Social Science & Medicine 110 (2014) 31e40
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Willingness to pay for public health services in rural Central Java, Indonesia: Methodological considerations when using the contingent valuation method Aiko Shono a, b, *, Masahide Kondo c, Hiroshi Ohmae d, Ichiro Okubo c a Department of Health Care Policy and Management, Doctoral Program in Human Care Science, Graduate School of Comprehensive Human Sciences, University of Tsukuba, Japan b Department of Public Health and Epidemiology, Meiji Pharmaceutical University, Japan c Department of Health Care Policy and Health Economics, Faculty of Medicine, University of Tsukuba, Japan d Department of Parasitology, National Institute of Infectious Diseases, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 February 2013 Received in revised form 10 March 2014 Accepted 24 March 2014 Available online 26 March 2014
In the health sectors of low- and middle-income countries, contingent valuation method (CVM) studies on willingness to pay (WTP) have been used to gather information on demand variation or financial perspectives alongside price setting, such as the introduction of user fees and valuation of quality improvements. However, WTP found in most CVM studies have only explored the preferences that consumers express through their WTP without exploring whether they are actually able to pay for it. Therefore, this study examines the issues pertaining to WTP estimation for health services using the conventional CVM. We conducted 202 household interviews in 2008, in which we asked respondents about three types of public health services in Indonesia and assessed WTP estimated by the conventional CVM as well as in the scenario of “resorting to debt” to recognize their budget constraints. We find that all the demand curves for both WTP scenarios show gaps. Furthermore, the gap for midwife services is negatively affected by household income and is larger for the poor. These results prove that CVM studies on WTP do not always reveal WTP in the latter scenario. Those findings suggest that WTP elicited by the conventional CVM is different to that from the maximum price that prevents respondents from resorting to debt as their WTP. In order to bridge this gap in the body of knowledge on this topic, studies should improve the scenarios that CVM analyses use to explore WTP. Furthermore, because valuing or pricing health services based on the results of CVM studies on WTP alone can exacerbate the inequity of access to these services, information provided by such studies requires careful interpretation when used for this purpose, especially for the poor and vulnerable sections of society. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Indonesia Willingness to pay Contingent valuation method Equity
1. Introduction User fees for public health services in the context of universal coverage continue to remain a policy issue in low- and middleincome countries (WHO, 2010). Not only does the introduction of user fees deter access to public health services for the poor, but their contribution to revenue and service quality improvement is also debatable (James et al., 2006; Palmer et al., 2004). There is evidence that user fees could affect equity of access for vulnerable
* Corresponding author. Department of Public Health and Epidemiology, Meiji Pharmaceutical University, Japan. E-mail address: [email protected] (A. Shono). http://dx.doi.org/10.1016/j.socscimed.2014.03.025 0277-9536/Ó 2014 Elsevier Ltd. All rights reserved.
sections of society and the utilization of health services, that is, their demand in economic terms (James et al., 2006; Lagarde and Palmer, 2008; Witter et al., 2000). Conceptually, demand refers to the quantity of a good or a service that an individual or a household chooses to buy at a given price. However, demand also represents the preference that consumers express through their ability and willingness to pay (WTP) (McPake and Normand, 2008). In other words, a general preference for acquiring a commodity only becomes a specific demand for that commodity when the consumer is willing and able to pay for it (Le Grand et al., 2008). Demand curves, which show this relation between price and demand for a given period, can be drawn by summing, at each price, the total quantities purchased by each individual in the market (Stiglitz and Walsh, 2006). The contingent valuation method (CVM)
32
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
is often used as an alternative when the market price is unavailable. In this survey method, individuals are queried about their hypothetical WTP for a good or a service (McPake and Normand, 2008). In the context of the health sector in low- and middle-income countries, CVM studies on WTP have been used to gather information on demand variation or financial perspectives alongside price setting, such as the introduction of user fees and valuation of quality improvements (Foreit and Foreit, 2003; Russell et al., 1995; Weaver et al., 1996). More recently, this method has often been employed to help set premiums for community-based insurance as a price strategy (Dong et al., 2003). However, in low- and middle-income countries, most CVM studies estimating WTP have only explored the preferences that consumers express through their WTP without exploring whether they are actually able to pay the value. In other words, their WTP estimated by these CVM studies might not reflect an unforeseen situation that may affect a household, such as the heavy burden of paying for health services (Russell, 1996; Waddington and Enyimayew, 1989). Therefore, this study examines the issues pertaining to WTP estimation for health services using the conventional CVM.
and midwives. Outpatient health services within office hours are either free or are nominally charged. Limited amounts of generic medicines can be dispensed. For inpatient services, the health center and hospital impose their own tariffs depending on the services provided. However, the poor are exempt from paying for these services. In terms of distance, the midwife is located in the village; the health center, in the central subdistrict; and the hospital, in the central district. We used the simple random sampling method to select 202 households from the list of 757 households provided by the village subheads. Two trained interviewers conducted face-to-face interviews using a structured questionnaire in the local language (Indonesian). The questionnaire was drafted in English and backtranslated. The questionnaire had four sections: respondents’ characteristics (age, sex, and years of schooling), household economic status (family size and land ownership), medical history of each family member, and a section for contingent valuation of WTP and WTPpayable for public health services. The questions were devised after referring to variables of similar existing WTP studies in lowand middle-income countries (Dong et al., 2003; Habbani et al., 2006; McNamee et al., 2010). Household income was calculated
2. Methods Firstly, we deliberated how to identify the amount that people are able to pay for a good or a service in low- and middle-income countries. While we considered the concept of ability to pay, an accepted definition of this measure has not yet been established in the literature. Moreover, we faced difficulty in examining respondents’ opportunity costs in detail, which are central to the concept of ability to pay in our research (Russell, 1996). Some CVM studies have proposed that the need to resort to debt to pay for a product or a service implies a lack of ability to pay (Habbani et al., 2003, 2006). These studies included respondents’ budget constraints in the scenario in order to ensure that the stated willingness-to-pay was indeed payable. Their approach was to improve the scenario rather than to deliberate the definition of ability to pay, which addresses one of the known CVM problems, namely “alternative expenditure possibility,” raised during the early development of CVM studies (Arrow et al., 1993). We took a similar approach for this research. Specifically, we defined and identified the maximum price that respondents can pay without resorting to debt as their WTP, thereby avoiding the inability to pay. In other words, we considered that the scenario of “resorting to debt” had to recognize respondents’ budget constraints. Then, we denoted this value as “WTPpayable,” which differs from the WTP determined by the conventional method. In the next step, we conducted CVM studies in the field by investigating households’ WTP as well as WTPpayable for a public health service. We chose Indonesia as our focus geographical area. Historically, like many low- and middle-income countries, Indonesia used to impose user fees for public health services and aims at achieving universal coverage. Although the Indonesian government is making efforts to increase access to health services for the poor, about half the population still lacks health insurance coverage, notably the informal sector (Rokx et al., 2009; Tangcharoensathien et al., 2011). The field research was conducted from November to December 2008 in the Batang district of Central Java. The district is a rural area and is representative of Central Java, where people mainly practice agriculture for a living. Our research site, “village A,” was officially designated as a disadvantaged village (Inpres Desa Tertinggal: IDT). Three types of public health services can be normally accessed: services of the midwife, the health center, and the hospital. The latter two allow hospitalization and are staffed by doctors, nurses,
Table 1 Summary statistics of variables. Variable
Summary statistics
Age Sex
Mean, years (SD) Male, % Female, % Years of schooling Mean, years (SD) Household size Mean, persons (SD) Prior experience by household member Village midwife Yes, % No, % Health center Yes, % No, % Hospital Yes, % No, % Knowledge of typhoid fevera Yes, % No, % Whether a household member Yes, % has been treated for typhoid fever No, % Satisfaction with care at public Satisfied, % health facilities Less satisfied, % Dissatisfied, % Annual household income Mean, Rpb (SD) Annual household income per head
Mean, Rp (SD)
Value 40.9 (12.7) 45.5 54.5 5.11 (3.15) 3.73 (1.54) 81.2 18.8 58.4 41.6 17.8 82.2 82.0 18.0 52.0 48.0 78.5 3.7 17.8 3 991 651 (6 034 936) 1 158 706 (2 137 004)
c
WTP for services by Village midwife Health center Hospital WTPpayabled for services by Village midwife Health center Hospital The WTP WTPpayable Village midwife Health center Hospital The (WTP WTPpayable)/WTP Village midwife Health center Hospital
Mean, Rp (SD) Mean, Rp (SD) Mean, Rp (SD)
56 139 (142 799) 147 960 (177 330) 369 005 (839 534)
Mean, Rp (SD) Mean, Rp (SD) Mean, Rp (SD)
21 572 (72 775) 61 903 (131 002) 110 311 (384 556)
Mean (SD) Mean (SD) Mean (SD)
34 567 (121 532) 86 057 (140 905) 258 694 (744 807)
Mean (SD) Mean (SD) Mean (SD)
0.505 (0.459) 0.657 (0.379) 0.690 (0.370)
a Respondents were asked “Have you ever heard of the disease named typhoid fever?”. b US$1 ¼ Rp12 195 (as of December 2008). c Willingness to pay. d The maximum price that prevents respondents from resorting to debt as their WTP.
33
0.070 0.187** 0.073 0.007 0.070 0.032 0.394** 0.268** 0.303** 0.319** 0.267** 0.315** 0.035 0.151* 0.240** 0.267** 0.110 0.096 0.006 0.063 0.018 0.266** 0.315** 0.304** 0.112 0.377** 0.284** 0.262** 0.021 0.165* 0.285** 0.305** 0.132 0.072 0.181* 0.214** **p < 0.01 *p < 0.05.
0.191** 0.081 0.207** 0.261** 0.127 0.312** 0.099 0.283** Age Years of schooling Annual household income Annual household income per head
Hospital
WTP WTPpayable WTP The (WTP WTPpayable WTPpayable)/WTP WTP
Health center
The (WTP WTPpayable)/WTP WTPpayable WTP WTPpayable WTP
Village midwife Variables
Table 2 Spearman’s correlation between WTP, WTPpayable, WTP WTPpayable, (WTP WTPpayable)/WTP, and continuous variables.
Fig. 1. Demand curves for typhoid fever treatment at different public health facilities in the study area. (a The maximum WTP and WTPpayable are Rp5 000 000.)
WTPpayable WTP The (WTP WTPpayable WTPpayable)/WTP
by taking into account the relation between household income and land ownership in Indonesia (Shiga et al., 2009), since self-reported income is considered to be unreliable in our field (Russell, 1996). Various approaches, such as payment cards, the take-it-orleave-it (TIOLI) method, and the bidding game method, have been applied in CVM studies on WTP in the health sector. Nonetheless, it is worth noting that all methodologies have some form of bias, such as the starting point bias in the bidding game method, the yessaying behavior in the TIOLI method, and the range bias in payment cards (Johannesson and Jonsson, 1991; Klose, 1999; McNamee et al., 2010). Which of these methods is more reliable in the context of in low- and middle-income countries remains controversial. Dichotomous choice methods such as the bidding game and TIOLI are often used in this type of research (Arrow et al., 1993; Cummings et al., 1995), with evidence of bias clearer in the former (Klose, 1999). Therefore, in this study we chose to employ the bidding game approach. We used a hypothetical scenario of the respondent contracting typhoid fever. This fever is a water- and food-borne infection caused by Salmonella enterica subspecies enterica serovar Typhi (S. Typhi) that continues to be a serious public health problem, especially in low- and middle-income Asian countries (Ochiai et al., 2008). Affected patients of all ages risk suffering from severe symptoms and complications without timely and appropriate care.
0.006 0.141 0.230** 0.261**
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
34
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 3 Mean WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP for public health services according to respondents’ characteristics. Variables
Village midwife Mean WTP
Age 29 30e39 40e49 50e Sex Male Female Prior utilization of village midwife by household member Yes No Prior utilization of health center by household member Yes No Prior utilization of hospital by household member Yes No Knowledge of typhoid fever Yes No Whether a household member has been treated for typhoid fever Yes No Satisfaction with care at public health facilities Satisfied Less satisfied Dissatisfied
p valuea
Health center Mean WTPpayable
0.597 75 38 49 61
870 137 022 864
761 765 152 924
0.085
23 792 18 452 0.627
66 667 53 855
0.224
0.219
0.996 56 190 56 082
24 262 18 660 0.654 23 817 22 143 14 559
187 083 139 424 0.000 161 982 87 639 0.636
0.520 0.489 0.731
39 517 7857 27 500
0.144
0.550
0.749
0.810
148 941 146 566
0.514 0.462
31 929 37 423
0.926
0.043
0.248
0.586
159 540 98 289
0.364 0.537
39 466 13 784
0.002
0.881
0.342
0.633 22 851 16 486
0.023 180 879 120 727
0.501 0.511
17 083 38 358
0.909
0.223
0.376
0.449 157 048 138 021
0.498 0.528 0.306 0.497
p value
130 569 761 603
0.257
0.486 0.591
28 157 43 571
49 583 15 497
62 317 30 270
63 333 30 000 42 059
25 351 74 342
Mean WTP
134 151 160 145
0.464 0.540 0.179
0.608
p value 0.795
0.189
0.483
0.622
Mean (WTP WTPpayable)/WTP 0.517 0.546 0.506 0.456
47 853 23 455
23 308 14 079
51 949 62 024
109 373 870 941
0.199
0.309
p value 0.566
55 21 35 28
28 777 15 545
48 659 88 421
Mean WTP WTPpayable
0.519 20 16 13 32
76 630 39 000
p value
0.669 147 633 114 286 172 206
a p values derived from t test for sex, prior utilization of a health service, knowledge of typhoid fever, and ANOVA for age, years of schooling, and satisfaction level with the health service.
The infection could burden a household with the high costs associated with its treatment (Poulos et al., 2011). Therefore, each respondent was provided with three hypothetical scenarios of types of care depending on the severity of typhoid fever: outpatient care provided by the midwife, hospitalization for three days attended to by doctors at the health center, and hospitalization for four days attended to by doctors at the hospital. We then provided the following five starting bids for each scenario: Midwife: Rp10 000 (US$0.82), Rp15 000 (US$1.23), Rp20 000 (US$1.64), Rp30 000 (US$2.46), and Rp40 000 (US$3.28) Health center: Rp60 000 (US$4.92), Rp90 000 (US$7.38), Rp120 000 (US$9.84), Rp180 000 (US$14.76), and Rp240 000 (US$19.68) Hospital: Rp160 000 (US$13.12), Rp240 000 (US$19.68), Rp320 000 (US$26.24), Rp480 000 (US$39.36), and Rp640 000 (US$0.82). At the start of the contingent valuation section, the respondent was asked whether s/he had ever heard of typhoid fever. If s/he answered “no,” the interviewer explained the common symptoms of typhoid fever to him/her (the scenario used in this study is given in the Appendix). In addition, s/he was also asked to assume that the required treatment/drugs were available at each health care facility for each of the three levels in the scenario. Secondly, s/he was provided with the first hypothetical scenario on outpatient care, that is, care provided by the midwife, and was randomly assigned to one of the five starting bids. The respondent was asked if s/he would be willing to pay the first bid. If s/he answered “yes,” a higher bid was provided to check the highest
possible price the respondent would be willing to pay. If the answer was “no,” a lower bid was provided to check the highest possible price the respondent would be willing to pay. We considered respondents’ WTP as the maximum bid they agreed to pay at the end of the interview. Next, we inquired about the price the respondent could pay without incurring any debts (i.e., their WTPpayable). We proceeded with the remaining two levels of care in the hypothetical scenario, that is, hospitalization at the health center and at the hospital, in the same manner. Thirdly, we estimated two types of demand curves using the results of each CVM study, that is, one based on WTP and the other on WTPpayable. By using the method outlined in McIntosh et al. (2010), we then drew two demand curves using the stated WTP and WTPpayable for the three scenarios. We also drew supply curves that were depicted with straight lines in line with our assumption that the respondent could avail of any of the three types of public health services should they wish to. In other words, it was assumed that appropriate health care would always be provided at the health facilities. Next, we explored the gap between WTP and WTPpayable as follows:
D ¼ WTP WTPpayable % ¼ ðWTP WTPpayableÞ=WTP If WTP was zero, we excluded such samples from our analysis for percentages. The excluded WTP cases were 10, 5, and 10 cases for the midwife, health center, and hospital services, respectively. We employed Spearman’s correlation in order to study the
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Health center
35
Hospital
Mean p value Mean WTP p value Mean (WTP p value Mean WTP p value Mean WTP p value Mean WTP p value Mean (WTP p value WTPpayable WTPpayable WTPpayable)/WTP payable WTPpayable WTPpayable)/WTP 0.454 57 73 37 75
065 137 500 216
0.570 77 78 123 70
065 431 261 388
0.055 82 720 44 682
0.292 98 159 76 045
0.002 69 954 27 368
0.190 98 194 53 985
0.895 88 889 85 439
0.585 64 497 51 250
0.000
0.567
0.049
100 214 70 573
0.349 0.681 0.630
0.725 85 750 67 143 104 853
0.294 0.684 0.522 0.591
79 119 144 427
association between the continuous variables, WTP, WTPpayable, D ¼ WTP WTPpayable, and % ¼ (WTP WTPpayable)/WTP. The t test and one-way analysis of variance (ANOVA) were used to ascertain whether WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP depended on their attributes or experiences. Furthermore, multiple regression analysis was used to examine the variables that influence WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP. We used SPSS V. 17.0 for the statistical analysis and considered differences significant at p < 0.05. This research protocol was approved by the Ethical Committee of the University of Tsukuba, Japan. The conduct of the field survey was approved by the State Ministry of Research and Technology, Indonesia. Respondents were interviewed with prior informed consent. 3. Results Table 1 shows the summary statistics of respondents’ characteristics. The mean age of respondents was 40.9 years and 45.5% of them were male. Their mean years of schooling and average household size were 5.11 and 3.73, respectively. Annual household income was Rp3 991 651 (US$327.3), while annual household income per head was Rp1 158 706 (US$95.0). We found mean individual WTP values of Rp56 139 (US$4.6), Rp147 960 (US$12.1), and Rp369 005 (US$30.3) for the services provided by the midwife, health center, and hospital, respectively. The corresponding mean individual WTPpayable values were Rp21 572 (US$1.8), Rp61 903 (US$5.1), and Rp110 311 (US$9.0), respectively. The corresponding
0.001
0.044 0.718 0.553
0.114 338 167 171 771
0.792 109 550 210 714 101 618
0.093 0.595 0.711
300 229 76 667 0.250
0.981 376 967 325 000 394 706
0.318
0.782
0.395
0.122 0.724 0.640
371 389 234 106
114 405 94 722
417 286 316 198
0.309
0.270
0.002
0.002 0.656 0.834
303 665 194 759
174 444 96 318
414 634 171 389
0.962
0.593
0.164
0.144 0.647 0.726
257 469 263 947
122 521 92 952
545 833 330 424
0.685 0.545 0.137
0.932 61 883 47 143 67 353
0.598 0.670
97 485 36 389
56 833 67 448
0.314
0.089
0.149
0.251
0.770 0.671 0.707 0.732 0.658
363 159 172 273
129 218 29 211
426 186 287 711
826 922 500 078
0.069
0.538
0.520
0.261 147 253 447 201
169 808 61 091
386 687 293 158
0.671 0.636
696 118 717 647
0.018
0.018
0.702
0.444 78 204 62 90
532 967 233 364
0.629 0.776
82 860 90 602
522 039 217 724
0.080
0.464
0.591
0.152 226 458 510 291
0.605 0.700
89 586 70 921
66 081 55 964
0.500 0.642 0.686 0.710 0.602
0.086 0.734 0.640
0.853 267 417 114 286 293 088
0.143 0.713 0.425 0.657
values of WTP WTPpayable were Rp34 567 (US$2.8), Rp86 057 (US$7.1), and Rp258 694 (US$21.2), respectively. The corresponding values of (WTP WTPpayable)/WTP were 0.505, 0.657, and 0.690, respectively. All the demand curves based on the WTP and WTPpayable values show gaps between respondents’ WTP and WTPpayable (log rank test: p < 0.001) (Fig. 1). The percentages of respondents whose WTP exceeded the approximate actual price, as shown by the supply line, were 70.8%, 24.4%, and 15.4% for the midwife, health center, and hospital, respectively. The corresponding percentages of respondents who indicated a WTPpayable exceeding the approximate price were 37.1%, 9.5%, and 4.5%. Table 2 describes the correlation between the continuous variables, namely WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP. The WTP values for the three services were correlated to years of schooling, while the WTPpayable values for the three services were correlated to years of schooling, annual household income, and annual household income per head. WTP WTPpayable for the midwife service was related to annual household income and annual household income per head, and for the hospital service, it was related to years of schooling. (WTP WTPpayable)/WTP for all three services was related to annual household income and annual household income per head, and for the midwife and health center services were also related to years of schooling. Table 3 provides the mean WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP, showing differences between groups for the three services. The mean WTP, WTPpayable, and WTP WTPpayable for the midwife service did
0.321 0.995 0.0160 0.012 0.280 1.221 0.006 2.769 0.0931 0.097 0.002 2.975 0.000 4.022 0.0200 0.187 0.000 5.231 0.001 3.310 0.0222 0.198 0.000 5.541 2.301 0.023
household. a
0.216
0.456
16 637 1.240
0.00430 0.746 0.084 0.004 2.679
0.00656 0.135 0.000 3.876
0.852 7003 0.187 0.777 22 248 0.284 0.803 2901 0.250 0.745 0.326 5083
0.514 15 359 0.654
8650 1.303 0.194
0.194 85 646 1.305 0.763 0.302 41 338 0.475 14 523 0.716 0.806 6711 0.246 0.647 0.519
0.442 0.771 24 371
7512
0.811 0.239 21 156 0.569 0.571 105 332 0.958 1432 0.052 0.151 52 979 12 099 0.773 0.441
1.442
0.340 0.957 71 297 0.262 1.126 174 982 0.087 1.720 39 583 0.303 1.034 31 995 2.499 0.013 32 955 0.772 0.290 7713
0.623 0.493 28 618 0.340 0.956 115 896 0.549 0.600 10 766 0.820 5496 0.228 0.394 0.694 0.461 15 330 0.738
4054
0.706 0.739 0.113 0.475 0.218 70 863 0.377 893 0.334 95 165 1.595 8464 0.715 90 582 1.236 0.976 0.949 0.083 0.208 0.737 11 631 0.03 354 0.063 216 904 1.741 31 221 1.264 51 363 0.336 0.805 0.980 0.237 0.069 0.093 14 339 0.247 20 0.025 21 880 1.186 6689 1.830 38 252 1.689 0.595 0.531 0.201 0.012 0.111 41 508 0.532 697 0.628 31 844 1.284 12 479 2.539 48 836 1.604 26 105 0.785 0.433 426 0.901 0.369 4077 0.386 0.700 5735 2.738 0.007 6650 0.512 0.609 0.714 0.455 0.289 0.000 0.047 24 690 0.368 715 0.749 22 717 1.064 15 892 3.756 52 552 2.005
Constant Age Sex (male:0, female:1) Years of schooling Prior utilization of village midwife by HHa member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic
p value Coefficient t value
WTPpayable WTP
p value Coefficient t value p value Coefficient t value p value Coefficient t value p value Coefficient t value Coefficient t value
Hospital
WTP
WTPpayable Health center
WTP
WTPpayable Village midwife Variables
not show any statistical significance between groups. However, the mean (WTP WTPpayable)/WTP for the same service showed a statistical significance between groups with and without a household member who had availed of the hospital service in the past. The mean WTP for the health center service indicated statistical significance between male and female respondents between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. The mean WTPpayable for the health center service showed a statistical significance between groups with and without a household member who had accessed the midwife service in the past. The mean WTP WTPpayable for the health center service showed statistical significance between groups with and without prior knowledge of typhoid fever. The mean (WTP WTPpayable)/WTP for the health center service showed statistical significance between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. The mean WTP for the hospital service indicated statistical significance between male and female respondents and between groups with and without prior knowledge of typhoid fever. The mean WTPpayable for the hospital service did not show any statistical significance. For the hospital service, the mean WTP WTPpayable showed statistical significance between groups with and without prior knowledge of typhoid fever. For the hospital service, the mean (WTP WTPpayable)/WTP showed statistical significance between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. Table 4 shows the results of the multiple regression analysis on WTP and WTPpayable for public health services. The WTP for midwife services was significantly affected by years of schooling and prior utilization of midwife services by a household member, whereas the WTPpayable for this service was significantly affected by years of schooling, prior utilization of hospital services by a household member, and annual household income per head. The WTP for health center services was significantly affected by years of schooling and annual household income per head, and WTPpayable, by annual household income per head. Finally, the WTP for hospital services was significantly affected by annual household income per head, while WTPpayable was not significantly affected by any of the variables. Table 5 shows the results of the multiple regression analysis on WTP WTPpayable for public health services. For both the full and stepwise models, the WTP WTPpayable for midwife services was significantly affected by years of schooling, prior utilization of midwife services by a household member, and annual household income per head. For the health center service, the WTP WTPpayable assessed using the full model was not significantly affected by any of the variables. For the health center service, the WTP WTPpayable assessed using the stepwise model was significantly affected by the respondent’s knowledge of typhoid fever. Finally, the WTP WTPpayable for hospital services was significantly affected by annual household income per head in both the full model and the stepwise model. Table 6 shows the results of the multiple regression analysis on (WTP WTPpayable)/WTP for these services. The (WTP WTPpayable)/WTP for midwife services was significantly affected by prior utilization of the hospital service by a household member and annual household income per head in both models. The (WTP WTPpayable)/WTP for health center services was significantly affected by prior utilization of midwife services by a household member, prior knowledge of typhoid fever, and annual
p value
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 4 Results of the multiple regression analysis on WTP and WTPpayable for public health services.
36
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
37
Table 5 Results of the multiple regression analysis on WTP WTPpayable for public health services. Stepwise model
Village midwife Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic Health center Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic Hospital Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic a
Full model p valuea
Coefficient
t value
42 762
1.831
0.069
9922 58 691
3.196 2.630
0.002 0.009
0.00966 0.065 0.002 5.239
2.127
0.035
36 717
1.538
0.126
60 472
2.288
0.023
149 389
2.483
0.014
0.0943 0.068 0.000 14.47
3.803
0.000
0.022 0.023 5.233
Coefficient
t value
p value
50 836 288 18 643 10 150 59 199 19 379 25 241 36 468 22 867 7977 0.0108 0.061 0.020 2.193
0.879 0.350 1.014 2.785 2.621 1.083 1.100 1.338 1.131 0.690 2.185
0.381 0.727 0.312 0.006 0.010 0.280 0.273 0.183 0.260 0.491 0.030
26 310 702 10 036 5684 10 629 16 205 7556 54 385 7895 7983 0.00235 0.006 0.356 1.111
0.381 0.716 0.458 1.307 0.395 0.759 0.276 1.673 0.327 0.579 0.396
0.704 0.475 0.648 0.193 0.694 0.449 0.783 0.096 0.744 0.563 0.692
82 561 1249 121 738 22 774 39 218 87 270 103 689 84 108 126 996 15 253 0.0770 0.064 0.017 2.254
0.233 0.248 1.082 1.021 0.284 0.797 0.738 0.504 1.026 0.216 2.536
0.816 0.804 0.281 0.309 0.777 0.426 0.461 0.615 0.306 0.829 0.012
p (in) ¼ 0.05, p (out) ¼ 0.1.
household income per head in the full and stepwise models. For the hospital service, the (WTP WTPpayable)/WTP assessed using the full model was significantly affected by prior utilization of midwife services by a household member and prior knowledge of typhoid fever. Finally, for the hospital service, the (WTP WTPpayable)/ WTP assessed using the stepwise model was significantly affected by respondent’s years of schooling, prior utilization of the midwife service by a household member, and prior knowledge of typhoid fever. 4. Discussion In this study, we conducted interviews to assess the WTP and WTPpayable for three types of public health services in Central Java, Indonesia, using the conventional CVM and recognizing respondents’ budget constraints, respectively. Firstly, we defined the maximum price at which respondents can avoid resorting to debt as their WTPpayable. Specifically, we considered that the scenario of resorting to debt had to recognize the budget constraints of respondents.
We anticipated that each household’s WTPpayable for health services would be fixed irrespective of the scenarios. However, many respondents indicated different WTPpayable values depending on the three types of public health services examined. We believe that this is because we used debt to differentiate WTP from WTPpayable. Respondents set household priorities and changed their expenditure patterns according to the situation at hand, as most have potential resources such as land, livestock, and crops that could be converted into money (Russell, 1996; Wallman and Baker, 1996). In other words, they make discretional choices between selling assets and borrowing money. All demand curves based on the WTP and WTPpayable values show gaps. This finding suggests that WTP elicited by the conventional CVM is different to that from WTPpayable (Fig. 1). However, few CVM studies on WTP have mentioned whether the WTP is able to be paid, although many authors have shown that economic indicators such as income correlate positively with WTP (Dong et al., 2003; Foreit and Foreit, 2003; Frick et al., 2003). Therefore, the findings of CVM studies on WTP should be interpreted with caution. Moreover, to bridge the gap between WTP and WTPpayable elicited
38
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 6 Results of the multiple regression analysis on (WTP WTPpayable)/WTP for public health services. Stepwise model
Village midwife Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head 2
Adjusted R p value F-statistic Health center Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head 2
Adjusted R p value F-statistic Hospital Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic a
Full model t value
p valuea
Coefficient
t value
p value
14.74
0.000
0.171
1.990
0.048
4.958 108
3.222
0.002
0.633 0.000 0.046 0.013 0.106 0.018 0.175 0.110 0.002 0.003 4.006 108
2.860 0.010 0.648 0.933 1.229 0.259 1.995 1.056 0.025 0.070 2.109
0.005 0.992 0.518 0.352 0.221 0.796 0.048 0.292 0.980 0.944 0.036
3.983 0.242 1.347 1.158 2.364 0.397 1.130 2.349 0.177 0.748 1.974
0.000 0.809 0.180 0.248 0.019 0.692 0.260 0.020 0.860 0.456 0.050
3.931 0.341 1.299 1.306 2.898 1.333 1.922 2.487 0.487 0.360 0.868
0.000 0.733 0.196 0.193 0.004 0.184 0.056 0.014 0.627 0.719 0.387
Coefficient 0.593
0.064 0.001 7.259
0.045 0.054 1.856
0.715
8.921
0.000
0.173
2.514
0.013
0.159
2.295
0.023
3.247
0.001
4.059 108 0.088 0.000 6.889
0.088 0.003 2.778
0.772
9.440
0.000
0.022 0.190
2.544 2.802
0.012 0.006
0.224
3.232
0.001
0.092 0.000 7.119
0.709 0.001 0.076 0.013 0.164 0.022 0.080 0.197 0.011 0.027 3.012 108
0.682 0.001 0.072 0.014 0.196 0.071 0.132 0.203 0.030 0.012 1.292 108 0.101 0.001 3.034
p(in) ¼ 0.05, p (out) ¼ 0.1.
by CVM, studies should improve the scenarios used in CVM studies on WTP. In particular, researchers must inform respondents that they are actually choosing to avail of a given health service at a given/ stated WTP and remind them that this would also reduce their other expenditures. This process might improve the reliability of estimating and valuing services (Arrow et al., 1993). Further, in studies examining respondents in low- and middle-income countries, there is a shortage of information regarding the source of the money for respondents’ stated WTP payments. Household income is limited, especially for the poorer population, although their potential income resources are diverse and unstable in contrast to developed countries (Donaldson, 1999; Russell, 1996). Therefore, WTP does not necessarily mean that this amount can be instantly paid (i.e., WTP s WTPpayable), which rather tends to be decided through their complex choices on how to spend their finite household income on health care and other competing needs (Russell, 1996; Waddington and Enyimayew, 1989). The empirical findings should be carefully interpreted when deciding on a new pricing strategy or quality improvements. If
prices are set in response to the results of CVM studies on WTP rather than WTPpayable, it is highly likely that poor and vulnerable groups would not benefit from service and quality improvement, or that repetitive payments for these services would impose an excessive financial burden on households, forcing them to borrow money to pay for medical treatment and thus pushing them even deeper into poverty (Poulos et al., 2011; Whitehead et al., 2001). Therefore, subsidizing this segment of the population may significantly correct the inequitable access to relatively high cost health services, although the issue of health finance policies is still relevant here. According to the presented results of the multiple regression analysis, for the WTP WTPpayable and (WTP WTPpayable)/ WTP values for midwife services (Tables 5 and 6, respectively), the gap is negatively affected by household income. Moreover, this gap is larger for the poor. That is, valuing or pricing health services based on the results of CVM studies on WTP rather than WTPpayable may increase the inequity of access to health services. For health center and hospital services, we must also be cautious
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
when contextualizing the results, because respondents’ economic status did not seem to significantly affect either WTP WTPpayable or (WTP WTPpayable)/WTP. In many low- and middle-income countries, policymakers are increasingly interested in social and community health insurance as a possible risk-sharing mechanism and a more equitable alternative to user fees (WHO, 2010). Moreover, many CVM studies on WTP have been conducted to help gain information to arrive at a premium-priced service (Asgary et al., 2004; Donfouet et al., 2011; Dong et al., 2003; Gustafsson-Wright et al., 2009; Onwujekwe et al., 2010). However, most such studies overlook whether respondents are able to pay for WTP, meaning that the objective of the insurance system, namely to correct the inequity in access to health services, would not be achieved. There is room for improving the scenario used in this study, which might have been too simplistic. First, the explanation of typhoid fever might have been too brief, and respondents that stated they knew about the disease might actually have confused typhoid fever with a different disease. A better description, such as a depiction of the disease with illustrations, could overcome this problem. Second, since an exemption scheme was available to the poor in the field, the scenario should have reflected this fact, which would have made the scenario more realistic for respondents. Third, it is important to explore other aspects of the budget constraint when designing a scenario. Since typhoid fever is a communicable disease, the risk of other household members falling ill may be added, or other typical types of major expenditure may be included in the scenario to minimize potential budget constraint bias (Mitchell and Carson, 1989). In the bidding game part of the interview, respondents were offered a higher bid if they accepted the first amount and a lower bid if they refused the first amount. We intended to increase/ decrease the amount in increments/decrements of the same interval for the next bid for each service by referring to the findings of previous research (McNamee et al., 2010). Practically, however, respondents were not offered the amount in increments/decrements of the same interval for the next bid after accepting the current bid, because there were large gaps between the rich and poor populations in our research site. While increasing/decreasing the interval systematically for richer respondents was a long and tedious process, the small interval was meaningful for poorer respondents. Therefore, we had to change the interval depending on the respondents. Future studies should aim to explore the size of the increment/decrement of the interval that might be adopted after respondents refuse each bid in the process of the bidding game in the context of low- and middle-income countries. Moreover, our survey was conducted in a situation where respondents could actually access public health services, and we explained this assumption to them. Hence, this research may be treated as a case study. Furthermore, we adopted typhoid fever as the disease of interest in the CVM survey. Therefore, we cannot assume that the results of this study also apply to other types of diseases, including other chronic diseases. Finally, we estimated annual household income by taking into account the relation between household income and land ownership instead of directly asking respondents to state their income levels. However, other methods, such as means tests, could be used to assess household income, which in low- and middle-income countries tends to be derived from diversified sources (Jehu-Appiah et al., 2010). 5. Conclusion The results of this study showed that CVM studies on WTP might not always clarify respondents’ demand, especially for poor and vulnerable groups in low- and middle-income countries. In
39
order to bridge this gap in the body of knowledge on this topic, studies should improve the scenario used in CVM studies on WTP. In particular, researchers must inform respondents that they are choosing to buy at a given WTP and remind them that this would also reduce their other expenditures. This process might improve the reliability of estimating and valuing health services. As a policy implication, it is important to note that information provided by CVM studies on WTP requires careful interpretation when used for pricing or valuing public health services, especially for poor and vulnerable groups in low- and middle-income countries, which might not be able to pay the stated WTP value. In this respect, the findings of this research can help poor and vulnerable populations avoid falling into the medical poverty trap when utilizing public health services. Competing interests None declared. Acknowledgment We appreciate the assistance provided by Mr. Farikhun Asror, Health Officer of Batang District, in introducing our team into the field area. We are also very grateful to the villagers for their participation in our survey. This study was funded by a grant from the University of Tsukuba. Appendix. The scenarios used in the CVM studies to estimate WTP and WTPpayable 1) Firstly, have you heard of the disease typhoid fever? If “no,” the symptoms of typhoid fever are described below. If suffering from typhoid fever, patients have a very high fever and feel a chill for several days. Anyone can suffer from the disease. In order to recover, treatment is needed. The following question is hypothetical. Please choose the answer that best suits what you think. 2) If you were suffering from typhoid fever, would you have the willingness to pay the first bid for the midwife service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of care. The service is assumed to be satisfactory. 3) How much can you afford to pay without going into debt? 4) Please consider the following situation. Although you visited the midwife, you were immediately transferred without any treatment by the midwife, namely without making a payment, to a health center in your subdistrict for hospitalization because of your condition. Would you be willing to pay the first bid to avail of the health center’s service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of hospitalization for three days. The service is assumed to be satisfactory. 5) How much can you pay without going into debt? 6) Please consider the following situation. Although you visited the health center, you were immediately transferred without any treatment by health center, namely without making a payment, to the hospital because of your condition. Would you be willing to pay the first bid to avail of the hospital’s service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of hospitalization for four days. The service is assumed to be satisfactory. 7) How much can you pay without going into debt? Note: We formulated the question for WTPpayable in a slightly different manner from that used for estimating the WTP using the
40
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
conventional method (i.e., what can you pay without going into debt?). We confirmed that respondents understood the word “debt” as borrowing money from neighbors or moneylenders inside or outside the village through prior interviews with villagers including the village chief, educated villagers, and moneylenders in the research area.
References Arrow, K., Solow, R., Portney, P.R., Leamer, E.E., Radner, R., Schuman, H., 1993. Report of NOAA Panel on Contingent Valuation, pp. 1e67. www.darrp.noaa.gov/library/ pdf/cvblue.pdf (accessed 27.05.13.). Asgary, A., Willis, K., Taghvaei, A.A., Rafeian, M., 2004. Estimating rural households’ willingness to pay for health insurance. The European Journal of Health Economics 5, 209e215. Cummings, R.G., Harrison, G.W., Rutström, E.E., 1995. Homegrown values and hypothetical surveys: is the dichotomous choice approach incentive-compatible? The American Economic Review 85, 260e266. Donaldson, C., 1999. Valuing the benefits of publicly-provided health care: does ‘ability to pay’ preclude the use of ‘willingness to pay’? Social Science & Medicine 49, 551e563. Donfouet, H.P., Makaudze, E., Mahieu, P.A., Malin, E., 2011. The determinants of the willingness-to-pay for community-based prepayment scheme in rural Cameroon. International Journal of Health Care Finance and Economics 11, 209e220. Dong, H., Kouyate, B., Cairns, J., Mugisha, F., Sauerborn, R., 2003. Willingness-to-pay for community-based insurance in Burkina Faso. Health Economics 12, 849e 862. Foreit, J.R., Foreit, K.G., 2003. The reliability and validity of willingness to pay surveys for reproductive health pricing decisions in developing countries. Health Policy 63, 37e47. Frick, K.D., Lynch, M., West, S., Munoz, B., Mkocha, H.A., 2003. Household willingness to pay for azithromycin treatment for trachoma control in the United Republic of Tanzania. Bulletin of the World Health Organization 81, 101e107. Gustafsson-Wright, E., Asfaw, A., van der Gaag, J., 2009. Willingness to pay for health insurance: an analysis of the potential market for new low-cost health insurance products in Namibia. Social Science & Medicine 69, 1351e1359. Habbani, K., Groot, W., Jelovac, I., 2003. User fees policy in Khartoum state of Sudan: the ability to pay for the public health services and its relation with willingness to pay. Maastricht University, Mimeo. Habbani, K., Groot, W., Jelovac, I., 2006. Household health-seeking behaviour in Khartoum, Sudan: the willingness to pay for public health services if these services are of good quality. Health Policy 75, 140e158. James, C.D., Hanson, K., McPake, B., Balabanova, D., Gwatkin, D., Hopwood, I., et al., 2006. To retain or remove user fees? Reflections on the current debate in lowand middle-income countries. Applied Health Economics and Health Policy 5, 137e153. Jehu-Appiah, C., Aryeetey, G., Spaan, E., Agyepong, I., Baltussen, R., 2010. Efficiency, equity and feasibility of strategies to identify the poor: an application to premium exemptions under National Health Insurance in Ghana. Health Policy 95, 166e173. Johannesson, M., Jonsson, B., 1991. Economic evaluation in health care: is there a role for cost-benefit analysis? Health Policy 17, 1e23. Klose, T., 1999. The contingent valuation method in health care. Health Policy 47, 97e123.
Lagarde, M., Palmer, N., 2008. The impact of user fees on health service utilization in low- and middle-income countries: how strong is the evidence? Bulletin of the World Health Organization 86, 839e848. Le Grand, J., Propper, C., Smith, S., 2008. The Economics of Social Problems, fourth ed. Palgrave Macmillan, New York. McIntosh, E., Clarke, P.M., Frew, E.J., Louviere, J.J., 2010. Applied Methods of Costbenefit Analysis in Health Care. Oxford University Press, Oxford. McNamee, P., Ternent, L., Gbangou, A., Newlands, D., 2010. A game of two halves? Incentive incompatibility, starting point bias and the bidding game contingent valuation method. Health Economics 19, 75e87. McPake, B., Normand, C., 2008. Health Economics: An international perspective, second ed. Routledge, Oxon. Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Resources for the Future, Washington, D.C. Ochiai, R.L., Acosta, C.J., Danovaro-Holliday, M.C., Baiqing, D., Bhattacharya, S.K., Agtini, M.D., et al., 2008. A study of typhoid fever in five Asian countries: disease burden and implications for controls. Bulletin of the World Health Organization 86, 260e268. Onwujekwe, O., Okereke, E., Onoka, C., Uzochukwu, B., Kirigia, J., Petu, A., 2010. Willingness to pay for community-based health insurance in Nigeria: do economic status and place of residence matter? Health Policy and Planning 25, 155e161. Palmer, N., Mueller, D.H., Gilson, L., Mills, A., Haines, A., 2004. Health financing to promote access in low income settings: how much do we know? Lancet 364, 1365e1370. Poulos, C., Riewpaiboon, A., Stewart, J.F., Clemens, J., Guh, S., Agtini, M., et al., 2011. Cost of illness due to typhoid fever in five Asian countries. Tropical Medicine & International Health 16, 314e323. Rokx, C., Schieber, G., Harimurti, P., Tandon, A., Somanathan, A., 2009. Health Financing in Indonesia: A Reform Road Map. The International Bank for Reconstruction and Development/The World Bank, Washington D.C. Russell, S., 1996. Ability to pay for health care: concepts and evidence. Health Policy and Planning 11, 219e237. Russell, S., Fox-Rushby, J., Arhin, D., 1995. Willingness and ability to pay for health care: a selection of methods and issues. Health Policy and Planning 10, 94e101. Shiga, K., Shono, A., Kondo, M., Masuda, M., 2009, November. Indonesia no nouson ni okeru totisyoyuu to setaisyuunyuu no kankei: naka-java syu blora-ken wo zirei to shite. In: Koichi, Miyoshi (Ed.), The 20th Annual Conference. The Japan Society for International Development, Beppu, November 21e22. Stiglitz, J.E., Walsh, C.E., 2006. Economics, fourth ed. W. W. Norton & Company, Inc., New York. Tangcharoensathien, V., Patcharanarumol, W., Ir, P., Aljunid, S.M., Mukti, A.G., Akkhavong, K., et al., 2011. Health-financing reforms in southeast Asia: challenges in achieving universal coverage. Lancet 377, 863e873. Waddington, C., Enyimayew, K.A., 1989. A price to pay: the impact of user charges in ashanti-akim district, Ghana. The International Journal of Health Planning and Management 4, 17e47. Wallman, S., Baker, M., 1996. Which resources pay for treatment? A model for estimating the informal economy of health. Social Science & Medicine 42, 671e 679. Weaver, M., Ndamobissi, R., Kornfield, R., Blewane, C., Sathe, A., Chapko, M., et al., 1996. Willingness to pay for child survival: results of a national survey in Central African Republic. Social Science & Medicine 43, 985e998. Whitehead, M., Dahlgren, G., Evans, T., 2001. Equity and health sector reforms: can low-income countries escape the medical poverty trap? Lancet 358, 833e836. WHO, 2010. World Health Report 2010: Health Systems Financing: The path to universal coverage. World Health Organization, Geneva. Witter, S., Ensor, T., Jowett, M., Thompson, R., 2000. Health Economics for Developing Countries: A practical guide. Macmillan Education, Oxford.
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Willingness to pay for public health services in rural Central Java, Indonesia: Methodological considerations when using the contingent valuation method Aiko Shono a, b, *, Masahide Kondo c, Hiroshi Ohmae d, Ichiro Okubo c a Department of Health Care Policy and Management, Doctoral Program in Human Care Science, Graduate School of Comprehensive Human Sciences, University of Tsukuba, Japan b Department of Public Health and Epidemiology, Meiji Pharmaceutical University, Japan c Department of Health Care Policy and Health Economics, Faculty of Medicine, University of Tsukuba, Japan d Department of Parasitology, National Institute of Infectious Diseases, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 February 2013 Received in revised form 10 March 2014 Accepted 24 March 2014 Available online 26 March 2014
In the health sectors of low- and middle-income countries, contingent valuation method (CVM) studies on willingness to pay (WTP) have been used to gather information on demand variation or financial perspectives alongside price setting, such as the introduction of user fees and valuation of quality improvements. However, WTP found in most CVM studies have only explored the preferences that consumers express through their WTP without exploring whether they are actually able to pay for it. Therefore, this study examines the issues pertaining to WTP estimation for health services using the conventional CVM. We conducted 202 household interviews in 2008, in which we asked respondents about three types of public health services in Indonesia and assessed WTP estimated by the conventional CVM as well as in the scenario of “resorting to debt” to recognize their budget constraints. We find that all the demand curves for both WTP scenarios show gaps. Furthermore, the gap for midwife services is negatively affected by household income and is larger for the poor. These results prove that CVM studies on WTP do not always reveal WTP in the latter scenario. Those findings suggest that WTP elicited by the conventional CVM is different to that from the maximum price that prevents respondents from resorting to debt as their WTP. In order to bridge this gap in the body of knowledge on this topic, studies should improve the scenarios that CVM analyses use to explore WTP. Furthermore, because valuing or pricing health services based on the results of CVM studies on WTP alone can exacerbate the inequity of access to these services, information provided by such studies requires careful interpretation when used for this purpose, especially for the poor and vulnerable sections of society. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Indonesia Willingness to pay Contingent valuation method Equity
1. Introduction User fees for public health services in the context of universal coverage continue to remain a policy issue in low- and middleincome countries (WHO, 2010). Not only does the introduction of user fees deter access to public health services for the poor, but their contribution to revenue and service quality improvement is also debatable (James et al., 2006; Palmer et al., 2004). There is evidence that user fees could affect equity of access for vulnerable
* Corresponding author. Department of Public Health and Epidemiology, Meiji Pharmaceutical University, Japan. E-mail address: [email protected] (A. Shono). http://dx.doi.org/10.1016/j.socscimed.2014.03.025 0277-9536/Ó 2014 Elsevier Ltd. All rights reserved.
sections of society and the utilization of health services, that is, their demand in economic terms (James et al., 2006; Lagarde and Palmer, 2008; Witter et al., 2000). Conceptually, demand refers to the quantity of a good or a service that an individual or a household chooses to buy at a given price. However, demand also represents the preference that consumers express through their ability and willingness to pay (WTP) (McPake and Normand, 2008). In other words, a general preference for acquiring a commodity only becomes a specific demand for that commodity when the consumer is willing and able to pay for it (Le Grand et al., 2008). Demand curves, which show this relation between price and demand for a given period, can be drawn by summing, at each price, the total quantities purchased by each individual in the market (Stiglitz and Walsh, 2006). The contingent valuation method (CVM)
32
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
is often used as an alternative when the market price is unavailable. In this survey method, individuals are queried about their hypothetical WTP for a good or a service (McPake and Normand, 2008). In the context of the health sector in low- and middle-income countries, CVM studies on WTP have been used to gather information on demand variation or financial perspectives alongside price setting, such as the introduction of user fees and valuation of quality improvements (Foreit and Foreit, 2003; Russell et al., 1995; Weaver et al., 1996). More recently, this method has often been employed to help set premiums for community-based insurance as a price strategy (Dong et al., 2003). However, in low- and middle-income countries, most CVM studies estimating WTP have only explored the preferences that consumers express through their WTP without exploring whether they are actually able to pay the value. In other words, their WTP estimated by these CVM studies might not reflect an unforeseen situation that may affect a household, such as the heavy burden of paying for health services (Russell, 1996; Waddington and Enyimayew, 1989). Therefore, this study examines the issues pertaining to WTP estimation for health services using the conventional CVM.
and midwives. Outpatient health services within office hours are either free or are nominally charged. Limited amounts of generic medicines can be dispensed. For inpatient services, the health center and hospital impose their own tariffs depending on the services provided. However, the poor are exempt from paying for these services. In terms of distance, the midwife is located in the village; the health center, in the central subdistrict; and the hospital, in the central district. We used the simple random sampling method to select 202 households from the list of 757 households provided by the village subheads. Two trained interviewers conducted face-to-face interviews using a structured questionnaire in the local language (Indonesian). The questionnaire was drafted in English and backtranslated. The questionnaire had four sections: respondents’ characteristics (age, sex, and years of schooling), household economic status (family size and land ownership), medical history of each family member, and a section for contingent valuation of WTP and WTPpayable for public health services. The questions were devised after referring to variables of similar existing WTP studies in lowand middle-income countries (Dong et al., 2003; Habbani et al., 2006; McNamee et al., 2010). Household income was calculated
2. Methods Firstly, we deliberated how to identify the amount that people are able to pay for a good or a service in low- and middle-income countries. While we considered the concept of ability to pay, an accepted definition of this measure has not yet been established in the literature. Moreover, we faced difficulty in examining respondents’ opportunity costs in detail, which are central to the concept of ability to pay in our research (Russell, 1996). Some CVM studies have proposed that the need to resort to debt to pay for a product or a service implies a lack of ability to pay (Habbani et al., 2003, 2006). These studies included respondents’ budget constraints in the scenario in order to ensure that the stated willingness-to-pay was indeed payable. Their approach was to improve the scenario rather than to deliberate the definition of ability to pay, which addresses one of the known CVM problems, namely “alternative expenditure possibility,” raised during the early development of CVM studies (Arrow et al., 1993). We took a similar approach for this research. Specifically, we defined and identified the maximum price that respondents can pay without resorting to debt as their WTP, thereby avoiding the inability to pay. In other words, we considered that the scenario of “resorting to debt” had to recognize respondents’ budget constraints. Then, we denoted this value as “WTPpayable,” which differs from the WTP determined by the conventional method. In the next step, we conducted CVM studies in the field by investigating households’ WTP as well as WTPpayable for a public health service. We chose Indonesia as our focus geographical area. Historically, like many low- and middle-income countries, Indonesia used to impose user fees for public health services and aims at achieving universal coverage. Although the Indonesian government is making efforts to increase access to health services for the poor, about half the population still lacks health insurance coverage, notably the informal sector (Rokx et al., 2009; Tangcharoensathien et al., 2011). The field research was conducted from November to December 2008 in the Batang district of Central Java. The district is a rural area and is representative of Central Java, where people mainly practice agriculture for a living. Our research site, “village A,” was officially designated as a disadvantaged village (Inpres Desa Tertinggal: IDT). Three types of public health services can be normally accessed: services of the midwife, the health center, and the hospital. The latter two allow hospitalization and are staffed by doctors, nurses,
Table 1 Summary statistics of variables. Variable
Summary statistics
Age Sex
Mean, years (SD) Male, % Female, % Years of schooling Mean, years (SD) Household size Mean, persons (SD) Prior experience by household member Village midwife Yes, % No, % Health center Yes, % No, % Hospital Yes, % No, % Knowledge of typhoid fevera Yes, % No, % Whether a household member Yes, % has been treated for typhoid fever No, % Satisfaction with care at public Satisfied, % health facilities Less satisfied, % Dissatisfied, % Annual household income Mean, Rpb (SD) Annual household income per head
Mean, Rp (SD)
Value 40.9 (12.7) 45.5 54.5 5.11 (3.15) 3.73 (1.54) 81.2 18.8 58.4 41.6 17.8 82.2 82.0 18.0 52.0 48.0 78.5 3.7 17.8 3 991 651 (6 034 936) 1 158 706 (2 137 004)
c
WTP for services by Village midwife Health center Hospital WTPpayabled for services by Village midwife Health center Hospital The WTP WTPpayable Village midwife Health center Hospital The (WTP WTPpayable)/WTP Village midwife Health center Hospital
Mean, Rp (SD) Mean, Rp (SD) Mean, Rp (SD)
56 139 (142 799) 147 960 (177 330) 369 005 (839 534)
Mean, Rp (SD) Mean, Rp (SD) Mean, Rp (SD)
21 572 (72 775) 61 903 (131 002) 110 311 (384 556)
Mean (SD) Mean (SD) Mean (SD)
34 567 (121 532) 86 057 (140 905) 258 694 (744 807)
Mean (SD) Mean (SD) Mean (SD)
0.505 (0.459) 0.657 (0.379) 0.690 (0.370)
a Respondents were asked “Have you ever heard of the disease named typhoid fever?”. b US$1 ¼ Rp12 195 (as of December 2008). c Willingness to pay. d The maximum price that prevents respondents from resorting to debt as their WTP.
33
0.070 0.187** 0.073 0.007 0.070 0.032 0.394** 0.268** 0.303** 0.319** 0.267** 0.315** 0.035 0.151* 0.240** 0.267** 0.110 0.096 0.006 0.063 0.018 0.266** 0.315** 0.304** 0.112 0.377** 0.284** 0.262** 0.021 0.165* 0.285** 0.305** 0.132 0.072 0.181* 0.214** **p < 0.01 *p < 0.05.
0.191** 0.081 0.207** 0.261** 0.127 0.312** 0.099 0.283** Age Years of schooling Annual household income Annual household income per head
Hospital
WTP WTPpayable WTP The (WTP WTPpayable WTPpayable)/WTP WTP
Health center
The (WTP WTPpayable)/WTP WTPpayable WTP WTPpayable WTP
Village midwife Variables
Table 2 Spearman’s correlation between WTP, WTPpayable, WTP WTPpayable, (WTP WTPpayable)/WTP, and continuous variables.
Fig. 1. Demand curves for typhoid fever treatment at different public health facilities in the study area. (a The maximum WTP and WTPpayable are Rp5 000 000.)
WTPpayable WTP The (WTP WTPpayable WTPpayable)/WTP
by taking into account the relation between household income and land ownership in Indonesia (Shiga et al., 2009), since self-reported income is considered to be unreliable in our field (Russell, 1996). Various approaches, such as payment cards, the take-it-orleave-it (TIOLI) method, and the bidding game method, have been applied in CVM studies on WTP in the health sector. Nonetheless, it is worth noting that all methodologies have some form of bias, such as the starting point bias in the bidding game method, the yessaying behavior in the TIOLI method, and the range bias in payment cards (Johannesson and Jonsson, 1991; Klose, 1999; McNamee et al., 2010). Which of these methods is more reliable in the context of in low- and middle-income countries remains controversial. Dichotomous choice methods such as the bidding game and TIOLI are often used in this type of research (Arrow et al., 1993; Cummings et al., 1995), with evidence of bias clearer in the former (Klose, 1999). Therefore, in this study we chose to employ the bidding game approach. We used a hypothetical scenario of the respondent contracting typhoid fever. This fever is a water- and food-borne infection caused by Salmonella enterica subspecies enterica serovar Typhi (S. Typhi) that continues to be a serious public health problem, especially in low- and middle-income Asian countries (Ochiai et al., 2008). Affected patients of all ages risk suffering from severe symptoms and complications without timely and appropriate care.
0.006 0.141 0.230** 0.261**
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
34
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 3 Mean WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP for public health services according to respondents’ characteristics. Variables
Village midwife Mean WTP
Age 29 30e39 40e49 50e Sex Male Female Prior utilization of village midwife by household member Yes No Prior utilization of health center by household member Yes No Prior utilization of hospital by household member Yes No Knowledge of typhoid fever Yes No Whether a household member has been treated for typhoid fever Yes No Satisfaction with care at public health facilities Satisfied Less satisfied Dissatisfied
p valuea
Health center Mean WTPpayable
0.597 75 38 49 61
870 137 022 864
761 765 152 924
0.085
23 792 18 452 0.627
66 667 53 855
0.224
0.219
0.996 56 190 56 082
24 262 18 660 0.654 23 817 22 143 14 559
187 083 139 424 0.000 161 982 87 639 0.636
0.520 0.489 0.731
39 517 7857 27 500
0.144
0.550
0.749
0.810
148 941 146 566
0.514 0.462
31 929 37 423
0.926
0.043
0.248
0.586
159 540 98 289
0.364 0.537
39 466 13 784
0.002
0.881
0.342
0.633 22 851 16 486
0.023 180 879 120 727
0.501 0.511
17 083 38 358
0.909
0.223
0.376
0.449 157 048 138 021
0.498 0.528 0.306 0.497
p value
130 569 761 603
0.257
0.486 0.591
28 157 43 571
49 583 15 497
62 317 30 270
63 333 30 000 42 059
25 351 74 342
Mean WTP
134 151 160 145
0.464 0.540 0.179
0.608
p value 0.795
0.189
0.483
0.622
Mean (WTP WTPpayable)/WTP 0.517 0.546 0.506 0.456
47 853 23 455
23 308 14 079
51 949 62 024
109 373 870 941
0.199
0.309
p value 0.566
55 21 35 28
28 777 15 545
48 659 88 421
Mean WTP WTPpayable
0.519 20 16 13 32
76 630 39 000
p value
0.669 147 633 114 286 172 206
a p values derived from t test for sex, prior utilization of a health service, knowledge of typhoid fever, and ANOVA for age, years of schooling, and satisfaction level with the health service.
The infection could burden a household with the high costs associated with its treatment (Poulos et al., 2011). Therefore, each respondent was provided with three hypothetical scenarios of types of care depending on the severity of typhoid fever: outpatient care provided by the midwife, hospitalization for three days attended to by doctors at the health center, and hospitalization for four days attended to by doctors at the hospital. We then provided the following five starting bids for each scenario: Midwife: Rp10 000 (US$0.82), Rp15 000 (US$1.23), Rp20 000 (US$1.64), Rp30 000 (US$2.46), and Rp40 000 (US$3.28) Health center: Rp60 000 (US$4.92), Rp90 000 (US$7.38), Rp120 000 (US$9.84), Rp180 000 (US$14.76), and Rp240 000 (US$19.68) Hospital: Rp160 000 (US$13.12), Rp240 000 (US$19.68), Rp320 000 (US$26.24), Rp480 000 (US$39.36), and Rp640 000 (US$0.82). At the start of the contingent valuation section, the respondent was asked whether s/he had ever heard of typhoid fever. If s/he answered “no,” the interviewer explained the common symptoms of typhoid fever to him/her (the scenario used in this study is given in the Appendix). In addition, s/he was also asked to assume that the required treatment/drugs were available at each health care facility for each of the three levels in the scenario. Secondly, s/he was provided with the first hypothetical scenario on outpatient care, that is, care provided by the midwife, and was randomly assigned to one of the five starting bids. The respondent was asked if s/he would be willing to pay the first bid. If s/he answered “yes,” a higher bid was provided to check the highest
possible price the respondent would be willing to pay. If the answer was “no,” a lower bid was provided to check the highest possible price the respondent would be willing to pay. We considered respondents’ WTP as the maximum bid they agreed to pay at the end of the interview. Next, we inquired about the price the respondent could pay without incurring any debts (i.e., their WTPpayable). We proceeded with the remaining two levels of care in the hypothetical scenario, that is, hospitalization at the health center and at the hospital, in the same manner. Thirdly, we estimated two types of demand curves using the results of each CVM study, that is, one based on WTP and the other on WTPpayable. By using the method outlined in McIntosh et al. (2010), we then drew two demand curves using the stated WTP and WTPpayable for the three scenarios. We also drew supply curves that were depicted with straight lines in line with our assumption that the respondent could avail of any of the three types of public health services should they wish to. In other words, it was assumed that appropriate health care would always be provided at the health facilities. Next, we explored the gap between WTP and WTPpayable as follows:
D ¼ WTP WTPpayable % ¼ ðWTP WTPpayableÞ=WTP If WTP was zero, we excluded such samples from our analysis for percentages. The excluded WTP cases were 10, 5, and 10 cases for the midwife, health center, and hospital services, respectively. We employed Spearman’s correlation in order to study the
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Health center
35
Hospital
Mean p value Mean WTP p value Mean (WTP p value Mean WTP p value Mean WTP p value Mean WTP p value Mean (WTP p value WTPpayable WTPpayable WTPpayable)/WTP payable WTPpayable WTPpayable)/WTP 0.454 57 73 37 75
065 137 500 216
0.570 77 78 123 70
065 431 261 388
0.055 82 720 44 682
0.292 98 159 76 045
0.002 69 954 27 368
0.190 98 194 53 985
0.895 88 889 85 439
0.585 64 497 51 250
0.000
0.567
0.049
100 214 70 573
0.349 0.681 0.630
0.725 85 750 67 143 104 853
0.294 0.684 0.522 0.591
79 119 144 427
association between the continuous variables, WTP, WTPpayable, D ¼ WTP WTPpayable, and % ¼ (WTP WTPpayable)/WTP. The t test and one-way analysis of variance (ANOVA) were used to ascertain whether WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP depended on their attributes or experiences. Furthermore, multiple regression analysis was used to examine the variables that influence WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP. We used SPSS V. 17.0 for the statistical analysis and considered differences significant at p < 0.05. This research protocol was approved by the Ethical Committee of the University of Tsukuba, Japan. The conduct of the field survey was approved by the State Ministry of Research and Technology, Indonesia. Respondents were interviewed with prior informed consent. 3. Results Table 1 shows the summary statistics of respondents’ characteristics. The mean age of respondents was 40.9 years and 45.5% of them were male. Their mean years of schooling and average household size were 5.11 and 3.73, respectively. Annual household income was Rp3 991 651 (US$327.3), while annual household income per head was Rp1 158 706 (US$95.0). We found mean individual WTP values of Rp56 139 (US$4.6), Rp147 960 (US$12.1), and Rp369 005 (US$30.3) for the services provided by the midwife, health center, and hospital, respectively. The corresponding mean individual WTPpayable values were Rp21 572 (US$1.8), Rp61 903 (US$5.1), and Rp110 311 (US$9.0), respectively. The corresponding
0.001
0.044 0.718 0.553
0.114 338 167 171 771
0.792 109 550 210 714 101 618
0.093 0.595 0.711
300 229 76 667 0.250
0.981 376 967 325 000 394 706
0.318
0.782
0.395
0.122 0.724 0.640
371 389 234 106
114 405 94 722
417 286 316 198
0.309
0.270
0.002
0.002 0.656 0.834
303 665 194 759
174 444 96 318
414 634 171 389
0.962
0.593
0.164
0.144 0.647 0.726
257 469 263 947
122 521 92 952
545 833 330 424
0.685 0.545 0.137
0.932 61 883 47 143 67 353
0.598 0.670
97 485 36 389
56 833 67 448
0.314
0.089
0.149
0.251
0.770 0.671 0.707 0.732 0.658
363 159 172 273
129 218 29 211
426 186 287 711
826 922 500 078
0.069
0.538
0.520
0.261 147 253 447 201
169 808 61 091
386 687 293 158
0.671 0.636
696 118 717 647
0.018
0.018
0.702
0.444 78 204 62 90
532 967 233 364
0.629 0.776
82 860 90 602
522 039 217 724
0.080
0.464
0.591
0.152 226 458 510 291
0.605 0.700
89 586 70 921
66 081 55 964
0.500 0.642 0.686 0.710 0.602
0.086 0.734 0.640
0.853 267 417 114 286 293 088
0.143 0.713 0.425 0.657
values of WTP WTPpayable were Rp34 567 (US$2.8), Rp86 057 (US$7.1), and Rp258 694 (US$21.2), respectively. The corresponding values of (WTP WTPpayable)/WTP were 0.505, 0.657, and 0.690, respectively. All the demand curves based on the WTP and WTPpayable values show gaps between respondents’ WTP and WTPpayable (log rank test: p < 0.001) (Fig. 1). The percentages of respondents whose WTP exceeded the approximate actual price, as shown by the supply line, were 70.8%, 24.4%, and 15.4% for the midwife, health center, and hospital, respectively. The corresponding percentages of respondents who indicated a WTPpayable exceeding the approximate price were 37.1%, 9.5%, and 4.5%. Table 2 describes the correlation between the continuous variables, namely WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP. The WTP values for the three services were correlated to years of schooling, while the WTPpayable values for the three services were correlated to years of schooling, annual household income, and annual household income per head. WTP WTPpayable for the midwife service was related to annual household income and annual household income per head, and for the hospital service, it was related to years of schooling. (WTP WTPpayable)/WTP for all three services was related to annual household income and annual household income per head, and for the midwife and health center services were also related to years of schooling. Table 3 provides the mean WTP, WTPpayable, WTP WTPpayable, and (WTP WTPpayable)/WTP, showing differences between groups for the three services. The mean WTP, WTPpayable, and WTP WTPpayable for the midwife service did
0.321 0.995 0.0160 0.012 0.280 1.221 0.006 2.769 0.0931 0.097 0.002 2.975 0.000 4.022 0.0200 0.187 0.000 5.231 0.001 3.310 0.0222 0.198 0.000 5.541 2.301 0.023
household. a
0.216
0.456
16 637 1.240
0.00430 0.746 0.084 0.004 2.679
0.00656 0.135 0.000 3.876
0.852 7003 0.187 0.777 22 248 0.284 0.803 2901 0.250 0.745 0.326 5083
0.514 15 359 0.654
8650 1.303 0.194
0.194 85 646 1.305 0.763 0.302 41 338 0.475 14 523 0.716 0.806 6711 0.246 0.647 0.519
0.442 0.771 24 371
7512
0.811 0.239 21 156 0.569 0.571 105 332 0.958 1432 0.052 0.151 52 979 12 099 0.773 0.441
1.442
0.340 0.957 71 297 0.262 1.126 174 982 0.087 1.720 39 583 0.303 1.034 31 995 2.499 0.013 32 955 0.772 0.290 7713
0.623 0.493 28 618 0.340 0.956 115 896 0.549 0.600 10 766 0.820 5496 0.228 0.394 0.694 0.461 15 330 0.738
4054
0.706 0.739 0.113 0.475 0.218 70 863 0.377 893 0.334 95 165 1.595 8464 0.715 90 582 1.236 0.976 0.949 0.083 0.208 0.737 11 631 0.03 354 0.063 216 904 1.741 31 221 1.264 51 363 0.336 0.805 0.980 0.237 0.069 0.093 14 339 0.247 20 0.025 21 880 1.186 6689 1.830 38 252 1.689 0.595 0.531 0.201 0.012 0.111 41 508 0.532 697 0.628 31 844 1.284 12 479 2.539 48 836 1.604 26 105 0.785 0.433 426 0.901 0.369 4077 0.386 0.700 5735 2.738 0.007 6650 0.512 0.609 0.714 0.455 0.289 0.000 0.047 24 690 0.368 715 0.749 22 717 1.064 15 892 3.756 52 552 2.005
Constant Age Sex (male:0, female:1) Years of schooling Prior utilization of village midwife by HHa member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic
p value Coefficient t value
WTPpayable WTP
p value Coefficient t value p value Coefficient t value p value Coefficient t value p value Coefficient t value Coefficient t value
Hospital
WTP
WTPpayable Health center
WTP
WTPpayable Village midwife Variables
not show any statistical significance between groups. However, the mean (WTP WTPpayable)/WTP for the same service showed a statistical significance between groups with and without a household member who had availed of the hospital service in the past. The mean WTP for the health center service indicated statistical significance between male and female respondents between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. The mean WTPpayable for the health center service showed a statistical significance between groups with and without a household member who had accessed the midwife service in the past. The mean WTP WTPpayable for the health center service showed statistical significance between groups with and without prior knowledge of typhoid fever. The mean (WTP WTPpayable)/WTP for the health center service showed statistical significance between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. The mean WTP for the hospital service indicated statistical significance between male and female respondents and between groups with and without prior knowledge of typhoid fever. The mean WTPpayable for the hospital service did not show any statistical significance. For the hospital service, the mean WTP WTPpayable showed statistical significance between groups with and without prior knowledge of typhoid fever. For the hospital service, the mean (WTP WTPpayable)/WTP showed statistical significance between groups with and without a household member who had accessed the midwife service in the past, and between groups with and without prior knowledge of typhoid fever. Table 4 shows the results of the multiple regression analysis on WTP and WTPpayable for public health services. The WTP for midwife services was significantly affected by years of schooling and prior utilization of midwife services by a household member, whereas the WTPpayable for this service was significantly affected by years of schooling, prior utilization of hospital services by a household member, and annual household income per head. The WTP for health center services was significantly affected by years of schooling and annual household income per head, and WTPpayable, by annual household income per head. Finally, the WTP for hospital services was significantly affected by annual household income per head, while WTPpayable was not significantly affected by any of the variables. Table 5 shows the results of the multiple regression analysis on WTP WTPpayable for public health services. For both the full and stepwise models, the WTP WTPpayable for midwife services was significantly affected by years of schooling, prior utilization of midwife services by a household member, and annual household income per head. For the health center service, the WTP WTPpayable assessed using the full model was not significantly affected by any of the variables. For the health center service, the WTP WTPpayable assessed using the stepwise model was significantly affected by the respondent’s knowledge of typhoid fever. Finally, the WTP WTPpayable for hospital services was significantly affected by annual household income per head in both the full model and the stepwise model. Table 6 shows the results of the multiple regression analysis on (WTP WTPpayable)/WTP for these services. The (WTP WTPpayable)/WTP for midwife services was significantly affected by prior utilization of the hospital service by a household member and annual household income per head in both models. The (WTP WTPpayable)/WTP for health center services was significantly affected by prior utilization of midwife services by a household member, prior knowledge of typhoid fever, and annual
p value
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 4 Results of the multiple regression analysis on WTP and WTPpayable for public health services.
36
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
37
Table 5 Results of the multiple regression analysis on WTP WTPpayable for public health services. Stepwise model
Village midwife Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic Health center Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic Hospital Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic a
Full model p valuea
Coefficient
t value
42 762
1.831
0.069
9922 58 691
3.196 2.630
0.002 0.009
0.00966 0.065 0.002 5.239
2.127
0.035
36 717
1.538
0.126
60 472
2.288
0.023
149 389
2.483
0.014
0.0943 0.068 0.000 14.47
3.803
0.000
0.022 0.023 5.233
Coefficient
t value
p value
50 836 288 18 643 10 150 59 199 19 379 25 241 36 468 22 867 7977 0.0108 0.061 0.020 2.193
0.879 0.350 1.014 2.785 2.621 1.083 1.100 1.338 1.131 0.690 2.185
0.381 0.727 0.312 0.006 0.010 0.280 0.273 0.183 0.260 0.491 0.030
26 310 702 10 036 5684 10 629 16 205 7556 54 385 7895 7983 0.00235 0.006 0.356 1.111
0.381 0.716 0.458 1.307 0.395 0.759 0.276 1.673 0.327 0.579 0.396
0.704 0.475 0.648 0.193 0.694 0.449 0.783 0.096 0.744 0.563 0.692
82 561 1249 121 738 22 774 39 218 87 270 103 689 84 108 126 996 15 253 0.0770 0.064 0.017 2.254
0.233 0.248 1.082 1.021 0.284 0.797 0.738 0.504 1.026 0.216 2.536
0.816 0.804 0.281 0.309 0.777 0.426 0.461 0.615 0.306 0.829 0.012
p (in) ¼ 0.05, p (out) ¼ 0.1.
household income per head in the full and stepwise models. For the hospital service, the (WTP WTPpayable)/WTP assessed using the full model was significantly affected by prior utilization of midwife services by a household member and prior knowledge of typhoid fever. Finally, for the hospital service, the (WTP WTPpayable)/ WTP assessed using the stepwise model was significantly affected by respondent’s years of schooling, prior utilization of the midwife service by a household member, and prior knowledge of typhoid fever. 4. Discussion In this study, we conducted interviews to assess the WTP and WTPpayable for three types of public health services in Central Java, Indonesia, using the conventional CVM and recognizing respondents’ budget constraints, respectively. Firstly, we defined the maximum price at which respondents can avoid resorting to debt as their WTPpayable. Specifically, we considered that the scenario of resorting to debt had to recognize the budget constraints of respondents.
We anticipated that each household’s WTPpayable for health services would be fixed irrespective of the scenarios. However, many respondents indicated different WTPpayable values depending on the three types of public health services examined. We believe that this is because we used debt to differentiate WTP from WTPpayable. Respondents set household priorities and changed their expenditure patterns according to the situation at hand, as most have potential resources such as land, livestock, and crops that could be converted into money (Russell, 1996; Wallman and Baker, 1996). In other words, they make discretional choices between selling assets and borrowing money. All demand curves based on the WTP and WTPpayable values show gaps. This finding suggests that WTP elicited by the conventional CVM is different to that from WTPpayable (Fig. 1). However, few CVM studies on WTP have mentioned whether the WTP is able to be paid, although many authors have shown that economic indicators such as income correlate positively with WTP (Dong et al., 2003; Foreit and Foreit, 2003; Frick et al., 2003). Therefore, the findings of CVM studies on WTP should be interpreted with caution. Moreover, to bridge the gap between WTP and WTPpayable elicited
38
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
Table 6 Results of the multiple regression analysis on (WTP WTPpayable)/WTP for public health services. Stepwise model
Village midwife Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head 2
Adjusted R p value F-statistic Health center Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head 2
Adjusted R p value F-statistic Hospital Constant Age Sex (female:1, male:0) Years of schooling Prior utilization of village midwife by HH member (yes:1, no:0) Prior utilization of health center by HH member (yes:1, no:0) Prior utilization of hospital by HH member (yes:1, no:0) Knowledge of typhoid fever (yes:1, no:0) Whether a household member has been treated for typhoid fever (yes:1, no:0) Satisfaction with care at public health facilities Annual HH income per head Adjusted R2 p value F-statistic a
Full model t value
p valuea
Coefficient
t value
p value
14.74
0.000
0.171
1.990
0.048
4.958 108
3.222
0.002
0.633 0.000 0.046 0.013 0.106 0.018 0.175 0.110 0.002 0.003 4.006 108
2.860 0.010 0.648 0.933 1.229 0.259 1.995 1.056 0.025 0.070 2.109
0.005 0.992 0.518 0.352 0.221 0.796 0.048 0.292 0.980 0.944 0.036
3.983 0.242 1.347 1.158 2.364 0.397 1.130 2.349 0.177 0.748 1.974
0.000 0.809 0.180 0.248 0.019 0.692 0.260 0.020 0.860 0.456 0.050
3.931 0.341 1.299 1.306 2.898 1.333 1.922 2.487 0.487 0.360 0.868
0.000 0.733 0.196 0.193 0.004 0.184 0.056 0.014 0.627 0.719 0.387
Coefficient 0.593
0.064 0.001 7.259
0.045 0.054 1.856
0.715
8.921
0.000
0.173
2.514
0.013
0.159
2.295
0.023
3.247
0.001
4.059 108 0.088 0.000 6.889
0.088 0.003 2.778
0.772
9.440
0.000
0.022 0.190
2.544 2.802
0.012 0.006
0.224
3.232
0.001
0.092 0.000 7.119
0.709 0.001 0.076 0.013 0.164 0.022 0.080 0.197 0.011 0.027 3.012 108
0.682 0.001 0.072 0.014 0.196 0.071 0.132 0.203 0.030 0.012 1.292 108 0.101 0.001 3.034
p(in) ¼ 0.05, p (out) ¼ 0.1.
by CVM, studies should improve the scenarios used in CVM studies on WTP. In particular, researchers must inform respondents that they are actually choosing to avail of a given health service at a given/ stated WTP and remind them that this would also reduce their other expenditures. This process might improve the reliability of estimating and valuing services (Arrow et al., 1993). Further, in studies examining respondents in low- and middle-income countries, there is a shortage of information regarding the source of the money for respondents’ stated WTP payments. Household income is limited, especially for the poorer population, although their potential income resources are diverse and unstable in contrast to developed countries (Donaldson, 1999; Russell, 1996). Therefore, WTP does not necessarily mean that this amount can be instantly paid (i.e., WTP s WTPpayable), which rather tends to be decided through their complex choices on how to spend their finite household income on health care and other competing needs (Russell, 1996; Waddington and Enyimayew, 1989). The empirical findings should be carefully interpreted when deciding on a new pricing strategy or quality improvements. If
prices are set in response to the results of CVM studies on WTP rather than WTPpayable, it is highly likely that poor and vulnerable groups would not benefit from service and quality improvement, or that repetitive payments for these services would impose an excessive financial burden on households, forcing them to borrow money to pay for medical treatment and thus pushing them even deeper into poverty (Poulos et al., 2011; Whitehead et al., 2001). Therefore, subsidizing this segment of the population may significantly correct the inequitable access to relatively high cost health services, although the issue of health finance policies is still relevant here. According to the presented results of the multiple regression analysis, for the WTP WTPpayable and (WTP WTPpayable)/ WTP values for midwife services (Tables 5 and 6, respectively), the gap is negatively affected by household income. Moreover, this gap is larger for the poor. That is, valuing or pricing health services based on the results of CVM studies on WTP rather than WTPpayable may increase the inequity of access to health services. For health center and hospital services, we must also be cautious
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
when contextualizing the results, because respondents’ economic status did not seem to significantly affect either WTP WTPpayable or (WTP WTPpayable)/WTP. In many low- and middle-income countries, policymakers are increasingly interested in social and community health insurance as a possible risk-sharing mechanism and a more equitable alternative to user fees (WHO, 2010). Moreover, many CVM studies on WTP have been conducted to help gain information to arrive at a premium-priced service (Asgary et al., 2004; Donfouet et al., 2011; Dong et al., 2003; Gustafsson-Wright et al., 2009; Onwujekwe et al., 2010). However, most such studies overlook whether respondents are able to pay for WTP, meaning that the objective of the insurance system, namely to correct the inequity in access to health services, would not be achieved. There is room for improving the scenario used in this study, which might have been too simplistic. First, the explanation of typhoid fever might have been too brief, and respondents that stated they knew about the disease might actually have confused typhoid fever with a different disease. A better description, such as a depiction of the disease with illustrations, could overcome this problem. Second, since an exemption scheme was available to the poor in the field, the scenario should have reflected this fact, which would have made the scenario more realistic for respondents. Third, it is important to explore other aspects of the budget constraint when designing a scenario. Since typhoid fever is a communicable disease, the risk of other household members falling ill may be added, or other typical types of major expenditure may be included in the scenario to minimize potential budget constraint bias (Mitchell and Carson, 1989). In the bidding game part of the interview, respondents were offered a higher bid if they accepted the first amount and a lower bid if they refused the first amount. We intended to increase/ decrease the amount in increments/decrements of the same interval for the next bid for each service by referring to the findings of previous research (McNamee et al., 2010). Practically, however, respondents were not offered the amount in increments/decrements of the same interval for the next bid after accepting the current bid, because there were large gaps between the rich and poor populations in our research site. While increasing/decreasing the interval systematically for richer respondents was a long and tedious process, the small interval was meaningful for poorer respondents. Therefore, we had to change the interval depending on the respondents. Future studies should aim to explore the size of the increment/decrement of the interval that might be adopted after respondents refuse each bid in the process of the bidding game in the context of low- and middle-income countries. Moreover, our survey was conducted in a situation where respondents could actually access public health services, and we explained this assumption to them. Hence, this research may be treated as a case study. Furthermore, we adopted typhoid fever as the disease of interest in the CVM survey. Therefore, we cannot assume that the results of this study also apply to other types of diseases, including other chronic diseases. Finally, we estimated annual household income by taking into account the relation between household income and land ownership instead of directly asking respondents to state their income levels. However, other methods, such as means tests, could be used to assess household income, which in low- and middle-income countries tends to be derived from diversified sources (Jehu-Appiah et al., 2010). 5. Conclusion The results of this study showed that CVM studies on WTP might not always clarify respondents’ demand, especially for poor and vulnerable groups in low- and middle-income countries. In
39
order to bridge this gap in the body of knowledge on this topic, studies should improve the scenario used in CVM studies on WTP. In particular, researchers must inform respondents that they are choosing to buy at a given WTP and remind them that this would also reduce their other expenditures. This process might improve the reliability of estimating and valuing health services. As a policy implication, it is important to note that information provided by CVM studies on WTP requires careful interpretation when used for pricing or valuing public health services, especially for poor and vulnerable groups in low- and middle-income countries, which might not be able to pay the stated WTP value. In this respect, the findings of this research can help poor and vulnerable populations avoid falling into the medical poverty trap when utilizing public health services. Competing interests None declared. Acknowledgment We appreciate the assistance provided by Mr. Farikhun Asror, Health Officer of Batang District, in introducing our team into the field area. We are also very grateful to the villagers for their participation in our survey. This study was funded by a grant from the University of Tsukuba. Appendix. The scenarios used in the CVM studies to estimate WTP and WTPpayable 1) Firstly, have you heard of the disease typhoid fever? If “no,” the symptoms of typhoid fever are described below. If suffering from typhoid fever, patients have a very high fever and feel a chill for several days. Anyone can suffer from the disease. In order to recover, treatment is needed. The following question is hypothetical. Please choose the answer that best suits what you think. 2) If you were suffering from typhoid fever, would you have the willingness to pay the first bid for the midwife service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of care. The service is assumed to be satisfactory. 3) How much can you afford to pay without going into debt? 4) Please consider the following situation. Although you visited the midwife, you were immediately transferred without any treatment by the midwife, namely without making a payment, to a health center in your subdistrict for hospitalization because of your condition. Would you be willing to pay the first bid to avail of the health center’s service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of hospitalization for three days. The service is assumed to be satisfactory. 5) How much can you pay without going into debt? 6) Please consider the following situation. Although you visited the health center, you were immediately transferred without any treatment by health center, namely without making a payment, to the hospital because of your condition. Would you be willing to pay the first bid to avail of the hospital’s service? (This question was repeated until the maximum price was reached.) The fee includes the total cost of hospitalization for four days. The service is assumed to be satisfactory. 7) How much can you pay without going into debt? Note: We formulated the question for WTPpayable in a slightly different manner from that used for estimating the WTP using the
40
A. Shono et al. / Social Science & Medicine 110 (2014) 31e40
conventional method (i.e., what can you pay without going into debt?). We confirmed that respondents understood the word “debt” as borrowing money from neighbors or moneylenders inside or outside the village through prior interviews with villagers including the village chief, educated villagers, and moneylenders in the research area.
References Arrow, K., Solow, R., Portney, P.R., Leamer, E.E., Radner, R., Schuman, H., 1993. Report of NOAA Panel on Contingent Valuation, pp. 1e67. www.darrp.noaa.gov/library/ pdf/cvblue.pdf (accessed 27.05.13.). Asgary, A., Willis, K., Taghvaei, A.A., Rafeian, M., 2004. Estimating rural households’ willingness to pay for health insurance. The European Journal of Health Economics 5, 209e215. Cummings, R.G., Harrison, G.W., Rutström, E.E., 1995. Homegrown values and hypothetical surveys: is the dichotomous choice approach incentive-compatible? The American Economic Review 85, 260e266. Donaldson, C., 1999. Valuing the benefits of publicly-provided health care: does ‘ability to pay’ preclude the use of ‘willingness to pay’? Social Science & Medicine 49, 551e563. Donfouet, H.P., Makaudze, E., Mahieu, P.A., Malin, E., 2011. The determinants of the willingness-to-pay for community-based prepayment scheme in rural Cameroon. International Journal of Health Care Finance and Economics 11, 209e220. Dong, H., Kouyate, B., Cairns, J., Mugisha, F., Sauerborn, R., 2003. Willingness-to-pay for community-based insurance in Burkina Faso. Health Economics 12, 849e 862. Foreit, J.R., Foreit, K.G., 2003. The reliability and validity of willingness to pay surveys for reproductive health pricing decisions in developing countries. Health Policy 63, 37e47. Frick, K.D., Lynch, M., West, S., Munoz, B., Mkocha, H.A., 2003. Household willingness to pay for azithromycin treatment for trachoma control in the United Republic of Tanzania. Bulletin of the World Health Organization 81, 101e107. Gustafsson-Wright, E., Asfaw, A., van der Gaag, J., 2009. Willingness to pay for health insurance: an analysis of the potential market for new low-cost health insurance products in Namibia. Social Science & Medicine 69, 1351e1359. Habbani, K., Groot, W., Jelovac, I., 2003. User fees policy in Khartoum state of Sudan: the ability to pay for the public health services and its relation with willingness to pay. Maastricht University, Mimeo. Habbani, K., Groot, W., Jelovac, I., 2006. Household health-seeking behaviour in Khartoum, Sudan: the willingness to pay for public health services if these services are of good quality. Health Policy 75, 140e158. James, C.D., Hanson, K., McPake, B., Balabanova, D., Gwatkin, D., Hopwood, I., et al., 2006. To retain or remove user fees? Reflections on the current debate in lowand middle-income countries. Applied Health Economics and Health Policy 5, 137e153. Jehu-Appiah, C., Aryeetey, G., Spaan, E., Agyepong, I., Baltussen, R., 2010. Efficiency, equity and feasibility of strategies to identify the poor: an application to premium exemptions under National Health Insurance in Ghana. Health Policy 95, 166e173. Johannesson, M., Jonsson, B., 1991. Economic evaluation in health care: is there a role for cost-benefit analysis? Health Policy 17, 1e23. Klose, T., 1999. The contingent valuation method in health care. Health Policy 47, 97e123.
Lagarde, M., Palmer, N., 2008. The impact of user fees on health service utilization in low- and middle-income countries: how strong is the evidence? Bulletin of the World Health Organization 86, 839e848. Le Grand, J., Propper, C., Smith, S., 2008. The Economics of Social Problems, fourth ed. Palgrave Macmillan, New York. McIntosh, E., Clarke, P.M., Frew, E.J., Louviere, J.J., 2010. Applied Methods of Costbenefit Analysis in Health Care. Oxford University Press, Oxford. McNamee, P., Ternent, L., Gbangou, A., Newlands, D., 2010. A game of two halves? Incentive incompatibility, starting point bias and the bidding game contingent valuation method. Health Economics 19, 75e87. McPake, B., Normand, C., 2008. Health Economics: An international perspective, second ed. Routledge, Oxon. Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Resources for the Future, Washington, D.C. Ochiai, R.L., Acosta, C.J., Danovaro-Holliday, M.C., Baiqing, D., Bhattacharya, S.K., Agtini, M.D., et al., 2008. A study of typhoid fever in five Asian countries: disease burden and implications for controls. Bulletin of the World Health Organization 86, 260e268. Onwujekwe, O., Okereke, E., Onoka, C., Uzochukwu, B., Kirigia, J., Petu, A., 2010. Willingness to pay for community-based health insurance in Nigeria: do economic status and place of residence matter? Health Policy and Planning 25, 155e161. Palmer, N., Mueller, D.H., Gilson, L., Mills, A., Haines, A., 2004. Health financing to promote access in low income settings: how much do we know? Lancet 364, 1365e1370. Poulos, C., Riewpaiboon, A., Stewart, J.F., Clemens, J., Guh, S., Agtini, M., et al., 2011. Cost of illness due to typhoid fever in five Asian countries. Tropical Medicine & International Health 16, 314e323. Rokx, C., Schieber, G., Harimurti, P., Tandon, A., Somanathan, A., 2009. Health Financing in Indonesia: A Reform Road Map. The International Bank for Reconstruction and Development/The World Bank, Washington D.C. Russell, S., 1996. Ability to pay for health care: concepts and evidence. Health Policy and Planning 11, 219e237. Russell, S., Fox-Rushby, J., Arhin, D., 1995. Willingness and ability to pay for health care: a selection of methods and issues. Health Policy and Planning 10, 94e101. Shiga, K., Shono, A., Kondo, M., Masuda, M., 2009, November. Indonesia no nouson ni okeru totisyoyuu to setaisyuunyuu no kankei: naka-java syu blora-ken wo zirei to shite. In: Koichi, Miyoshi (Ed.), The 20th Annual Conference. The Japan Society for International Development, Beppu, November 21e22. Stiglitz, J.E., Walsh, C.E., 2006. Economics, fourth ed. W. W. Norton & Company, Inc., New York. Tangcharoensathien, V., Patcharanarumol, W., Ir, P., Aljunid, S.M., Mukti, A.G., Akkhavong, K., et al., 2011. Health-financing reforms in southeast Asia: challenges in achieving universal coverage. Lancet 377, 863e873. Waddington, C., Enyimayew, K.A., 1989. A price to pay: the impact of user charges in ashanti-akim district, Ghana. The International Journal of Health Planning and Management 4, 17e47. Wallman, S., Baker, M., 1996. Which resources pay for treatment? A model for estimating the informal economy of health. Social Science & Medicine 42, 671e 679. Weaver, M., Ndamobissi, R., Kornfield, R., Blewane, C., Sathe, A., Chapko, M., et al., 1996. Willingness to pay for child survival: results of a national survey in Central African Republic. Social Science & Medicine 43, 985e998. Whitehead, M., Dahlgren, G., Evans, T., 2001. Equity and health sector reforms: can low-income countries escape the medical poverty trap? Lancet 358, 833e836. WHO, 2010. World Health Report 2010: Health Systems Financing: The path to universal coverage. World Health Organization, Geneva. Witter, S., Ensor, T., Jowett, M., Thompson, R., 2000. Health Economics for Developing Countries: A practical guide. Macmillan Education, Oxford.
