In the regions affected by Human Immunod efficiency Virus / Acquired Immune Deficiency Syndrome , prime-age adult mortality negatively affects household welfare by decreasing household income and consumption. Previous studies on the effects of prime-age adult mortality on household agricultural production show that the mortality decreases household size and productive assets such as land and livestock. In this study, we further ask whether prime-age adult mortality due to HIV/AIDS decreases the endowment of knowledge for agricultural production in Kagera, Tanzania, reducing total factor productivity. Equivalently, we ask whether prime-age adult mortality due to HIV/AIDS destroys household agricultural production by magnitude beyond the decreases in observed productive assets such as household members, land, and livestock. We also quantify how much decreased TFP growth contributes to the decrease in long-term household agricultural income growth compared to the decreased accumulation of each productive asset. Kagera was estimated to be one of the regions in Tanzania most affected by the HIV/AIDS epidemic , Beegle. Kagera is also the region where AIDS cases were reported first in hospitals in Tanzania. In 1983, the first 3 AIDS cases were reported and the number of cases increased rapidly to 5,116 cases in 1994. On the other hand, the share of reported AIDS cases in Kagera to Tanzania decreased from 100% in 1983 to 10% in 1994. In 2003, the percentage of HIV positive in Kagera among age 15-49 is 3.7% while the figure in Tanzania is 7.0% and thus HIV/AIDS pandemic in Kagera has been alleviated compared to other regions in Tanzania. We use the Kagera Health and Development Survey which collects the detailed information on households in Kagera in 1991-94 and 2003. The survey samples households hit by prime-age adult mortality more than households without the mortality and the data allow us to study the long-term effects of prime-age adult mortality on agricultural production. In the data, 36.7% of prime-age adult mortality is considered to be due to HIV/AIDS by deceased individuals’ families.
We will focus on agricultural production among other income generating activities and we will study the effects of prime-age adult mortality on agricultural production in the region. Agriculture is the major income source in Kagera and also in Tanzania. In Kagera, 85% of household heads engage in agriculture in 2000/01 while 70% in Tanzania , Tanzania NBS. In Kagera,grow lights households engage in subsistence and traditional agriculture. Male adult members produce coffee and banana with or without cattle manure. Female adult members produce crops such as maize and yams mainly for own consumption.As shown below, households hit by prime-age adult mortality between 1990 and 2003 have less increase in household members by 1 person from 1991 and 2003 than households without the mortality.They also accumulate less other productive assets; land and livestock. As a consequence, their agricultural income growth is also smaller. However, we do not find such clear differences in per capita asset accumulation and income growth between households with prime-age adult mortality and those without it. In order to explore the effects of the mortality on agricultural production more, we will study the difference in TFP growth. We study the hypothesis that a household hit by the mortality cannot increase TFP as much as a household not hit by it. We also decompose agricultural income growth into the contribution of the accumulation of each productive asset and TFP growth and compare the differences in those factors between households with and without the mortality. The remainder of this paper is organized as follows. Section 2 reviews the previous studies on the effects of prime-age adult mortality on households’ welfare based on household level micro data and the differences between the previous studies and this study. Section 3 outlines our conceptual model, hypothesis, and framework of empirical methods. Section 4 explains the characteristics of the original data, especially with respect to prime-age adult mortality, how we construct our data for the analysis from the original KHDS data, and discuss the relevancy of our specification of the model to study the data. Our empirical methods are explained in more details in Section 5 and the empirical results are shown and discussed in Section 6. Section 7 concludes this paper.Whether and how much HIV/AIDS epidemic affects a household welfare is the important topic. We can categorize the literature of the effects of prime-age mortality due to HIV/AIDS on household welfare into consumption studies and production studies. Beegle, de Weerdt and Dercon studies the effects of prime-age mortality on long-term consumption growth based on KHDS.
Their regression equations have change in logarithm of per capita consumption from 1991 to 2003 as the dependent variable and dummy variables for deaths as explanatory variables. They use household fixed effects methods in order to control unobserved time-invariant characteristics and relax the endogeneity and self-selection problem of HIV/AIDS as other previous studies based on panel data do. They take into account which year each death occurred by using dummy variables for deaths in 1991-1995, 1996-1999, and 2000-2004. Their results show that the coefficients of dummy variables for deaths are negative but only dummy variables for deaths in 2000-2004 are statistically significantly different from zero. This characteristics of the results are robust in various specification of regression equations. Their results imply that there are negative effects of prime-age adult mortality on consumption growth but households may recover from the negative shock of the mortality after 5 years. They find that a prime-age adult death results in a 7% drop in consumption in the first 5 years after the death. Carter, May, Aguero and Ravindranath use KwaZulu-Natal Income Study , South Africa data and study the effects of prime-age mortality due to HIV/AIDS on long-term growth rate of per capita consumption and find the negative coefficients for dummy variables for deaths although they are not statistically significantly different from zero. They also find the large magnitude of the negative effects: a prime-age adult death lowers a household’s 5 year growth rate by 21%. Although the consumption studies above find the negative effects of prime-age adult mortality on household 5-year consumption growth, channels of the causality has not been made clear. Production studies analyze some potential channels of the causality. Beegle uses the first 4 waves of KHDS from 1991 to 1994 and studies the short-term effects of prime age adult mortality in a household on the household members’ labor supplies. She constructs dummy variables for male and female deaths in future and past 0-6 months and 7-12 months and uses them as explanatory variables in regression equations. The dependent variables are the probabilities of being in wage employment, non-farm self-employment, working on coffee production, banana production or maize, cassava, or beans production. She finds coefficients of some dummy variables for deaths are negative and statistically significantly different from zero in regression equation of being in wage employment, working on coffee production and maize, cassava, or beans production.
Yamano and Jayne use two-year panel of rural Kenyan households and study the effects of prime-age adult mortality on households’ size and composition, crop production, asset levels and off-farm income. They find the mortality decreases households’ size, area under high-valued crops, gross and net outputs, farm equipment, small animals, and off-farm income. They find that the death of a male household head is associated with a 68% reduction in the net value of the household crop production implying large negative effects of the mortality on households welfare and that channels of the causality are decreases in productive inputs above. Chapoto and Jayne use nationally representative 3-year panel data in Zambia and find the results similar to Yamano and Jayne . These production studies show that the negative effects of prime-age adult mortality on household income and channels of the causality. HIV/AIDS also increases an household’s expenditure for medical care for the sick and funeral for the deceased. Tibaijuka finds that this expenditure is almost equivalent to the cash income for the 10 households in her data from Kagera, Tanzania. We can think that decreased income and increased expenditure for health care and funeral due to HIV/AIDS and prime-age adult mortality contribute to the decreased consumption which is found in the consumption studies above. Households hit by HIV/AIDS have to face tighter budget constraints and invest less in productive assets than the other households. Smaller investment in productive assets brings smaller income in the future. We contribute to the literature with the following three points. First, we provide an answer to the question whether prime-age adult mortality decreases total factor productivity in the long run. Previous studies do not ask this question although it is an important question to study the channels from prime-age adult mortality to decreased income and welfare. This question is closely linked to the question how important an adult’s knowledge stock of agriculture is for his/her household income generation. Since subsistence agriculture in Kagera, Tanzania depends on weather and is erratic,led grow lights the knowledge may be important. On the other hand, its agriculture is traditional and does not depend on new technologies and new market opportunities so much, the knowledge may not be important. If the knowledge is important, prime-age adult mortality destroys not only household members but also the quality of household as an agricultural enterprise. Second, we decompose the agricultural income growth into TFP growth and the contribution of each productive asset. Previous production studies analyze the effects of prime-age adult mortality on each productive asset separately and cannot show how much change in each productive asset due to prime-age adult mortality contributes to change in agricultural income or output.
We quantify this channel from change in each productive asset to change in agricultural income by estimating an agricultural production function and decomposing the long-term change in agricultural income growth into TFP growth and change in contribution of each productive asset for households with and without prime-age adult mortality. Third, we study the effects of prime-age adult mortality on long-term agricultural production and link the previous studies on long-term consumption with the previous studies on short-term change in production mentioned above in this section.We can categorize channels through which the mortality changes the investment decision into two: First, the household changes future asset accumulation path as a response to changes in current asset levels due to the mortality and inheritance. For example, the household may sell land and livestock in order to achieve efficient and smaller productive asset level as a response to decreased household members and productivity due to the mortality. Second, the household’s budget constraint becomes tighter due to the mortality and the household has to change its allocation of income into consumption and investment over time. The household lost labor for income generation since the member who was sick and deceased did not and will not contribute to the household as labor and other members take care of the sick and thus the household income decreases. Furthermore, the household faces expenditure for medical care and funeral. Tibaijuka finds that this expenditure is almost equivalent to the cash income for the 10 households in her study. Although there is no consensus on what adult age range we should use to study the effects of adult mortality on household welfare1, we set the age range for prime-age adults is from 15 and 50. In this subsection, we discuss the relevancy of this age range. Our focus is the effects of prime-age adult mortality on agricultural production. We will focus on prime-age adult’s death rather than other household members’ deaths since prime-age adults contribute to their household as main labor force for agricultural production and they are the age group who are affected by HIV/AIDS directly.We set the lower bound of prime-age adult to be 15 since 15 year old individuals is physically adult and start to face the risk of HIV/AIDS through heterosexual sex. Although under 15 year old children can contribute to their households with their labor, we do not think that decreasing the lower bound would change the results since most of them do not die due to HIV/AIDS shown below. On the other hand, we set the upper bound of the age range at 50. Figures 1 and 2 show the distribution of age by gender in the data. Figures 3 and 4 show the distribution of deceased individuals’ age by gender. Figures 5, 6, 7, and 8 show the distribution of age of deceased individuals due to HIV/AIDS .