CHAPTER 11. COST-EFFECTIVENESS ANALYSIS
11.3 Effectiveness
383. The MVP had multiple goals, but these goals can be essentially reduced to three core ones:
1. To raise income through increased agricultural productivity, market development, microfinance and better infrastructure;
2. To improve health through development of a better health system delivering primary care at village level with referral to secondary and tertiary care through improved diagnostics, the provision of transport and electronic communications;
3. To improve educational achievements through improvements in schooling.
384. All three goals are intertwined. For example, better incomes provide better nutrition, leading to improved health, which in turn leads to lower workplace and school absenteeism. Education can also become cheaper for families through the provision of low-cost educational materials, while rising family incomes lower the need for child labour and increase the relative returns to schooling.
385. A precise mapping of interventions to results (and thus costs to effects) within any given sector is difficult to achieve. This is because we were given sectoral costs and broad cost categories which do not map to a specific result. Thus, we cannot say precisely that an amount was spent to achieve a result, and in any case for a multisector intervention, this would only be used for benchmarking.
Instead, the important effect that we should expect is the synergy that all productive activities are produced more efficiently. We nonetheless point to some heuristic observations that sectoral costs undertaken to achieve results within MVP far exceed what would be spent had some other means been undertaken. One should note also that the lack of closely comparable projects in the literature leaves questions as to whether the observed results could truly have been delivered at such low costs given the limited pre-existing system-level infrastructure in health and education in the project area.
We discuss this further, below.
11.3.1 Income
386. In addressing income gains, we answer two questions: (i) what was the overall income growth rate attributable to the MVP? And, (ii) is there any pattern in the growth rate of income? In response to the first question, as noted in Chapter 5, the MVP achieved significant income gains, which are summarised in Table 47. Using the DD approach to compare the MVs with the CVs, total gains for
Itad in association with IDS, PDA Ghana and LSHTM Back to contents Page | 152
the period 2012–16 averaged US$141 per capita (2012 PV) which have been transformed from the results reported regarding adult equivalent gains in Chapter 5.
Table 47. Gains in income over 2012–16
Av. DD effect Fixed effects Gains (Cedi) Cedi/Dollar Gains (US$)
2013* 0.27 123 2.09 59
2014 0.22 100 3.20 31
2015* 0.29 132 3.81 34
2016** 0.29 132 3.95 33
Total (US$) 157
Total (2012 PV US$) 141
* P-value = 0.1, ** P-value = 0.05
Notes: Dependent variable: Adult equivalent income / baseline standard deviation. Average adult equivalent adjustment:
5.285. Baseline standard deviation: 592. Values are per capita.
387. To ensure comparability with costs for the purpose of the CEA, income gains are quoted in US$ terms.
The value of US$141 may only slightly understate the local purchasing power parity (PPP) in the villages. This is because exchange rate fluctuations reflect not just inflation alone but also current trade balance and a future assessment of export-import potential. The PPP downward fluctuation is lower than the unfavourable exchange rate fluctuation. Thus, there is a bit of downward over-adjustment in the exchange rate in terms of purchasing power.
388. For the CEA, we consider three sectors that were primarily targeting increases in income: agriculture, infrastructure and community development. Summing the per capita costs in each of these three gives a total of US$160 per capita to mid-2016, slightly higher than the observed benefits. This suggests that, while income gains were achieved, there may have been a negative rate of return on investment, and thus limited cost-effectiveness. This assessment is potentially severe as one does not expect infrastructure or community development to affect income alone and further impact may occur in the future. Agricultural activities alone contributed mostly to the rise during the last few years. On the other hand, one should expect such factors as health to also affect income.
Apportioning costs from a complex project with multisectoral outcomes towards isolating costs of production of a particular outcome is fraught with conceptual difficulties. For example, Banerjee et al. (2015) for a poverty graduation programme implemented in various low and middle-income countries, report from their Ghana site that they achieved a benefit cost ratio of 133%. Thus, to produce US$141, a per capita amount of US$ 106 per capita would be invested.
389. While it is possible that MVP activities were successful in increasing income and earned a positive return, it would be highly difficult to adjust for the over-valued exchange rate in the past. In that case, the baseline dollar value would be lower, and, correspondingly, the DD measure would be higher.
390. Overall, in terms of income gain trends, it is difficult to say with any certainty whether the MVs have shifted to a higher growth trajectory, given the limited data points available for the 2012–16 period.
The absence of significant changes in expenditure or consumption despite the income gains remains a puzzle (as discussed in Chapters 5 and 9), and this analysis suggests that income gains are not being viewed as permanent, and that patterns of growth are not likely to have changed.
11.3.2. Health
391. Estimating the cost-effectiveness of health interventions is made complicated by the holistic approach taken to improving the health system as a whole, as well as the limited time within which impacts are to be observed.
Itad in association with IDS, PDA Ghana and LSHTM Back to contents Page | 153 392. Facilities. The MVs saw improvements in health infrastructure relative to the CVs in terms of both
physical infrastructure (buildings, rooms, vehicles and storage facilities) and staff (including a range of medical professionals from midwives to nurses to CHWs). While these facility improvements are significant, measuring their effectiveness is more challenging, particularly over a relatively small sample size and limited period of time.
393. Mortality rates. While many health outcomes will be measurable only in the longer term, we would still have expected improvements in IMR and U5MR over the period of the evaluation. Nevertheless, no significant improvements were observed under either measure.
394. Vaccinations. A total of 832 additional vaccinations were carried out as a result of the MVP.
According to UNICEF (2016) vaccine cost estimates, the cost of the vaccines themselves should have amounted to around only US$60 per year under the MVP (Cooke et al. 2016). Once the necessary infrastructure (cold storage, delivery, staff etc.) is factored in, however, UNICEF estimates that the total unit cost may be around US$32 per vaccine. This would lead to a cost-effectiveness threshold of US$26,624 for the delivery of 832 vaccinations. Although we cannot say exactly how much of the MVP costs are associated with vaccinations, it is notable that this cost-effectiveness threshold amounts to just 0.7% of the total MVP health budget.
395. Malaria. There was an attributable decrease in the incidence of malaria of 5.71% across the MVP area, which translates to an estimated prevention of 891 cases. In a systematic review of malaria prevention costs, White et al. (2011) find the median cost per case to be US$25. In light of this, we assume a cost-effectiveness threshold of US$22,275 for the prevention of the 891 cases. While we cannot say exactly what proportion of the MVP budget contributed to malaria reduction, it is notable that US$22,275 would amount to just 0.6% of health spending to date – this suggests that the cost of reducing malaria under the MVP may have been significantly higher than the median value observed in White et al. (2011). Any impact of reducing malaria is also likely to yield economic benefits. Laxminaryan (2004) reported that a 10% decline in malaria resulted in 0.3% increase in overall household expenditure. Under the MVP, however, no increase in household expenditure was observed. Usually, expenditure is a result of increased income; while the latter was observed, the former was not in the MV area.
396. Stunting. Across the MVs, 467 fewer children showed stunting over the five-year period. Fink et al.
(2014) find that every unit dropped in the height-to-age z-score (HAZ, a standard measure of stunting) is associated with 0.6 to 0.7 lost years of schooling, while each year of schooling is worth US$1,450 to US$2,000 in potential future incomes. As such, we estimate the MVP gains from reduced stunting to be in the range of US$0.4 million to US$1.1 million (equivalent to 10–28% of the total MVP health costs to date).168. No unit cost estimates were found in the literature relating to stunting.
Although this is a health impact it may not be due only to health interventions. It could be due to a lower malaria rate and perhaps some reduction in food insecurity as noted earlier in our evaluation.
It could also be that households with higher incomes pay more attention to their children’s needs, both in terms of time and making purchases.
397. Contacts with health workers. Over the period, the MVP achieved 48,500 more contacts with health workers. It is expected that this increased coverage will lead to improved health outcomes in future.
The quality of care provided to date is unclear. Doherty and Govender (2004) found that estimated costs per contact ranged from US$2 to US$7. Taking the midpoint of US$4.50, we arrive at a cost-effectiveness threshold of US$218,250 for 48,500 contacts. Once again, while we cannot say exactly how much of the MVP budget can be associated with these 48,500 contacts, the cost-effectiveness threshold of US$218,250 is equivalent to just 6% of the total MVP health budget, suggesting that the cost per contact may have been significantly higher than costs observed elsewhere. Further, the combined years of women between the ages of 15 and 49 contracepting over the 4.5 years was
168 Lower bound: one-unit HAZ improvement x 0.6 years x US$1,450 x 467 children. Upper bound: two-unit HAZ improvement x 0.7 years x US$2,000 x 467 children.
Itad in association with IDS, PDA Ghana and LSHTM Back to contents Page | 154
2,112. Programmes observed in sub-Saharan Africa can achieve the contraception protection noted above for US$60,000 in 2001 (Levine et al. 2006). While no cost estimates were available relating to stunting, the estimated potential benefits are substantial. The expected benefits from reduced stunting would only cover 10–28% of the total health costs. The above analysis, combined with a lack of evidence of improvements in the IMR and U5MR, suggest that the total health costs to mid-2016 are difficult to justify on cost-effectiveness grounds.
In summary, and with the caveat noted above regarding sectoral costs, most of the health achievements could have been achieved through other interventions spending about
US$500,000–600,000. This amounts to one eighth the cost of the what was spent by MVP on health.
11.3.3 Education
398. As with health, the MVP education interventions sought to intensively improve the education system as a whole, thus making assessment of the cost-effectiveness ratio difficult, given a lack of comparable projects. The full impact of system-wide changes may take longer than the five years in which the current evaluation took place. The MVP area, however, saw a considerable increase in educational inputs such as increase in number of teachers, text books and equipment.
399. Facilities. In terms of overall improvements to the education system, there were significant improvements in the MVs in terms of number of classrooms, provision of school supplies and number of qualified teachers. Teacher salary payments became more regular, and student-to-teacher and functioning classrooms ratios improved. Participation in parent–teacher associations has increased by 17%, and wage expectations are now higher in the MV areas. However, teacher training remains limited, which may owe to the recruitment of older and more experienced teachers under the MVP, who would require less training at present. The number of children travelling more than 3 km to school also remains high, while the number of schools being built has decreased since 2012.
400. Additional years of schooling. The main positive impact has been the increased number of school years at the primary level. The number of years of schooling per child gained through the MVP in this age group is 0.21 years, or a total of 2,153 additional years of schooling in total. This was observed through asking direct questions about school attendance. Taking the Fink et al. (2014) and our own estimated gains in income per additional school year, we have estimated a projected future income gain of US$1,450 to US$2,000 per year of schooling. Applying this to all those who achieved an extra year of education, we may obtain a total income increase of US$3.1 million to US$4.3 million. This compares favourably with the total cost of US$1.6 million spent on education to date under the MVP.
However, when comparing this with other projects, we find a range of values in the literature for the cost of retaining a child in school for a year. Baird et al. (2009) reports a cost of US$5 and Evans et al. (2009) find a cost US$91. While this is a wide range, these figures fall far below the MVP cost of US$736 per additional year of schooling.169
401. Although the returns to education appear significantly positive, in cost-effectiveness terms it seems that less intensive projects have been able to generate additional years of schooling at a far lower cost than the MVP.
402. In summary:
• Across the key impact areas of income, health and education, MVP compares unfavourably with other projects in the cost-effectiveness of service delivery.
169 US$1.6 million divided by 2,153 years of schooling.
Itad in association with IDS, PDA Ghana and LSHTM Back to contents Page | 155
• Returns to investment in education appear to be highest, although it is believed that similar results could have been achieved at a significantly lower cost.
• The cost-effectiveness of income and health-related interventions is highly questionable.
• Nevertheless, impacts are expected to continue to arise in future years. At the same time, a lack of precise comparators means the difficulties (and required costs) of delivering such a programme in northern Ghana may be understated in the CEA.