report on the results

of a health worker incentive survey that examined the

composition and determinants of both public and private

income of rural midwives. The paper will focus on the extent

to which the incentives faced by rural midwives support

government objectives to deliver maternal health services to

remote, rural areas.

MethodsThe study was carried out in Banten province, a new province

on the island of Java and part of West Java Province until 2000.

Banten is divided into four districts and three cities (kota).

Two districts—Serang and Pandeglang—were selected by the

Ministry of Health for the study based on their relatively

inferior health outcomes: Banten has an infant mortality rate of

around 55, substantially higher than the national average of 28,

whilst public spending per capita on health is amongst the

lowest in the country (Somanathan et al. 2004). Serang and

Pandeglang have slightly different profiles. Serang is more

geographically concentrated, with a higher proportion of the

population living in urban areas. Pandeglang is larger, with

more remote areas and a very low population density in the

coastal areas in the south which merge into a national park.

The incentive survey was included as a part of a larger costing

study of the rural midwife scheme (see Quayyum et al. 2007a).

A stratified random sample of 207 midwives (out of a total of

737) providing services in 227 villages (out of a total of 711)

was obtained. Stratification was based on the distance and type

of midwife assignment. Distance was defined as the distance

between the midwife’s main place of work and the nearest

district hospital in kilometres. This was based on information

from the District Health Office (DHO), ratified using local

maps. Three categories of midwife assignment were defined:

where a midwife resides in a village and has sole responsibility

only for that village; where the midwife is responsible for one

village but does not reside there; and where a midwife is

responsible for more than one village. Characteristics of the

midwives, including responsibility, were collected during a

previous census of midwives undertaken as part of the same

research project.

The survey was conducted using a structured questionnaire

originally developed in English, translated into Bahasa and

then retranslated to check for any distortions in the meaning of

the question. The questionnaire was piloted in six villages

during July 2005 and the main data collection took place from

August to October 2005. Interviews were conducted, as far

as possible, so as not to disrupt the working routine of

respondents. Data were double entered in Epi-Info. Data

cleaning focused on outliers. These were identified by examin-

ing the distribution of income from different sources and

examining those where the deviation from the third quartile

was more than 150% of the interquartile range. These

observations were then re-checked first for accuracy in data

entry and then by going back to interviewers, and ultimately

respondents, for clarification.

After allowing for non-responses, data were obtained on

incomes from all sources for 190 midwives, including 128

village midwives (bidan di desa), 46 health centre midwives

(Puskesmus) and 16 other midwives. A series of 17 questions

were asked to obtain information on different types of income,

including sources and size of typical monthly official public

2 HEALTH POLICY AND PLANNING

income (main salary and miscellaneous sources such as

per training per diems), and private income from clinical

and non-clinical sources. Information was also sought on

income received from the State Insurance Company (Askes).

This was a system established in 2005 which pays midwives

for providing services, including deliveries for the small

proportion of the population, around 8%, that are registered

as poor because their household income falls below the official

poverty line.

Most questions were closed and quantitative, although an

open question at the end aimed to pick up other sources of

income. Information on reasons for becoming and remaining

midwives were collected using a list of closed questions,

developed at the design stage and from open responses

obtained during piloting. The final questionnaire also permitted

‘other’ open answers at the end which were subsequently

coded.

Instruments were administered by interviewers from outside

the area who were not currently working in the health

sector. Many were students from the University of Indonesia.

In order to permit analysis by distance, experience and contract

status, survey results were combined with data from another

survey on midwives and information available on area

characteristics from the District Health Office (Makowiecka

et al. 2007).

Data were first analysed using summary statistics and cross-

tabulations to provide descriptive information on composition

and size of average incomes by type of midwife.

In order to investigate the interactions and determinants

of the income of midwives in more depth, a series of

multivariate models were estimated that describe the income

of a midwife (Y). Private income is likely to be determined

principally by the number of services sold, vi, and the price

charged for each service, pi. A reduced form equation was

derived by assuming that price and quantity are determined by

the (perceived) quality of service offered, qi, local market

variables that influence the degree of competition in providing

services similar to those delivered by the rural midwife

(from other midwives, doctors, clinics, traditional birth atten-

dants), mi, and strength of local demand, xi.

Yi ¼ vi pi ¼ f ðqi;mi; xiÞ

The data set provides a series of variables that can be used

to proxy these main dimensions (Table 1). Quality is partly

proxied by the experience of the midwife. In addition, a

knowledge score was available based on an instrument

completed by all midwives working in the two districts,

which provides a direct measurement of technical knowledge.

This might be expected to translate into service (output)

quality. We also included an interaction term between

experience and knowledge (score) on the basis that more

experienced midwives are likely to be more effective in

converting extensive knowledge into improved quality.

Several measures of market competition (m) are available.

Numbers of alternative providers of delivery services are

available from DHO records. In addition, distance to the district

centre can be regarded as a measure of how easy it is to access

hospital services. Demand for services (x) is likely to be highly

influenced by the size of the local economy, both through

individual socio-economic characteristics and the overall size of

the market reflecting ability to pay for private health services.

Although there is no direct measure of socio-economic status

available, sub-district data on the proportion of people in

poverty can be used as a proxy for individual socio-economic

circumstances, while population and population density provide

a measure of the size of the economy. A district dummy is also

included given the generally higher socio-economic status of

the population of Serang and the differences in geographic

characteristics.

Public income is composed of basic salary, top-up allowances

and public income from the insurance scheme. Basic public

salary is likely to be determined mainly by contract status and

experience. Market variables (m) and quality variables (q) are

likely only to affect public insurance (Askes) income which is

determined by a woman’s choice to use a midwife and so is

workload related. The equation for public income is thus similar

Table 1 Indicators and proxy-indicators for income equations

Variable Indicator or proxy indicator

q Number of years since qualification (midwife experience 8-10¼ 1 for 8-10 years as midwife, midwife experience 10-14¼ 1 for 10-14 years,midwife experience >14¼ 1 for more than 14 years; otherwise zero)

Competency score (score based on knowledge test normalized between 0 and 100)

Contract status:

Central PTT Contract (1 if hold central contract, otherwise 0)

Local PTT Contract (1 if hold local contract, otherwise 0)

m Doctors (Doctors per 10 000; number by sub-district divided by population)

Midwives (Midwives per 10 000; number by sub-district divided by population)

TBAs in area (TBAs per 10 000; number by sub-district divided by population)

Distance from nearest hospital (hospital_distance 15-30 km¼ 1, between 15 and 30km from district centre; hospital_distance 30-60 km¼ 1,30-60 km; hospital_distance > 60 km¼ 1, greater than 60km)

x Population (population in 2004)

Population density (population density; population of sub-district divided by area in square kilometres)

Proportion of poor in each sub-district (proportion poor; proportion of the adult population)

Location (Serang; Serang district¼ 1, Pandeglang¼ 0)

INCENTIVES FOR VILLAGE MIDWIVES IN INDONESIA 3

to that for private income but we expect, a priori, that the effect

of quality (other than experience and contract) and demand

variables will be weaker than in the private income equation

given that income is largely determined by experience rather

than performance.

ResultsComposition of incomes by type of midwife(bivariate results)

Midwives obtain income from a variety of sources. Public

income is composed of basic salary and top-up allowances. Top-

up allowances combine both festival bonuses paid to everyone

and a very small performance bonus that is assessed at an

individual level. Since the latter only adds at most a few

percent to public income, it is unlikely to have much impact on

motivation. Payment from Askes is channelled through the

local health centre (Puskesmus). In addition to their public

income, most midwives (93% in the survey with no significant

difference between urban and rural areas) also earn income

from private clinical practice. Midwives are providers of both

maternal and other services such as child immunization, family

planning and adult services. Many also undertake private non-

clinical activities that, in some cases, provide an important

supplementary source of income.

The study found some differences in the total public income

of those on different contracts (Table 2, mean and median

public income with 95% confidence intervals).2 The most

coveted position (representing more than 57% of the sample)

is to be employed as a civil servant (PNS) ensuring a

permanent position with pension. Midwives in permanent

PNS posts reported annual mean total public income of

around $1768.3 Central limited term contracts (PTT), financed

by the Ministry of Health, or district government contracts

(TKK) are held by the remainder. For many, this status is seen

as a stepping stone to a permanent civil service appointment.

Those on PTT contracts reported annual mean public income of

$1164, while those on local TKK contracts reported annual

mean public income of $866.

Substantial income is derived from private earnings from

clinical practice, gifts from patients, and income from other

private activities such as farming, domestic work and cooking

for other people. In total, estimates suggest that midwives earn

around US$364 per month or US$4368 per year (median $3113).

Almost 60% of total income is derived from private income

from clinical practice, while other private sources contribute

around 8% of total income (Figure 1). It should also be noted

that there is no clear distinction between private and public

clinical practice. Midwives are employed by the state to deliver

services but can, with the exception of services provided at

health centres, also charge private fees for those same services.

There is substantial variation in income from private clinical

practice, both maternal and non-maternal, mostly reflecting

large differences in private clinical income earned by different

midwives. While mean private clinical income is $2508 per year

(median $1263), the top 10% of midwives earn, on average,

more than $11 000 from private activities.

Levels of income appear to be closely related to the amount

of experience a midwife has, represented by the number of

years since qualification (Figure 2). The study suggests that a

midwife with more than 15 years’ experience earns more than

twice the income of a midwife with less than 5 years’

experience.4 The share of private practice income appears to

increase with experience so that for very experienced midwives

it accounts for more than two-thirds of total income.

There is no clear cut bi-variate relationship between total

income and distance from the main centres of population

(defined as distance between the village of main work and the

district hospital). It is true that incomes are, on average, lower

for those living more than 60 km from the centre than for those

living within 15 km of the centre, but for distances in between,

the relationship is more ambiguous (Figure 3).5

The survey investigated reasons for becoming and remaining

a midwife (Table 3). These reasons were classified into four

main groups. The responses found a fairly even split between

those becoming a midwife for reasons of career, family

convenience—proximity of relatives, husband’s place of work

35%

58%

7%

Public

Private clinical

Private non-clincal

Figure 1 Structure of midwife’s annual total income

Table 2 Structure of annual public income of midwives by contracttype (US dollars)

PNS PTT TKK All midwives

Mean income (US$)

Askes 166 149 78 157

Salary 1442 903 679 1232

Top-up allowances 160 112 109 141

Total 1768 1164 866 1530

Confidence interval (95%) þ/� 290 142 537 188

Median income (US$)

Askes 126 102 82 126

Salary 1379 758 632 1137

Top-up allowances 155 100 109 132

Total 1660 960 823 1395

PNS¼ civil servant permanent position with pension; PTT¼ central limited

term contract, financed by Ministry of Health; TKK¼ district government

contract.

4 HEALTH POLICY AND PLANNING

and children’s school—and service to the community. The

importance of parents in decision-making is also suggested,

with 12% (n¼ 24) mentioning parental wishes or advice as the

most important motivation. The main reasons for remaining as

a midwife, rather than leaving and practicing privately or

entering another profession, are career protection and advance-

ment (good salary, guaranteed pension, good promotion pros-

pects), with 71% citing this as the main motivator. A desire to

serve the community appears to be less important. Answers did

not differ significantly by location. There were, however,

significant differences by experience (P < 0.01): 30% of mid-

wives with more than 14 years experience cited family and 7%

career reasons, while 46% of midwives with fewer than 8 years

experience cited career and 8% family reasons. It is unclear,

however, whether these differences can be attributed to a

genuine difference in approach to the job or a difference in the

way older respondents recalled reasons for past decisions.

Respondents were generally reluctant to move their village

of responsibility largely because of family reasons (Figure 4).

While it is difficult to be sure whether ‘family reasons’ mask

other more sensitive issues such as missing out on promotion

or a better salary should they move, it should be noted that

these answers were true of midwives that already have received

promotion. In fact it was the younger midwives, who may have

more to gain from promotion in their setting, who were more

willing to move when asked. Furthermore, information on the

midwives indeed confirms that many have established family

settings that they would be reluctant to give up. Fewer than

37% said they were willing to move to a village 1 hour away

and just 12% to a village 5 hours away. Those who were willing

to move said they would do so out of a sense of duty to the

profession or the community, although a few suggested that it

could lead to promotion or to securing civil service status.

Interestingly, very few midwives (four in the case of the 1 hour

move, three in the case of the 5 hour move) cited disturbance

to their private practice as a reason for not moving.

The reluctance to move to a more remote area was also

reflected in the very high payments most midwives demanded

in order to move, which often exceeded total incomes from all

sources (public and private). On average, midwives, initially

reluctant to move, suggested that to move 1 hour away from

their current location they would require their public salary

to double, while for a 5 hour move their salary would have to

triple. Figures derived from hypothetical scenarios should

always be treated with caution, as individuals may often

exaggerate values placed on preferences or may simply

misunderstand the question. The rather high figures placed

on a move may well suggest exaggeration of the figures.

Nevertheless, it is also true that asking a midwife to move her

location of work, even by 1 hour, is likely to necessitate far

more inconvenience and financial loss than is expressed by the

value of their private practice alone. Most midwives are

married, and moving place of work may necessitate a long

journey, or living separately from their spouse, or moving

residence and, possibly, the spouse moving way from his

employment. All of this implies costs to the midwife and her

family. These results echo an earlier study on incentives for

doctors to practice in rural Indonesia, which also found that for

many doctors the necessary payment exceeded anything that

was realistically affordable (Chomitz et al. 1998).

Multivariate analysis of income

Multivariate analysis based on the model described under

‘Methods’ above was undertaken to explore the association

$0

$1000

$2000

$3000

$4000

$5000

$6000

$7000

<8 years 8–10 years 10–14 years >14 years

Experience of midwife

Total publicincome

Total income

An

nu

al in

com

e (U

S d

olla

rs)

Figure 2 Variation in gross income by experience of midwifeNote: Public income¼ salary, top-up allowancesþAskes income.All income¼ public incomeþ private clinical and non-clinical income.

$0

$1000

$2000

$3000

$4000

$5000

$6000

< 15 km 15–30 km 30–60 km > 60 km Total

Distance from nearest district hospital

Private non-clinical

Private clinical

Public

An

nu

al in

com

e (U

S d

olla

rs)

Figure 3 Income (all sources) by distance from district hospital

Table 3 Reasons for becoming and remaining avillage midwife

Becoming amidwife

Remaining amidwife

No response 10% 8%

Career/income 21% 71%

Family 25% 6%

Community 27% 11%

Parental 12% 0%

Other 6% 4%

Total 100% 100%

INCENTIVES FOR VILLAGE MIDWIVES IN INDONESIA 5

between income and the various measures of market size and

quality of service. It was hypothesized that the characteristics of

the market could vary quite markedly in urban, rural and

remote areas, and between districts, affecting not only the

constant but also the slope coefficients. To test for this, separate

regressions were run for these sub-groups and compared with

the joint specification (pooled data) using a Chow test. No

significant differences between specifications were revealed,

however, so the pooled estimates were retained. Specifications

were estimated for total income and then separately for private

clinical income and public income. Natural logs of the depen-

dent variable and continuous independent variables were used.

This largely corrected any skewness in the variables.

The specification was tested for heteroscedasticity (non-

constant variance of residuals) and omitted variable bias but

these tests proved insignificant. A Ramsey test, which includes

powers of fitted variables as explanatory variables to test for

evidence of omitted variables, found no evidence of significant

bias (P > 0.17).

Proxy measures of quality exhibit strong associations with

income (Table 4). Both the knowledge and experience score

were significant in the private income equation. So, for

example, a midwife with more than 14 years experience with

a median knowledge score (assuming other variables such as

distance are held at their median values) might earn more than

three times the private income of a colleague with less than

8 years experience (Figure 5).

Civil service status does not appear to be an important

positive determinant of private income once the effect of

experience is taken into account. Midwives with PTT contracts

appear to earn more than their civil servant counterparts,

perhaps because they compensate for lower and less stable

public incomes. It should be noted, however, that the presence

of multicollinearity (correlation between contract and experi-

ence, P < 0.05) may mean that the effect of contract is reflected

in the experience variables.6

The impact of location is important. Private incomes

are significantly higher in Serang. Incomes also increase

progressively in villages that are further away from the district

centre. Although there are differences in socio-economic status

between the districts, part of which is controlled for by the

inclusion of poverty rates, the most striking difference is that

Serang is more urbanized with far fewer remote villages. We

believe, therefore, that the Serang dummy effect captures some

part of the distance effect on private incomes. Since, in general,

villages further away from the centre are poorer than those that

are closer, this effect is likely to reflect a local market effect

where the distance from hospitals and other formal providers

makes local services provided by midwives more attractive.

A competition effect can be detected for doctors—suggested by

the negative coefficient—but not for other categories of medical

personnel, but it is not significant. Other combinations of staff

including a joint professional staff variable (doctors, midwives

and nurses) were introduced into the equation but the

coefficients remained non-significant (P > 0.1).

The PTT contract appears to have a positive but not significant

(P > 0.15) effect. The lack of a positive effect of PNS contract on

private income, which was evident in the bivariate results, is

likely to be because PNS midwives will also generally be more

experienced and, assuming selection procedures are approxi-

mately merit based, more knowledgeable. When these variables

are included, any additional contract effect appears to be

negated.

In contrast to the private income equation, the contract

variables demonstrate a significant association with public

income (Table 5). PTT local has the greatest negative impact on

income compared with a PNS contract. Much of the explana-

tory power of the regression is in the constant term, although

experience also has a significant, if modest, impact on public

income.

The distance variables show a significant association

with public income, with those living closer to an urban

area having higher incomes. This is likely to be the effect

of more experienced midwives having choice over where they

live and demonstrating a preference for areas closer to the

centre.

Yes 2512%

Community 13Career/PNS 3

Yes 77 Other 937%

Community 41Career/PNS 13Other 23 No 52

25%Family 22Private practice 3Other 27

No 12963%

Family 65 No 129 + 52 = 181Distance 23 88%Private practice 4Other 37

Willing to move(5 hours)

Willing tomove (1 hour)

Figure 4 Willingness to accept a move to a more remote location

6 HEALTH POLICY AND PLANNING

Discussion of limitations and policyimplicationsStudy limitations

The results from a study of this type are necessarily limited by

the type of data and method of collection. Some of the

questions are sensitive and this may lead to misreporting.

In many contexts, obtaining information on income can be

problematic. In the Indonesian context it helps considerably

that dual clinical practice is positively encouraged rather than

outlawed. The data collected on motivations and willingness to

move to other areas of the country is likely to suffer from

misreporting. Some of the questions rely on past memories of

choices made. They also ask women to be honest about their

motivations. That so many midwives were willing to express an

interest in their own career and family rather than the wider

community is perhaps evidence that this bias was minimized.

The study raises a number of questions that are difficult to

answer with the available data. The most important relates to

the type of patients served. While we suspect from the evidence

on incentives and also information from other studies, that the

poor make less use of services because midwives prefer to serve

fee-paying patients, this study alone does not provide adequate

information on the type of patients. Although we questioned

midwives about the type of patients served, most vaguely

pointed towards serving vulnerable and the poor but no

concrete information was available on the precise level of

poverty—in some areas the entire population can be classified

as poor against the incomes of professional urban dwellers.

The priority for policy is to focus on the relative poor within

these communities.

The results suggest that a village midwife position within the

public sector, particularly with civil service status, is a highly

valued one. Midwives do not generally want to move fully into

private practice and prefer instead to earn income from

Table 4 Determinants of private midwife income

Coef. Std. Err. t P > t [95% Conf. Interval]

Serang 0.538567 0.219156 2.46 0.015 0.105691 0.971442

Experience 8-10 0.250015 0.280583 0.89 0.374 �0.30419 0.804219

Experience 10-14 0.501778 0.277438 1.81 0.072 �0.04621 1.049769

Experience >14 1.35458 0.337741 4.01 0 0.687477 2.021682

Ln Competency score 1.550169 0.736291 2.11 0.037 0.095854 3.004483

Hospital distance 15-30 kms 0.32689 0.230807 1.42 0.159 �0.129 0.782776

Hospital distance 30-60 kms 0.504299 0.260494 1.94 0.055 �0.01023 1.018824

Hospital distance >60 kms 0.773377 0.359143 2.15 0.033 0.064002 1.482752

Doctors per 10 000 �0.11363 0.093638 �1.21 0.227 �0.29858 0.071325

Midwives per 10 000 0.030734 0.030683 1 0.318 �0.02987 0.091338

TBAs per 10 000 �0.00895 0.009847 �0.91 0.365 �0.0284 0.010496

Central PTT Contract 0.099778 0.201066 0.5 0.62 �0.29737 0.49692

Local PTT Contract l 0.926736 0.629006 1.47 0.143 �0.31567 2.169142

Proportion poor 0.564058 2.983092 0.19 0.85 �5.32811 6.45623

Population density �0.00013 0.000089 �1.48 0.14 �0.00031 4.37E-05

Population in 2004 0.000095 5.58E-05 1.7 0.091 �1.5E-05 0.000205

Constant �0.74262 3.172745 �0.23 0.815 �7.0094 5.524147

Number of obs ¼ 174

F(16, 157)¼ 3.13

Prob > F¼ 0.0001

R-squared¼ 0.2416

Adj R-squared¼ 0.1643

Root MSE¼ 1.0577

-

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

4,500

50 55 60 65 70 75 80 85 90

Knowledge score

US

Do

llars

per

yea

r

< 8 years 8 to 10 years 10 to 14 years > 14 years

Lowest 10% Median score Upper 10%

Figure 5 Influence of experience and knowledge on annual privateincome of midwivesNote: All other variables are held constant at their median values.

INCENTIVES FOR VILLAGE MIDWIVES IN INDONESIA 7

charging users as part of their public health service role. This is

unsurprising given that the evidence indicates it is possible to

earn substantial private fees even in remote areas.

Discussion of implications

The results presented accord reasonably well with the

salary figures reported in a study of financing flows which

found an average of $1550 for PNS and $936 for PTT (Ensor

et al., forthcoming). In common with earlier studies, the survey

found that the income of midwives is substantially supple-

mented from other sources (Nadjib et al. 2004).

The effect of distance in the multivariate analysis is some-

what surprising, with private incomes rising further from the

district centre. A possible explanation is the difference in

market (demand) conditions for the services of midwives. Close

to the district centre, women have many more choices both

because they are able to travel to a hospital for delivery and

because there is a higher concentration of practitioners offering

delivery services. Further from the district centre, competition

declines, making it easier to develop a successful practice.

An examination of the determinants of private income

indicates that proxy measures of midwife quality correlate

highly with income. This suggests that patients select midwives

that are likely to provide good standards of service. Greater

demand for the services of more experienced midwives is

evidenced both in higher volumes of service (data suggest that

more patients are seen per week) and higher charges for each

service rendered. How consumers obtain this information

is unclear, although unofficial networks of opinion on the

quality of health providers have been shown elsewhere to

operate and to provide quite reliable levels of information on

quality (e.g. Tanzania, see Leonard 2002). It is likely that such

networks operate here.

The findings have important implications for the extent to

which the village midwife system operates to deliver services to

poor populations living in rural areas. The substantial depen-

dence on private income may ensure that midwives remain in

public service and may possibly establish a serious potential

barrier to the poor accessing skilled maternity services. This is

confirmed by evidence from a companion study of households

in the same districts, which found that the use of skilled

attendance amongst the richest quintile is more than eight

times the use amongst the poorest (Achadi et al. 2007;

Quayyum et al. 2007b). A cost analysis shows that even

though public resources for maternal health are distributed

quite evenly to sub-districts, the poorest 20% only benefits from

an estimated 10% of spending (Ensor et al. 2006).

A core issue for public policy is how to develop incentives

which ensure that midwives are attracted to remote areas and

deliver services to poorer women. The study reveals a strong

reluctance amongst midwives to work in a more remote area.

It is tempting to conclude that the reason for this reluctance is

that it is more difficult to earn a private income in these areas.

Yet the evidence indicates otherwise, with an unwillingness to

move strongly associated with existing ties to the community

mainly related to family—parents, children’s schools and

Table 5 Determinants of public midwife income

Coef. Std. Err. t P > t [95% Conf. Interval]

Serang 0.023969 0.0611 0.39 0.695 -0.09667 0.144613

Experience 8-10 �0.04927 0.076388 �0.65 0.52 �0.2001 0.101558

Experience 10-14 0.075519 0.074731 1.01 0.314 �0.07204 0.223079

Experience >14 0.264224 0.092532 2.86 0.005 0.081516 0.446932

Ln Competency score 0.01739 0.203153 0.09 0.932 �0.38374 0.418522

Hospital distance 15-30 kms 0.153975 0.064035 2.4 0.017 0.027536 0.280414

Hospital distance 30-60 kms 0.16668 0.073323 2.27 0.024 0.021901 0.31146

Hospital distance >60 kms �0.0041 0.101643 �0.04 0.968 �0.2048 0.196596

Doctors per 10 000 0.033965 0.02628 1.29 0.198 �0.01793 0.085856

Midwives per 10 000 �0.00737 0.008681 �0.85 0.397 �0.02451 0.009775

TBAs per 10 000 0.00308 0.002784 1.11 0.27 �0.00242 0.008577

Central PTT Contract �0.30851 0.055795 �5.53 0 �0.41868 �0.19834

Local PTT Contract l �0.7169 0.158849 �4.51 0 �1.03056 �0.40325

Proportion poor �0.09247 0.850238 �0.11 0.914 �1.7713 1.586352

Population density 3.63E-05 2.51E-05 1.45 0.149 �1.3E-05 8.58E-05

Population in 2004 �3.55E-06 1.56E-05 �0.23 0.82 �3.4E-05 2.73E-05

Constant 6.954366 0.867775 8.01 0 5.240914 8.667817

Number of obs¼ 181

F(16, 164)¼ 8.25

Prob > F¼ 0

R-squared¼ 0.4461

Adj R-squared¼ 0.392

Root MSE¼ 0.30264

8 HEALTH POLICY AND PLANNING

husband’s place of work. This is likely to mean that willingness

to work in a remote area is more likely amongst those midwives

that have yet to settle into a community. In fact, adjusting for

other confounding factors, distance has a significant positive

association with private income, probably because of the lack of

competition from health facilities. This effect is not captured by

the numbers of doctors and nurses which summarize the

competition from individual practitioners. Midwives in general

do not cite loss of private income or practice as a reason not to

move to a more remote location. Rather they mention family

reasons including proximity to their husband’s place of work,

children’s school and parents home.

Once in an area, a greater focus on the incentive to deliver

services to the poor is required. Recently, the Indonesian

government has begun to address this barrier by developing a

system of insurance for the poor, financed from the proceeds

of a reduction in the fuel subsidy and administered by a

government-owned health insurance company, Askes. This

provides a payment to midwives for providing delivery care to

poor women. Yet there are concerns that the incentive may not

be sufficient to compensate fully for lost private income and

that only a relatively small proportion of the population qualify

for the health card for the poor which permits them to gain

benefits from the insurance. So far, little increase in skilled

attendance has been observed, although the scheme is still at

an early stage of development.

These two concerns suggest that the focus of incentive policy

should be on developing contracts for yet-to-be settled mid-

wives (settled midwives are too expensive to move) which

encourage them to deliver services to the poor.7 New midwives

are usually employed initially as contracted midwives (non civil

servants) and progression to civil service status often requires a

move to facilities situated nearer the district centre. Developing

a longer term contract, perhaps with a promise of civil service

status, for younger, less experienced midwives in return for

working in remote areas might be one way of ensuring that

midwives are tied earlier to more remote areas. Such a contract

is likely to be more potent if midwives are recruited from those

areas. At the same time, measures to ensure that practitioners

are compensated by bridging the gap between public and

private income are required. One way of doing this is to pay a

substantive supplement for those working in high poverty, low

population density areas, perhaps based on the number of poor

living in the area. Such a payment would be in addition to any

service fee allocated through Askes.

Further work will be required to assess the level of incentive

required to provide sufficient enticement to deliver services

in remote areas. The doubling or tripling in salaries suggested

by midwives as a condition for moving their place of work

are substantial and may well be exaggerated for several

reasons. Nevertheless, the resistance to working in remote

areas suggests that the increment in income will need to be

substantial in order to ensure that midwives are retained in

remote areas and serve target groups.

Motivating frontline public health workers is a key policy

issue in many low income countries. So often the thrust of

policy is to ensure that sufficient workers are trained rather

than ensuring that they are retained in the public sector and

motivated to deliver services. Yet so often these key workers are

absent when tracking studies are undertaken (Reinnikka and

Smith 2004). Although many countries, including Indonesia,

claim to pay rural workers allowances for working in remote

areas, these increments are often derisory. This study suggests

that retaining and motivating workers in priority areas requires

substantial a benefit of substantial value—in cash or in kind.

This issue remains an important policy concern in Indonesia

and elsewhere.

AcknowledgementsThis work was undertaken as part of an international

research programme—Immpact (see: http://www.abdn.ac.uk/

immpact)—funded by the Bill & Melinda Gates Foundation,

the Department for International Development, the European

Commission and USAID. The funders have no responsibility for

the information provided or views expressed in this paper. The

views expressed herein are solely those of the authors.

Endnotes1 Initiative for Maternal Mortality Impact Assessment based at the

University of Aberdeen.2 Oneway ANOVA testing for differences in public income by contract

status, F ¼ 2.9, P¼ 0.002.3 To permit easier international comparisons, all figures are presented in

US dollars. At the time of the study, the exchange rate was 1 USdollar to 9200 Indonesian Rupiah.

4 F-test for differences across experience groups P value < 0.001, t-testfor difference between group 1 (<8 years) and group 4 (>14 years)P value¼ 0.01.

5 Oneway ANOVA F-test by distance group not significant at 5%(F¼ 1.58, P¼ 0.19) nor is a t-test between group 1 (<15km) andgroup 4 (>60km) (t¼ 1.05, P¼ 0.29). T-test between group 3(30–60km) and group 4 is significant at 5% level (t¼ 1.99,P¼ 0.049).

6 Though none of the variation-inflation factors (vif), a commonindicator of multicollinearity, are greater than 5 so multicollinear-ity is unlikely to be a substantial issue.

7 This is somewhat similar to Hammer and Jack’s model whichexamines the optimal package required to retain workers inremote areas based on the assumption that staff are split into(two) groups according to willingness to locate in remote areas(Hammer and Jack 2002). The model suggests that a choice ofcontracts should be offered; one offering a high total payment thatpays less per year for a longer period of service in the remote area;the other offering a higher average but smaller total period inreturn for a relatively short period of service.

ReferencesAchadi E, Scott S, Pambudi ES et al. 2007. Midwifery provision and

uptake of maternity care in Indonesia. Tropical Medicine and

International Health 12: 1490–7.

Berman P, Cuizon D. 2004. Multiple public-private jobholding of health

care providers in developing countries: an exploration of theory

and evidence. London: Harvard School of Public Health for DFID

Health Systems Resource Centre.

Chaudhury N, Hammer JS. 2004. Ghost doctors: absenteeism in rural

Bangladeshi health facilities. World Bank Economic Review 18:

423–44.

Chomitz KM, Setiadi G, Azwar A, Ismail N, Widiyarti. 1998. What do

doctors want? Developing incentives for doctors to serve in

INCENTIVES FOR VILLAGE MIDWIVES IN INDONESIA 9

Indonesia’s rural and remote areas. Policy Research Working Paper

Series, 1888. Washington, DC: World Bank.

Dieleman M, Cuong PV, Anh LV, Martineau T. 2003. Identifying factors

for job motivation of rural health workers in North Viet Nam.

Human Resources for Health 1: 10.

Ensor T, Thompson R. 2006. The unofficial health care economy in low

and middle-income countries. In: Jones A (ed.). The Elgar

companion to health economics. Cheltenham, UK: Edward Elgar.

Ensor T, Megraini A, Nadjib M, Quayyum Z. 2006. Analysis of financial

flows for maternal health in Serang and Pandeglang. Aberdeen &

Jakarta: Policy and Health Systems/Economic Outcomes, Initiative

for Maternal Mortality Impact Assessment.

Ensor T, Nadjib M, Quayyum Z, Megraini A. 2008. Public funding for

community-based skilled delivery care in Indonesia – to what

extent are the poor benefiting? European Journal of Health Economics

9: 385–92.

Hammer J, Jack W. 2002. Designing incentives for rural health care

providers in developing countries. Journal of Development Economics

69: 297–303.

Hatt L, Stanton C, Makowiecka K et al. 2007. Did the strategy of skilled

attendance at birth reach the poor in Indonesia? Bulletin of the

World Health Organization 85: 774–82.

Leonard KL. 2002. Active patients in rural African health care:

implications for welfare, policy and privatization. New York:

University of Colombia.

Makowiecka K, Achadi E, Izati Y, Ronsmans C. 2007. Midwifery

provision in two districts in Indonesia: how well are rural areas

served? Health Policy and Planning 23: 67–75.

Nadjib M, Junadi P, SoewondoP, Sucahya PK. 2004. Economic analysis

study of BDD Program. Jakarta: Centre for Health Research,

University of Indonesia.

Quayyum Z, Nadjib M, Sucahya PK, Ensor T. 2007a. Cost analysis of

village midwifery services in two districts of Indonesia: efficiency

and financial sustainability of the programme. Aberdeen & Jakarta:

Initiative for Maternal Mortality Impact Assessment.

Quayyum Z, Ensor T, Nadjib M, Sucahya P, Pambudi ES. 2007b.

Inequality in access to and financing of obstetric care in two

districts of Indonesia: are poor benefiting from the Bidan Di Desa

Programme and public provision of care? Aberdeen & Jakarta:

Initiative for Maternal Mortality Impact Assessment.

Reinnikka R, Smith N. 2004. Public expenditure tracking surveys in

education. Paris: International Institute for Educational Planning.

Online at: http://www.unesco.org/iiep/PDF/pubs/Reinikka.pdf.

Shiffman J. 2003. Generating political will for safe motherhood in

Indonesia. Social Science and Medicine 56: 1197–207.

Somanathan A, Rannan-Eliya R, Fernando T. 2004. Indonesia Public

Health Expenditure Review. Colombo: Institute for Policy Studies

for WHO Indonesia.

Stilwell B. 2001. Health worker motivation in Zimbabwe. Geneva: World

Health Organization (mimeo).

10 HEALTH POLICY AND PLANNING