The Fight Against Malaria: A New Index for Quantifying and Assessing Policy Implementation Actions to Reduce Malaria Burden in Sub-Saharan Africa

Abstract

We aggregate information from the World Malaria Reports on 61 different antimalarial policies to develop an extensive synthetic index (MaPI) for 44 SSA countries between 1990 and 2017. We observe an increase of convergence in polices’ implementation in the region, starting in more developed countries and reaching less developed countries from the mid-2000s. Using a difference-in-difference events study design, we find that prevention, diagnosis and treatment are key policies to reduce malaria mortality and prevalence: an increase of about 10 p.p. of these policies generates a mortality cumulative decrease of around 8 p.p. and a prevalence cumulative decrease of 13 p.p. after five years. We also use available data on policy coverage for a reduced subset of policies to construct an intensive margin version of the index. Main results are consistent to the ones derived from the extensive version. Finally, we prove the robustness of our results with a large battery of checks related to model specifications, econometric techniques, data sources and falsification tests.

Keywords:

• Antimalarial policies
• composite index
• malaria burden
• Sub-Saharan Africa
• external health
• event study design

Acknowledgment

We would like to acknowledge the editor and two anonymous referees for their valuable comments. We also thank Nidia García-Marín, José-Víctor Ríos-Rull, Fernando Perera-Tallo, David Pérez-Mesa and participants at the Economic Research seminar at U. Autónoma de Barcelona, the XXVIII Meeting on Public Economics, the 46 SAE meeting (Barcelona), for valuable comments and suggestions. We specially thank Raúl Santaeulalia-Llopis and Shaun da Costa for their helpful comments in these meetings.

Disclosure statement

No potential conflict of interest was reported by the autor(s).

Data availability statement

Data and replication codes of the paper are available (to bona fide researchers) on the web of the journal or upon request to the corresponding author. Any remaining errors are our own.

Notes

1 Many studies (e.g. Cutler, Fung, Kremer, Singhal, & Vogl, 2010; Kuecken et al., 2021; Lucas, 2010) point out that policy interventions are crucial to fight against malaria.

2 We classify policies in 7 categories incorporated in 2 pillars: pillar 1 contains 4 categories related to prevention and vectors’ control, diagnosis, treatment and case management, and surveillance measures; pillar 2 includes 3 categories related to antimalarial therapies, therapeutic efficacy tests and insecticide susceptibility bioassays.

3 Grouping by pillar and categories is a very convenient feature that allows us to focus on particular subsets of measures when needed. Moreover, this approach is flexible enough to consider alternative versions of the MaPI grouping together policies that, for instance, target vulnerable populations (pregnant women and children ages 5 and under) or focusing only on policies demonstrated to have the most significant short-run health impacts.

4 We have also considered a principal component approach to aggregate categories and pillars. Results are quite robust to alternative aggregation strategies (see S-Appendix I.4; Figure A1). Thus, all pairwise rank correlations are above 0.93 in all cases.

5 Information about policies’ implementation is only available from 2004 on; data on implementation from 1990 to 2003 has been obtained from available data on adopted policies (see S-Appendix II.1; Figure A2). We obtain the same conclusions in both cases, implemented, and adopted policies, before and after 2003. Hence, our analysis for implemented policies should not be affected by this extrapolation.

6 In 2003, the bottom-five countries in policies’ implementation are Congo Republic, Liberia, Somalia, Equatorial Guinea, and Mauritania, whereas the top-five are Zambia, Gabon, Ethiopia, South Africa, and Comoros.

7 Results defining the event as “any increase” or as “any decrease” are available upon request. As commented in Main Appendix, estimated lagged coefficients (for MaPI and Pillar 1) show the expected signs (negative for “any increase” and positive for “any decrease”), but they are non-significant in most cases. For Pillar 2, results are noisy and inconclusive. Moreover, many policy increases coincide in the same event window, which hinders a correct identification of the event’s impact on within-country malaria mortality.

8 Our results are consistent with the literature that concludes that certain policies (such as ITN, IRS, etc.) can alleviate malaria burden and will end up reducing malaria mortality (see, for instance Yang et al., 2018; Howard et al., 2000; Kim, Fedak, & Kramer, 2012).

9 We further show that the inclusion of dynamic regressors, $\Delta X_{i,t-j},$ does not remove the autoregressive structure of the error term in Equation (1)(1) $$\Delta ln\left(y_{i,t}\right)={\lambda }_i+{\delta }_t+\sum _{j=-j_0}^{j_1}{\gamma }_j{D}_{i,t}^j+v_{i,t};i=1,\dots ,44;t=1990,\dots ,2017,$$(1) (i.e., the conditional-within country malaria mortality growth rate still shows a significant inertia), hence, as a robustness check, we include lagged terms of the endogenous variable in (1), $\Delta ln\left(y_{i,t-1}\right).$ We lag $\Delta ln\left(y_{i,t}\right)$ 3 and 4 periods to reduce endogeneity concerns. Both coefficients are highly significant, which illustrates the existence of autoregressive structure in the residuals.

10 This result is consistent with the literature emphasizing that antimalarial policies affect especially to children under 5 years old (Konaté et al., 2020; Telemu, 2012; Biteghe-Bi-Essone et al., 2022).

11 If we use Pillar 2, we obtain that lagged coefficients are negative but none of them are individually significant; only the contemporaneous and the first lead terms are close to being significant at 10%. Moreover, for all lags and leads, we cannot reject that all terms are zero in this case. In the case of the MaPI, we obtain similar results as Pillar 1. However, they are noisier than Pillar 1 since MaPI contains Pillar 2 information.

Additional information

Funding

This paper has received financial support from the Ministerio de Economía y Competitividad of Spain through project PID2019-107161GB-C33, and from Gobierno de Canarias through the ProID2017010088 (María del Carmen Betancourt y Molina program) R&D project, co-funded by the Operative Program FEDER 2014–2020. This work was supported by Consejo Superior de Investigaciones Científicas.