カテゴリー: Munenobu Ikegami

Index Based Livestock Insurance (IBLI) 09

　In my last 8 posts, we have been reviewing academic papers on IBLI. In this post, I would like to introduce Russell (2021), which is a much shorter document than academic papers but I have sentimental value with. I would like you to read his post before this post.
　The reason why I have sentimental value with his post is that not only it mentions me several times but it summarizes my work over 16 years from 2005 to 2021.
　My 3rd post https://apsf.jp/en/2020/11/04/everyday-policy-studies-no-en22/ mentions that I was so excited to go to Marsabit for household survey after a year of office work in Nairobi. The first sentence of his post mentions that I was very surprised by how dry Marsabit is. This video https://www.standardmedia.co.ke/ktnnews/ktn-prime/video/2000210938/pangs-of-neglect-how-marsabit-residents-battle-drought-water-scarcity-as-country-battles-covid-19 uploaded a week ago by one of major TV news companies in Kenya might give you an idea. Both Bubisa and Lontolio in the video are study sites for the IBLI Marsabit Survey. Here, I would like to add another surprise I got before I reached at Marsabit. At that time in 2009, the main road from Isiolo to Marsabit has not been paved with tarmac yet. I saw Marsabit was dry and scarcely populated but the most surprising for me might be the way our land cruiser went over corrugated dirt road with high speed.
　The seventh paragraph of his post mentions Ikegami et al. (2020) as social protection paradox but does not mention that I started working on it in 2005 when I started working as an research assistant for Michael Carter. It was also the paper I have submitted to a peer reviewed journal for the first time. I remember I was so excited when we submitted it.
　The fourth last paragraph of his post introduces Janzen et al. (forthcoming). We have started working on this paper in 2011. This paper follows up the previous paper by focusing index insurance as a tool to prevent households from falling into poverty traps. I am planing to summarize these two papers in my next post.

Reference
Ikegami, M., M. R. Carter, C. B. Barrett, and S. Janzen (2019) “Poverty Traps and the Social Protection Paradox” in C. Barrett, M.R. Carter and J.-P. Chavas eds. The Economics of Poverty Traps. University of Chicago Press. Chapter 6. pp.223-256 https://press.uchicago.edu/ucp/books/book/chicago/E/bo28559644.html
Janzen, S., M. R. Carter, M. Ikegami (forthcoming) “Can Insurance Alter Poverty Dynamics and Reduce the Cost of Social Protection in Developing Countries?” Journal of Risk and Insurance https://onlinelibrary.wiley.com/doi/full/10.1111/jori.12322
Russell, Alex (2021) “Insurance for Vulnerable Families Cuts Rural Poverty and the Cost of Aid by Half” https://basis.ucdavis.edu/news/insurance-rural-families-cuts-poverty-and-cost-aid-half Accessed on May 7th, 2021

（Author: Munenobu Ikegami）

Index Based Livestock Insurance (IBLI) 08

　The previous essay reviews Takahashi et al. (2016) which study the determinants of IBLI uptake in the first two sales periods in Borena, Ethiopia. In this essay, we will review Takahashi et al. (2020) analyzing IBLI uptake not only in the first two but also the following four sales periods.
　Let us take a look at descriptive statistics of the uptake first. The study households are the 458 households who were interviewed in the all of the four rounds of our annual survey from 2011 to 2015. As mentioned in the previous essay, the insurance coverage period of IBLI contract is a year but potential buyers can purchase IBLI twice in a year. Because of this feature, it might be better to study uptake as ratio of households having valid policies (households who purchase IBLI not only in the current sales period but also in the previous sales period) rather than ratio of households purchasing in the current sales period. The uptake ratio changed from 28, 42, 38, 36, 29, to 28 percents from the first to sixth sales period. The average amount of livestock insured among households purchasing IBLI did not change so much (2.5, 2.5, 2.6, 2.4, 2.0, 2.3 Tropical Livestock Unit, TLU) over the six sales periods. IBLI uptake did not grow as I wished.
　Let us move to determinants of the uptake. As the potential determinants of the uptake, Takahashi et al. (2020) pick up price (discount rate of insurance premium), previous own uptake, other households’ uptake, and vegetation condition (Normalized Difference Vegetation Index, NDVI). We find that the current discount rate affects uptake but the previous discount rate does not, implying there is no price anchoring effects and that price subsidy in the initial stage of insurance market development would not hurt uptake in the future when subsidy is removed. Vegetation condition also matters. Pastoralists purchase IBLI more likely when vegetation condition is bad and payout is more likely. This damages sustainability of insurance market and an idea for a solution is to change insurance premium based on current vegetation condition. The results on the remaining two potential determinants, other households’ uptake and previous own uptake, are not so clear in the sense that the results depends on whether we use discount rate as instrument variable for endogeneity of IBLI uptake.
　Payout or no-payout may affect uptake in the following sales periods and there was no payout until November 2014. Timu et al. (2018) study this and thus Takahashi et al. (2020) do not study this to avoid an overlap of research questions.

Reference
Takahashi, K., Ikegami, M., Sheahan, M., and Barrett, C. B. (2016). “Experimental Evidence on the Drivers of Index-Based Livestock Insurance Demand in Southern Ethiopia.” World Development, 78, 324-340.
Takahashi, K., Noritomo, Y., Ikegami, M., & Jensen, N. D. (2020). “Understanding pastoralists’ dynamic insurance uptake decisions: Evidence from four-year panel data in Ethiopia.” Food Policy. 95.
Timu, A. G., Gustafson, C. R., Ikegami, M., & Jensen, N. D. (2018). “Indemnity Payouts and Index Insurance Demand in Ethiopia.” Working paper.

（Author: Munenobu Ikegami）

Index Based Livestock Insurance (IBLI) 07

　The previous essay https://apsf.jp/en/2021/02/02/everyday-policy-studies-no-en28/ reviews Jensen, Barrett, and Mude (2017) as the second paper among the nine papers on impacts of IBLI. Instead of continuing to review the remaining papers on the impacts, we will start reviewing papers on uptake of IBLI.
　Takahashi et al. (2016) study the determinants of IBLI uptake in the first two sales periods in Borena, Ethiopia. The number of sample households is 474. 30% of the households purchased IBLI in the first sales period in August-September 2012. 18% of the household purchased IBLI in the second sales period in January-February 2013. Insurance coverage period is a year but 5% of the households purchased IBLI in the both sales periods. This complication is due to the fact that one does not have to insure all of her livestock (and even she does not have to prove that she has livestock). Due to this contract feature, she can purchase insurance for some of her livestock in August-September 2012 and she can purchase another insurance in January-February 2013 although her first insurance is still valid and will expire in September 2013.
　Uptake of index insurances has been low despite of enthusiasm in development industry and the numbers of IBLI uptake above are comparable to or even larger than other index insurances. What disappointed us more was that policy holders insured only 2.7 tropical livestock unit (TLU) on average although the 474 sample households have 14.7 TLU on average. That is, the policy holders insure only 18% (= 2.7 / 14.7) of their livestock. 1 TLU is equivalent to 1 cow, 0.7 camel, 10 goats or 10 sheep.
　In order to study not only correlation between household characteristics and IBLI uptake but also some determinants/causes of IBLI uptake, we generated external variation in price and knowledge of IBLI by some experimental tools. The first tool is insurance premium discount coupon, which we have already reviewed in one of the previous essays. The second one is comic, with which insurance extension agents can introduce IBLI to our sample households and the households can review by themselves. The third one is skit tape, with which insurance extension agents can play audio skit and introduce IBLI to our sample households though the audio skit. By randomizing provision of these 3 tools, we can generate external variation in potential determinants of uptake. Discount coupons allow us to study causality from insurance premium price to uptake and the all three tools allow us to study causality from knowledge to uptake.
　Takahashi et al. (2016) find the following. First, comic and skit tape improve knowledge of IBLI but the knowledge does not increase uptake. Second, lower insurance premium increases uptake as expected and insurance demand is sensitive to price. Third, lower insurance premium in the first sale period and increase in insurance premium from the first to second sale period do not decrease uptake in the second sale period. This implies that a policy encouraging uptake by subsidizing insurance premium and let potential policy holders learn the benefit of the insurance can be effective since it will not decrease uptake in the future when the subsidy is decreased or removed.
　In the next essay, we will review a paper or two which study IBLI uptake in a longer time span.

Reference
Jensen, N. D., Barrett, C. B., and Mude, A. (2017). “Cash Transfers and Index Insurance: A Comparative Impact Analysis from Northern Kenya.” Journal of Development Economics, 129, 14–28.
Takahashi, K., Ikegami, M., Sheahan, M., and Barrett, C. B. (2016). “Experimental Evidence on the Drivers of Index-Based Livestock Insurance Demand in Southern Ethiopia.” World Development, 78, 324-340.

（Author: Munenobu Ikegami）

Index Based Livestock Insurance (IBLI) 06

　My previous essay https://apsf.jp/en/2021/01/08/everyday-policy-studies-no-en25/ reviews Janzen and Carter (2019). In this essay, I will review Jensen, Barrett, and Mude (2017), which is the second most cited paper among the nine papers on impacts of IBLI, following Janzen and Carter (2019).
　One of the major differences between Jensen et al. (2017) and the other eight papers is that Jensen et al. (2017) study not only the impacts of IBLI but also a conditional cash transfer program called Hunger Safety Net Programme (HSNP). HSNP provides 2,150 Kenya Shillings (KSH, about 29 United States dollar, USD) to beneficiary households once two months since April 2009. Marsabit County is one of the poorest counties in Kenya and it happened to be the pilot county not only for IBLI but also for HSNP and our households data have beneficiaries for either or both programs.
　Following our notation, equation of interest is Y=a+bX+e where Y is outcome variable, X is explanatory variable, a is constant term, and e is error term. Jensen et al. (2017) study many equations which have different X and Y. Explanatory variable X is having IBLI or having HSNP. Outcome variable Y is livestock sale, livestock herd size, veterinary expenditure, herding livestock with satellite camps, income from livestock milk, loss of livestock, household total income, school absenteeism, and health (measured by mid-upper arm circumference (MUAC) of children aged from zero to five). Jensen et al. (2017)’s central question is whether IBLI and HSNP have impacts on these outcome variables of household behavior and welfare. In order to control endogeneity of having IBLI or HSNP, they use instrument variable method. Instrument variable for having IBLI is insurance premium discount coupon as reviewed in my previous essay. Instrument variable for having HSNP is household eligibility status for HSNP.
　Jensen et al. (2017) find that either IBLI or HSNP has impacts on all outcome variables except school absenteeism. Based on estimated impacts (magnitude of b), they calculate and compare impact to cost ratios for IBLI and HSNP. Both total program cost per beneficiary and impact are similar between IBLI and HSNP and thus impact to cost ratios are also similar. On the other hand, marginal cost (how much it will cost for the next additional one beneficiary) for IBLI is 10 times smaller. It is because both IBLI and HSNP needs large initial cost for developing insurance product and setting up system for insurance transaction or cash transfer. As for marginal cost, HSNP incurs cost of cash to transfer but IBLI does not. Impact to marginal cost ratio is thus larger for IBLI.
　Jensen et al. (2017) do not find synergy between IBLI and HSNP as additional impacts when households have both IBLI and HSNP. This is studied further by Jensen, Ikegami, and Mude (2017).
　In the next essay, I will review another paper on the impacts of IBLI or proceed to studies on IBLI uptake.

Reference
Janzen, Sarah and Carter, Michael R. (2019). “After the Drought: The Impact of Microinsurance on Consumption Smoothing and Asset Protection,” American Journal of Agricultural Economics, 101(3), 651-671.
Jensen, N. D., Barrett, C. B., and Mude, A. (2017). “Cash Transfers and Index Insurance: A Comparative Impact Analysis from Northern Kenya.” Journal of Development Economics, 129, 14–28.
Jensen, N., Ikegami, M., and Mude, A. (2017). “Integrating Social Protection Strategies for Improved Impact: A Comparative Evaluation of Cash Transfers and Index Insurance in Kenya.” The Geneva Papers on Risk and Insurance – Issues and Practice, 42(4), 675–707.

（Author: Munenobu Ikegami）

Index Based Livestock Insurance (IBLI) 05

　My previous essay https://apsf.jp/en/2020/12/09/everyday-policy-studies-no-en23/ reviews how to measure the impacts of the insurance using the household data and the insurance premium discount coupon as a instrument variable. In this essay, I will review previous studies on the impacts using this method.
　Following the notation of my previous essay, our equation of interest is Y = a + bX + e where Y is dependent variable, X is explanatory variable, a is constant term, and e is error term. In our case, since we are studying the impacts of the insurance, X is purchasing and having the insurance and for simplicity, let us say X takes one if a household has insurance and zero otherwise (instead of that X is how much livestock the household covers with the insurance). Y is outcome variable which insurance can improve, for example, Y can be how much livestock a household owns after a drought. We hope that the insurance will help households save their livestock from a drought and so we hope that b in the equation above is positive and large. As we discussed in my previous essay, a more able household buys the insurance more likely and she can save her livestock more regardless the insurance and thus we would overestimate the impact b. We will alleviate this overestimation using the insurance premium discount coupon as an instrument variable.
　There are seven papers in academic journals and two working papers studying the insurance impacts on using the household data and insurance premium discount coupon as instrument variable. Five out of the nine papers use IBLI Borena Ethiopia data and the remaining four use IBLI Marsabit Kenya data. Each of the papers studies different outcome variables Y and impact channel b. Based on Google Scholar https://scholar.google.com, the most cited paper among the nine papers is Janzen and Carter (2019).
　There was a drought in 2011 and an insurance payout in November 2011. In October 2011, Janzen and Carter (2019) ask households whether they would sell livestock and reduce meal in the following three months (October-December 2011) in order to cope with the drought in 2011. Note that Janzen and Carter (2019) ask households their expectation on their drought coping strategies in the future instead of what households did in the past. One of two outcome variables Y in Janzen and Carter (2019) is to sell livestock to cope with the drought and they expect and find b is negative. Our next question is how large this impact is. 27 percents of the households say that they would sell livestock and the insurance decreases this by 61 percentage points (from 27 percents to 11 percents). The other outcome variable Y in Janzen and Carter (2019) is to reduce mead to cope with the drought. They expect b is negative and find b is negative for the poor half of the households but b is zero (no impact) for the remaining rich half. 62 percents of the households say that they would reduce meal and the insurance decreases this by 49 percentage points (from 62 percents to 32 percents) for the poor half of the households.
　In my next essay, some of the remaining eight papers will be reviewed.

Reference
Janzen, Sarah and Carter, Michael R, (2019), “After the Drought: The Impact of Microinsurance on Consumption Smoothing and Asset Protection,” American Journal of Agricultural Economics, 101(3), 651-671.

（Author:Munenobu Ikegami)

Index Based Livestock Insurance (IBLI) 04

　My previous essay reviews the household data we collected to study insurance uptake and impacts. In this essay, I will review how to measure the impact using the data and insurance premium discount coupons.
　The major challenge in measuring the impact is endogeneity of insurance uptake. Our central question is how much insurance improves households’ welfare. The cause is insurance and the result is welfare. Studying this causality is difficult due to unobserved household characteristics and endogeneity of insurance uptake. To simplify this reasoning, let us say unobserved household characteristics is household ability and more able households buy insurance more likely (and endogenously). Let us call the latter endogeneity of insurance uptake. If so, we cannot tell whether insurance improves household welfare or more able households improve their welfare and buy insurance at the same time. In this case, if we do not control this endogeneity of insurance uptake, our estimate of the impacts of insurance on household welfare will end up larger than the true value.
　To control this endogeneity, we designed a randomized controlled trial (RCT). Since it is difficult to ask insurance companies to sell insurance to particular households or in particular villages and not to sell to the others, we provided insurance premium discount coupons to particular households among our 924 sample households in Marsabit, Kenya and 515 sample households in Borena, Ethiopia. We call households who received the coupons treatment group and the other households who did not receive the coupons control group. We randomly split the households into treatment group and control group. The simplest RCT compares average welfare in the treatment group with average welfare in the control group and the impacts is the difference in welfare. In our case, if we did so, the difference in average welfare would be the impacts of discount coupon rather than the impacts of insurance.
　In order to obtain the impacts of insurance, we use an Econometircs method called instrument variable (IV) method. Our dependent variable Y is household welfare such as income and explanatory variable X is insurance uptake. Our equation of interest is Y = a + bX + e where a is average welfare (called constant term) and e is other factors influencing welfare (called error term). Coefficient b is the estimate we would like to have and the impacts of insurance (X) on household welfare (Y). IV method allows us the following. Even if insurance uptake (X) is influenced by ability captured in error term (e), we can estimate coefficient b using an instrument variable (Z) which is strongly related to insurance uptake (X) but unrelated to error term (e). Let us consider whether discount coupon can be an instrument variable (Z). First, discount coupon encourages insurance uptake and thus our instrument variable (Z) can strongly related to insurance uptake (X). Second, we randomly split households into treatment group who receives discount coupons and control group who does not and thus our instrument variable (Z) should be unrelated to other factors affecting household welfare (error term, e).
In this essay, I review that insurance premium discount coupon allows us to measure the impact of insurance on household welfare. In the next essay, I will review studies on insurance uptake and impacts using the household data and the discount coupons.

(Author: Munenobu Ikegami)

Index Based Livestock Insurance (IBLI) 03

　My previous essay reviews the process from insurance design to pilot implementation. In this essay I will start reviewing how we study insurance uptake and impacts.
　Mude et al. (2009) proposes to study insurance uptake and impacts and we designed and implemented a baseline household survey in September 2009 before an insurance company started selling the insurance in January 2010. The number of sample households is 924 and they are located in 16 sublocations in Marsabit County in northern Kenya. Initially we had a research grant only for the baseline survey but we obtained several grants subsequently and could have annual follow up survey five times.
　The baseline survey in 2009 was new experience for me. Before that, I have been studying the data which were designed and collected by someone else. This was my first time to design and implement a survey. This was my first fieldwork as well. I have travelled in rural areas in developing countries but the length of each trip is less than two weeks and the purpose is to satisfy my curiosity rather than research or work. I remember I was very excited to go to Marsabit for the survey after a year of office work in Nairobi.
　We piloted the insurance in Borena Zone in southern Ethiopia as well. Borena Zone and Marsabit County are sharing the border and my colleagues had known the area from their previous research project. In Borena Zone, the baseline household survey was in March 2012 and the initial insurance sale was in August 2012. 515 sample households are located in 17 study areas. We could have annual follow up survey three times. Ikegami and Sheahan (2017, 2018) explains more details of each survey and the data are publicly available at http://data.ilri.org/portal/dataset/ibli-marsabit-r1 and http://data.ilri.org/portal/dataset/ibli-borena-r1.
　In my next essay, I will review how to measure the impacts of insurance using the data. Our main tool was insurance premium discount coupon.

References
Ikegami, M. and M. Sheahan (2017) “Index Based Livestock Insurance (IBLI) Marsabit Household Survey Codebook” http://data.ilri.org/portal/dataset/ibli-marsabit-r1
Ikegami, M. and M. Sheahan (2018) “Index Based Livestock Insurance (IBLI) Borena Household Survey Codebook” http://data.ilri.org/portal/dataset/ibli-borena-r1
Mude, A., C. B. Barrett, M. R. Carter, S. Chantarat, M. Ikegami, and J. McPeak (2009) “Index Based Livestock Insurance for Northern Kenya’s Arid and Semi-arid Lands: The Marsabit Pilot” http://ssrn.com/abstract=1844758

(Author: Munenobu Ikegami)

Index Based Livestock Insurance (IBLI) 02

　My previous essay https://apsf.jp/en/2020/09/16/everyday-policy-studies-no-en15/ review why and how my colleague designed Index Based Livestock Insurance (IBLI) in Northern Kenya. In this essay, I will review the process from insurance design to pilot implementation.
　When I joined IBLI project at International Livestock Research Institute (ILRI) as a post-doc in September 2008, my colleague has already designed the insurance. One of my tasks was to contribute to our efforts to modify the insurance design for pilot implementation. For example, area boundary for insurance premium calculation and area boundary for index calculation can be modified based on what an insurance company think as the best. Also, insurance premium changes as insurance payout trigger, deducible, and exit (maximum payout) change.
　We talked to several insurance companies in Kenya and several reinsurance companies as potential implementation partners for the insurance pilot. Fortunately, an insurance company and an insurance broker company in Kenya, and a reinsurance company in Europe agreed to pilot the insurance with us. Several international development donors also agreed to provide grants to support the insurance pilot including insurance premium subsidy. We agreed on that all of insurance parameters (premium, premium subsidy, trigger, deducible, exit, area boundary). The pilot area is Marsabit County (previously Marsabit District), area boundary for insurance premium calculation divides Marsabit County into two areas (Upper and Lower Masabit), and area boundary for index calculation divides Marsabit County into 5 areas (1. Central and Gadamoji, 2. Laisamis, 3. Loiyangalani, 4. Maikona, 5. North Horr).
　The insurance was launched in January 2010. My task to contribute to the pilot implementation continued since IBLI project team at ILRI is responsible to calculate the index and support insurance companies. In retrospect, my transition to a graduate student to a post-doc is quite a change. My study as a graduate students is academic. On the other hand, my work as a post-doc is development project implementation more than academic research. I wrote scripts to download remote sensing data, calculate index, upload update on index to the project web site. I wrote documents explaining how index is calculated for insurance companies and had meetings with them.
　In my next essay, I will move to the next topics: insurance uptake and impact.

(Author: Munenobu Ikegami)

Index Based Livestock Insurance (IBLI) 01

　Hello, my name is Munenobu Ikegami. This is my first essay in English in this forum. I have worked for Index Based Livestock Insurance (IBLI) project from September 2008 to February 2018 at International Livestock Research Institute (ILRI) and I would like to review what we learned in this essay series.
　Pastoralists in Northern Kenya and Southern Ethiopia face drought and associated risks that are large in magnitude (frequently causing a 20-40% livestock mortality rate) and frequent (once every 4-5 years). ILRI and its research and implementation partners launched a commercial IBLI product in January 2010 in an effort to mitigate the negative consequences of livestock mortality risk. Research on IBLI can be divided into three groups: 1) insurance design; 2) insurance uptake; and 3) insurance impacts. In this essay I will review 1) insurance design.
　Chantarat et al. (2013) explains the design. What we would like to insure is the negative economic shock due to drought and livestock mortality. Livestock is the second largest productive asset the pastoralists have following their own human capital. When drought occurs, all of forage and water for livestock, livestock milk production, birth rate would decrease and livestock mortality would increase. Note that decreased livestock herd size due to decreased birth date and increased mortality rate affect the pastoralists negatively even in the following seasons after the drought.
　Traditional insurance for livestock mortality would not be commercially viable due to large transaction costs due to that pastoralists are herding livestock in remote and large space and it is difficult to verify each animal’s death by a insurance company. Index insurance can overcome these problems. Instead of making indemnity payout based on each animal’s death like traditional insurance, index insurance makes indemnity payout based on an index which represents average loss in a region. IBLI applies this idea to livestock mortality due to droughts in the region.
　Index should have the following properties: 1) closely related with the loss to be insured; 2) not manipulated; 3) available timely; 4) constructed with low costs. The index Chantarat et al. (2013) provides is predicted livestock mortality based on Normalized Difference Vegetation Index (NDVI). NDVI is measurement on how green earth surface is based on satellite. Chantarat et al. (2013) shows that NDVI can explain livestock mortality due to droughts in the area in the past well and suggests to use predicted livestock mortality based on NDVI as the index for IBLI.
　In the next essay, I will continue reviewing the design of IBLI.

(Author: Munenobu Ikegami)

Reference
Chantarat, Sommarat, Andrew G. Mude, Christopher B. Barrett, and Michael R. Carter. 2013. “Designing Index-Based Livestock Insurance for Managing Asset Risk in Northern Kenya.” Journal of Risk and Insurance. Vol. 80, No. 1, pp. 205-237.