Intra-Household Resource Allocation under Negative Income Shock: A Natural Experiment

Intra-Household Resource Allocation under Negative Income Shock: A Natural Experiment

World Development Vol. 66, pp. 557–571, 2015 0305-750X/Ó 2014 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev http://dx.doi.org/1...

2MB Sizes 1 Downloads 18 Views

World Development Vol. 66, pp. 557–571, 2015 0305-750X/Ó 2014 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2014.09.012

Intra-Household Resource Allocation under Negative Income Shock: A Natural Experiment KHONDOKER A. MOTTALEB a,b, SAMARENDU MOHANTY a and ASHOK K. MISHRA c,a,* a International Rice Research Institute (IRRI), Manila, Philippines b International Maize and Wheat Improvement Center (CIMMYT), Gulshan-2, Dhaka, Bangladesh c Louisiana State University, Baton Rouge, USA Summary. — Using households from coastal districts in Bangladesh, as a case, and applying the difference-in-difference estimation method, this paper demonstrates a gender-differentiated outcome of negative income shocks for education expenditures of households. The cyclonic disaster that reduced crop income substantially increased the demand for labor as well as wages for rebuilding damaged farms. This in turn increased the opportunity costs of boys’ schooling, as reconstruction is a male-friendly sector. Consequently, parents withdrew their sons from school and engaged them in their households’ repair work and/or in wage-earning activities. However, girls’ schooling expenditure was unchanged in the affected farm and non-farm households. Ó 2014 Elsevier Ltd. All rights reserved. Key words — cyclone, income shocks, farm household, gender-differentiated impacts, intra-household resource allocation

1. INTRODUCTION

This indicates the necessity of undertaking more countryspecific empirical studies with well-defined variables and new datasets to clearly understand whether the intra-household resource allocation to health and education under negative income shocks conditional on gender is warranted. This paper attempts to fill that gap by examining the intra-household resource allocation behavior of both farm and non-farm households to health and education in the face of negative income shocks caused by natural disasters. This particular case involved tropical cyclone “Aila,” which hit the coastal region of Bangladesh on May 25–27, 2009. By matching this natural disaster with the government’s Household Income and Expenditure Survey (HIES) data (HIES 2005 and HIES 2010) and applying the “difference-in-difference” estimation approach in a natural experimental setting, this article examines both farm and non-farm households’ expenditure behavior separately for their male and female family members on food, health and education, and participation in non-farm income activities that particularly boomed because of the reconstruction of houses and farms after the cyclone. This article shows that “Aila” caused enormous losses in crop income for farm households located in the affected areas and forced them to reallocate resources within the household to cope with the loss in crop income. Because of the loss of the rice crop, farm households had to allocate more resources to ensure food security by spending more on food. Although health-related expenditures did not reveal any bias, this article confirms that the cyclone-affected households with at least one male child in high school and above reduced their expenditures on boys’ schooling, while girls’ schooling expenditures were unchanged in the affected farm and non-farm households compared with the unaffected households. This article also confirms that, as the cyclone increased the demand for male labor in post-disaster mitigation and recovery construction relative to female labor, farm households allocated more male members to non-farm income-generating activities. This factor actually

There is a growing concern among development economists and policymakers on whether negative income shocks can generate gender-differentiated impacts on intra-household resource allocation to health and education (World Bank, 2012). A clear understanding of household choices relating to gender-conditional expenditure on health and education under negative income shocks is critical not only to ensure household welfare but also to attain gender parity as per the Millennium Development Goals (MDG) of the United Nations. A question arises: If there are gender-differentiated impacts on intra-household resource allocation, particularly to health and education, what is the direction of the bias? More specifically, if households need to reduce health and education expenditures in the face of negative income shocks, do they reduce expenditures more for females or males? Unfortunately, only a few empirical studies address this issue (e.g., Cameron & Worswick, 2001; Rose & Al-Samarrai, 2001; Tansel, 2002). These studies do not reach a consensus as to whether a negative income shock has a greater impact on males or females. For example, in response to a reduction in household budget in Turkey (e.g., Tansel, 2002) and to crop failure in Indonesia (e.g., Cameron & Worswick, 2001), it was mostly girls who were pulled out of school. By contrast, it was mostly boys who were pulled out of school in Ethiopia in the face of the economic crisis (Rose & Al-Samarrai, 2001) and, in Coˆte d’Ivoire, boys’ school enrollment fell more than girls’ in response to drought (Jensen, 2000). There are also a few contrasting findings that demonstrate that, if economic shocks lower the opportunity costs of schooling, this can actually bring students of both genders back to school. For example, a reduction in coffee price in Nicaragua (Maluccio, 2005) and a financial crisis in Argentina (Lo´pez Bo´o, 2010) led to an increase in school participation of boys from rural areas. These examples point to the fact that the nature, location, and duration of the shock, as well as the tradition and sociocultural factors of a country (World Bank, 2012), determine whether negative income shocks adversely impact males or females.

*

557

Final revision accepted: September 21, 2014.

558

WORLD DEVELOPMENT

explained the lower education expenditure on boys than on girls in the cyclone-affected areas. The novelty of the study is threefold. First, the study highlights the impact of income shock on the income of households engaged in both farm and non-farm sectors in Bangladesh. Note that, in Bangladesh, agriculture is the principal source of livelihood of its burgeoning population, where more than 45% of the total 54.1 million economically active labor force is directly engaged in agriculture (GOB, 2014). Ironically, as Bangladesh is mostly a low-lying delta with a long coastal area, floods, droughts, and cyclones are recurrent phenomena that generate frequent negative income shocks to farm households (del Ninno, Dorosh, & Smith, 2003; Khandker, 2007; Paul, 1998; Paul & Rashid, 1993; Sarker, Alam, & Gow, 2012). For example, because of seasonal floods alone, the average annual loss in rice production in India and Bangladesh amounts to more than 4 million tons (IRRI, 2010). Unlike developed countries, farm households in Bangladesh do not receive any direct government support that can help them to absorb these shocks. This article thus intends to provide important insights into the gender-differentiated impacts of negative income shocks on health and education expenditures, particularly of the farm households in a developing country, where negative income shocks caused by natural disasters are more frequent. Second, this study also investigates the impact of a unique female education stipend program in Bangladesh on intrahousehold resource allocation to health and education. In the four decades since its independence in 1971, Bangladesh has achieved a remarkable improvement in reducing gender differences, particularly as reflected in social, political, cultural, and educational attainments or attitudes. For example, the average number of children per woman has decreased from seven to two; girls’ school enrollment has increased dramatically; and, since 1990, female labor participation has doubled (World Bank, 2012). For all these reasons, Bangladesh is ranked the highest in terms of gender gap index among all Islamic as well as South Asian countries (World Bank, 2012). Importantly, mainly to delay adolescent girls’ marriage and motherhood and to increase the school attainment of female students, the government of Bangladesh, jointly with the World Bank and Asian Development Bank, initiated the Female Secondary School Assistance Project (FSSAP) in 1994. Under this program, each girl in grades 6–10 receives tuition-free education, book allowances, a monthly stipend, and free secondary school examination. Although a few studies already examine the impact of the FSSAP on female educational attainment (e.g., Khandker, Pitt, & Fuwa, 2003; Raynor & Wesson, 2006), this study also indirectly assesses the impact of the female stipend program on female education expenditure by a farm household under negative income shocks. Finally, although this paper focuses on Bangladesh, the circumstances that are examined are closely replicated in millions of households across South Asia and sub-Saharan Africa. Like Bangladesh, most of these developing countries depend on agriculture for their livelihoods, where natural disasters frequently cause negative income shocks to farm households. This juxtaposition of agricultural income volatility caused by natural disasters with households’ dependence on agriculture highlights the similarity between Bangladesh and many other agrarian economies. This paper therefore intends to suggest policies based on the case study of farm and non-farm households in Bangladesh that can be applied in many other developing countries. The rest of the paper is organized as follows. Section 2 describes the study area and the selection of experimental

and control groups. Section 3 presents the sources of data and sampling methods and the characteristics of the sample households. Section 4 discusses the empirical model and presents the estimation results. Section 5 contains the conclusions and policy implications. 2. STUDY AREA, RESEARCH DESIGN, AND EXPERIMENTAL AND CONTROL GROUPS Bangladesh is one of the countries in the world most prone to natural disasters. Particularly, tropical cyclones are common in Bangladesh from March to November. From 1960 to 2009, a total of 45 major cyclonic storms hit Bangladesh, causing severe damage to human life and property (e.g., BBS, 1999, 2011). For example, on April 29, 1991, the cyclone that hit Chittagong Division severely affected 19 districts and 102 subdistricts, killed 0.14 million and injured 1.39 million people, and damaged 0.13 million acres of cropland (BBS, 2011). Also, the intrusion of saltwater into bodies of fresh water destroyed freshwater fish across the coastal districts. Unfortunately, the majority of the cyclones strike from April to November, during monsoon season, when the sea level in the Bay of Bengal is usually higher than average. This is also when major wet-season rice crops (Aus and Aman) are in the field. Consequently, the strong winds in the coastal area, together with the higher sea level and heavy monsoon rain, increase the severity of cyclones on lives, crops, and property. On May 25, 2009, Aila, a tropical cyclone, severely hit the coastal divisions of Khulna and Barisal, and a few districts in Chittagong Division. The wind blew at 70 kph minimum and 90 kph maximum, which created a tidal surge as high as 4–6 ft (BBS, 2011). The tidal surge, along with the heavy rain and strong winds, flooded farmlands and smashed embankments, roads, schools, and houses. A total of 190 people were killed almost instantly and several thousand were injured. According to the report of the International Federation of Red Cross and Red Crescent (IFRCRC, 2010), cyclone Aila affected more than 3.9 million people across the coastal districts. The tidal surge and heavy rain caused the heaviest impact, destroying coastal embankments and leading to the intrusion of saline water from the sea into farmlands, which damaged the late Boro rice and Aman rice seedlings and plants and destroyed thousands of households’ homes and property in the area (e.g., Schiermeier, 2014). Even five months after the cyclone, nearly 0.20 million people had not regained access to their houses and were still living in temporary shelters. Table 1 presents the number of people affected (in thousands) by division and district. According to data obtained from the IFRCRC and the Information Technology for Humanitarian Assistance Corporation and Action (ITHACA, 2009), cyclone Aila severely affected the districts of Barguna, Barisal, Bhola, Jhalokathi, Patuakhali, and Pirojpur in Barisal Division; the districts of Bagerhat, Khulna, and Satkhira in Khulna Division; and the district of Noakhali in Chittagong Division, a total of more than 0.1 million people at the minimum (Table 1). In contrast, Comilla, Chandpur, Feni, Laskmipur, Chittagong, and Cox’s Bazar districts of Chittagong Division also felt the effects of the cyclone, but only a few thousand households were negatively affected (Figure 1). Using this natural disaster, we classified the sample households from the severely affected districts, where at least 0.1 million households were affected by Aila, as the treatment group and the sample households from the partly affected districts as the control group. It was assumed that the probability

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

559

Table 1. Number of affected people in sampled districts and sampled households by cyclone Aila Experimental design

Experimental group

Division

Barisal

Khulna

Chittagong Total Control group

Chittagong

No. of affected people (000)a

Severely damaged districts

Barguna Barisal Bhola Jhalokati Patuakhali Pirojpur Bagerhat Khulna Satkhira Noakhali Partly or not affected districts Comilla Chandpur Feni Laskmipur Chittagong Cox’s Bazar

Total

No. of sampled householdsc 2005

2010

390 357 128 302 293 359 480 363 170 116 2958

74 118 88 68 88 64 72 122 89 97 880

95 132 109 89 93 76 107 170 125 119 1115

94 91 30 29 13.53b 29.45b 286.98

158 99 71 73 197 73 671

185 90 74 65 268 87 769

Data sources: a ITHACA. (2009). Bangladesh-Tropical Cyclone Aila-affected areas as of May 27, 2009. Online: http://reliefweb.int/sites/reliefweb.int/files/resources/ FAADD2B8FAD44CA0C12575C4004229E3-map.pdf, accessed June 10, 2012. b IFRCRC. (2009). Bangladesh: Cyclone Aila, June 2, 2009. Online: http://www.ifrc.org/docs/appeals/09/MDRBD004_OU2.pdf, accessed June 11, 2012. c Sources: BBS Household Income and Expenditure Surveys 2005 & 2010.

of being hit by cyclone Aila was the same for all the sampled districts. Table 1 also presents the number of sampled households, by district and sampled years, conditional on whether they belong to the experimental or control group. These sampled households and their demographic and other information on household income and expenditures for this study were collected from the HIES data collected in 2004–05 (2005), and 2009– 10 (2010), 1 which were made available by the BBS (Bangladesh Bureau of Statistics), government of Bangladesh. BBS used a two-stage stratified random sampling to ensure greater precision. In the first stage, more than 500 primary sampling units (PSUs) were selected across the country; in the second stage, 20 households were selected randomly per PSU to represent rural, urban, and statistical metropolitan areas (SMAs). In the 2005 HIES, a total of 10,080 households were randomly selected from 6 divisions, 64 districts, and 364 subdistricts. Finally, in the 2010 HIES, a total of 12,240 households were randomly selected from 6 divisions, 612 PSUs, 64 districts, and 384 subdistricts. In this study, however, as we are particularly interested in examining whether negative income shock generates any gender-differentiated impacts on household non-farm income activity participation, food expenditures, and expenditures on education and health, we consider both farm and non-farm households that were located in the coastal districts in Bangladesh. 2 This restriction is mainly intended to capture the effect of negative income shocks caused by natural disasters on households that were strictly located in the coastal area, and were equally exposed to cyclonic disasters. Including households from non-coastal districts might introduce problems of parameter heterogeneity into estimated results, as the chance of being hit by a cyclone is lower for off-shore households than for households in the coastal districts. Thus, essentially, we considered only 3,435 farm and non-farm households located in Barguna, Barisal Bhola, Jhalokathi, Patuakhali, and

Pirojpur districts of Barisal Division; Bagerhat, Khulna, and Satkhira districts of Khulna Division; and Noakhali, Comilla, Chandpur, Feni, Laskmipur, Chittagong, and Cox’s Bazar districts of Chittagong Division. Table 1 shows that, out of 3,435 households, 1,995 were located in severely affected districts (experimental group) and the rest were from unaffected or partly affected districts (control group). (a) Basic demographic and other information on sampled households Table 2 presents the mean of variables at the household level in different years, conditional on whether the households were affected by tropical cyclone Aila and also whether the households are farm or non-farm households. 3 Note that we define a household as a farm household if it produced a crop at least on one parcel of land in 2005 and 2010. Table 2 shows that, out of a total of 1,345 farm households, 831 were from cyclone-affected districts, and, out of a total of 2,090 non-farm households, 1,164 were from cyclone-affected districts. Table 2 indicates that, in the affected and unaffected and farm and non-farm groups, on average, the head of the household was a male, older than 47 years, with 3.23–5.23 years of formal education. The same table also shows that years of schooling of the spouse ranged between 2.20 and 4.64 years, and household size averaged close to five members, with a minimum of 2.46 being females and 2.49 being males. Table 2 also presents information on the average number of children per household who were going to school in the different grades (6–10), and college or above, and the percentage of households with at least a female and male child in high school or above. Importantly, Table 2 also presents the percentage of children in high school or beyond who received a full tuition waiver. It shows that at least 3.58% of both farm and non-farm households in cyclone-affected groups and more than 3.62% of both farm and non-farm households from the unaffected group received

560

WORLD DEVELOPMENT

Figure 1. Map showing the districts severely affected by cyclone Aila in Bangladesh on May 25, 2009. Source: http://www.ithacaweb.org/media/maps/ ITHACA_TC_”AILA”_COSMO_Water_30may.jpg, accessed June 25, 2012.

a full tuition waiver for their children in the secondary school and above level of education. Table 2 also presents the mean values of crop income per farm and non-farm households, yearly total income, food and non-food expenditures, cropland, value of households’ durable assets, and expenditures according to gender on health and education in the sampled years, conditional on whether the household was affected by the cyclone. Total income, annual, consisted of income from crops, livestock, poultry, forestry, fishery, transfers, remittances (both foreign and domestic sources), and agricultural rent. Total nonfood expenditures, annual, consisted of expenditures on all items, except food, for example, expenditures on fuel, housing, clothes, education, and health. Total food expenditures consisted of expenditures on all food items (except spices) in two weeks. The annual health expenditures consisted of doctor’s fee, cost of hospitalization, and medicine. The annual education expenditures consisted of registration fee, examination fee, private tuition fee, and hostel (boarding) expenses. The value of durable household assets consisted of articles such as a radio, camera, tape recorder, motorbike, microwave oven, refrigerator, cutlery and crockery, and carpet. Note that all of the monetary

values were computed in terms of real Bangladeshi taka (BDT) using GDP deflator 1995/96 = 100. Table 2 also illustrates crop values. It shows that, on average, a sample farm household in a cyclone-affected district cultivated 1.87 acres of land (2.5 acres = 1 ha) and earned a total crop income of about BDT 15,480 in 2010. In contrast, a farm household from an unaffected district, on average, cultivated 1.66 acres of land and earned roughly BDT 19,880 in 2010. Although the average size of the total farmland tended to be larger for farm households in cyclone-affected districts than for those in the unaffected districts, Table 2 shows that, in 2005, crop income was higher for the cyclone-affected farm households; however, crop income in 2010 for farm households in cyclone-affected districts was significantly lower. The effect of the cyclone is also similar in the case of non-farm households, in which crop income and total income of the non-farm households in cyclone-affected districts in 2010 were lower compared with 2005 and non-farm households located in the unaffected districts. Note that cyclone Aila hit Bangladesh on May 25–27, 2009. May is the time for late Boro, dry-season rice harvesting, and for Aman, rainfed rice transplanting. The losses in Boro and Aman rice (major crops in

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

561

Table 2. Mean values of different variables at the sample household level in different years conditional on those affected by tropical cyclone Aila on May 25–27, 2009 Household characteristics Group

Farm household Affected districts

Non-farm household

Not or partly affected districts

Affected districts

Not or partly affected districts

Year

2005

2010

2005

2010

2005

2010

2005

2010

No. of households Years of schooling, household head Years of schooling, spouse Age, household head % Female-headed households Total female members in the household Total male members in the household No. of children in high school No. of children in college or above % of children receiving full waiver of tuition fee in high school or above % of households with a male child in high school or college % of households with a female child in high school or college Crop income (BDT 000) Yearly total income (BDT 000) Persons engaged in non-farm sector Total food expenditures (BDT 000) Annual non-food expenditures (BDT 000) Health expenditures, females (BDT 000) Health expenditures, males (BDT 000) Education expenditures, males (BDT 000) Education expenditures, females (BDT 000) Cropland owned (acres) Durable asset value (BDT 000) Real wage rate at the subdistrict level (BDT)

330 4.16 3.67 47.92 0.91 2.76 3.03 0.63 0.17 12.72 53.03 52.42 18.62 68.60 0.70 1.05 25.47 0.61 0.74 1.73 1.22 2.68 9.84 65.71

501 4.08 3.49 47.76 4.19 2.60 2.68 0.59 0.24 7.79 51.50 52.50 15.48 58.34 0.75 1.15 30.31 1.41 1.44 2.17 1.66 1.87 7.23 77.42

266 3.30 2.20 50.89 5.26 3.24 3.50 0.86 0.15 11.27 57.89 59.77 16.43 77.02 0.78 1.38 29.38 0.83 0.75 1.52 1.01 1.97 13.95 68.75

248 3.23 2.73 51.75 9.27 3.06 3.23 0.72 0.22 3.62 55.65 53.63 19.88 89.03 0.68 1.37 38.38 2.00 1.37 2.47 1.83 1.66 11.54 80.70

550 4.81 4.03 45.74 7.64 2.61 2.69 0.54 0.21 7.27 48.73 49.82 0.11 67.42 1.16 0.95 21.83 0.87 0.82 1.54 1.26 0.00 11.39 66.37

614 4.91 4.27 44.65 5.86 2.46 2.49 0.48 0.22 3.58 43.16 42.67 0.01 62.91 1.22 1.06 33.01 1.32 0.93 2.45 1.90 0.00 12.03 77.90

405 4.95 4.60 45.96 10.62 2.87 2.99 0.59 0.20 8.89 48.15 48.89 0.01 87.79 1.17 1.29 39.16 1.21 0.87 2.42 1.91 0.00 24.30 70.35

521 5.23 4.64 45.84 10.75 2.77 2.70 0.47 0.20 3.45 41.46 42.80 0.03 87.99 1.21 1.28 48.90 1.38 1.16 3.99 2.93 0.00 20.27 79.94

Note: All monetary values are computed in terms of real BDT using general price index 1995–96 = 100.

Bangladesh) caused by the cyclone could be the primary cause of the crop and yearly total income reduction in 2010 compared with 2005 for the cyclone-affected farm households. The findings indicate the severity of the cyclone that caused losses in total crop income of the affected farm households. The severity of the cyclone damage is also reflected in the values of farm household durables in 2010 compared with 2005 and compared with those of non-farm households in cyclone-affected districts. Table 2 shows that the value of household durables in 2010, for the farm households, in affected districts plunged drastically compared with that in 2005 and with farm households in unaffected districts. By contrast, annual household income and the value of household durables in 2010 for non-farm households in unaffected districts increased significantly in 2010 compared with 2005. Interestingly, the annual income in 2010 for the non-farm households in affected districts also decreased drastically compared with 2005 and with that in non-farm households in unaffected districts. This provides a unique opportunity to examine the gender-differentiated expenditure behavior of both farm and non-farm households under a similar type of negative income shock regime. The last few rows in Table 2 present the number of persons engaged in non-farm income-generating activities, total food expenditures, and yearly nonfood expenditures, including separate figures on household health and education expenditures for male and female members, conditional on being affected by the cyclone and affiliated to a farm. Although, in general, all of the sampled farm and non-farm households’ food and nonfood expenditures, including the expenditures on health and

education, had increased from 2005 to 2010, the food and non-food expenditures by cyclone-affected farm and non-farm households had increased much more dramatically in percentage terms compared with other households located in the unaffected districts. To clearly depict this picture, Table 3 was developed, which presents the percentage changes of some of the major variables during 2005–10 conditional on whether a household was affected by the cyclone and whether a household is a farm or non-farm household. Table 3 shows that, in the case of farm households located in a cyclone-affected district, from 2005 to 2010, crop income and yearly total income decreased by nearly 15% at the lowest, while food expenditures increased by nearly 10% overall. In the case of farm households in the unaffected districts, from 2005 to 2010, real food expenditures actually decreased while crop and yearly total income increased. The same can also be observed in the case of nonfarm households in the affected districts compared with unaffected districts. Importantly, overall annual non-food expenditures of all of the sampled households, in both affected and unaffected districts, increased in 2010 compared with 2005. In particular, health expenditures increased by 94.59% for male family members and by 131.15% for female family members, and education expenditures increased by 25.43% for male family members and by 36.07% for female family members. By contrast, during the same time period in the case of farm households from unaffected districts, while non-food expenditures increased by 30.6% overall, health expenditures for male family members increased by 82.67% and for female family members by 140.9% (however, education expenditures did

562

WORLD DEVELOPMENT

Table 3. % Changes in mean values of some of the variables of interest from 2005 to 2010 at the sample household level conditional on farm and non-farm affiliation and those affected by tropical cyclone Aila on May 25–27, 2009 Change

% Change (2010–2005)/2005

Characteristics

Farm household

Non-farm household

Status

Affected

Partly or not affected

Affected

Partly or not affected

Crop income (BDT 000) Yearly total income (BDT 000) Persons engaged in non-farm sector Total food expenditures (BDT 000) Annual non-food expenditures (BDT 000) Health expenditures, females (BDT 000) Health expenditures, males (BDT 000) Education expenditures, males (BDT 000) Education expenditures, females (BDT 000) Durable asset value (BDT 000)

16.86 14.96 7.14 9.52 19.00 131.15 94.59 25.43 36.07 26.52

21.00 15.59 12.82 0.72 30.63 140.96 82.67 62.50 81.19 17.28

90.91 6.69 5.17 11.58 51.21 51.72 13.41 59.09 50.79 5.62

200.00 0.23 3.42 0.78 24.87 14.05 33.33 64.88 53.40 16.58

Note: All monetary values are computed in terms of real BDT using general price index 1995–96 = 100.

increase by 62.5% for male family members and by 81.19% for female family members). For both farm and non-farm households in cyclone-affected districts, the major sources of increased non-food expenditures were extra expenses for the reconstruction of cyclone-damaged houses, farmland, seedbeds, terraces, and business premises. In addition, the health-related expenditures of both farm and non-farm households in affected districts increased drastically compared with households in unaffected districts, mainly because of the spread of water-borne diseases after the floods. Besides direct casualties, the spread of water-borne diseases in cycloneaffected districts (the result of the contamination of drinking-water sources, such as tube wells, with floodwater) raised the households’ health-related expenditures for both male and female members. Table 3 shows that, compared with the unaffected farm households, the cyclone-affected farm households spent relatively less in 2010 on the education of male family members than female family members than they had in 2005. For example, the average expenditures on the education on male and female members of a farm household in affected districts increased in percentage terms by 25% and 36%, respectively, from 2005 to 2010. On the other hand, in unaffected districts, the average farm household’s expenditures on the education of male and female members, respectively, increased by 62% and 81% from 2005 to 2010. These findings indicate that, within farm households, in cyclone-affected districts compared with unaffected districts, education expenditures on male family members in the cyclone-affected districts have not increased much in light of the increase in education expenditures on female family members in 2010. The opposite scenario can be observed in the case of non-farm households located in the cyclone-affected districts, where the overall education expenditures on both male and female family members were lower in 2010 than in the non-farm households in the unaffected districts, and male education expenditures were higher in percentage than female education expenditures. In our econometric model, we tried to determine the factors that influence the contrasting behavioral outcome of the farm and non-farm households. Note that there are two major reasons for the reduced education expenditures on male family members by farm households in the cyclone-affected districts. First, a cyclonic disaster usually increases the demand for workers in the construction and restoration sectors, which are by nature more male-friendly. An increase in wages in the construction and repair sectors after a disaster increases the opportunity costs

of schooling for boys rather than for girls. The findings in Table 3 confirm that, on average, the number of persons engaged in non-farm income-earning activities in farm households, in 2010, increased by 7% compared with the same in 2005 and also compared with non-farm households during the sampled period. Second, after a disaster, farm households’ need for manual outdoor labor increases, as they must rebuild their seedbeds and reconstruct bunds and terraces on the farmland. Traditionally, households in Bangladesh prefer male workers for outdoor farm labor. Thus, after the cyclone, when the demand for labor increased, households withdrew their boys from school and engaged them in family farm work, while letting the girls continue their schooling. The female secondary stipend program also cannot be ruled out as a factor in explaining females’ decision to continue schooling even after negative income shocks. Recall that school attendance is a necessary condition for receiving the stipend, and it is not surprising to see that parents in the cycloneaffected districts allowed their girls to continue schooling even after the devastating effects of the cyclone. Findings in Table 3 indicate that the income shock may produce mixed effects when it comes to intra-household resource allocation, particularly to investment in education conditional on the gender and major livelihood source of a household. This shows that, under a negative income shock caused by a tropical cyclone, farm households in Bangladesh increased health expenditures for both male and female family members, but reduced their expenditures on education of males compared with females. One might point out the exogeneity of the effects of a natural disaster such as a cyclone on intra-household resource allocation. To examine this, in the treatment group, with a focus on intra-household time allocation to earnings from the nonfarm sector and resource allocation to food, education, and health by gender, we conducted a balancing test in Table 4. We mainly compared the t-statistics for the null of the same mean values for both farm and non-farm household groups in Table 4. For example, yearly total income of the farm households in cyclone-affected districts was BDT 8,420 lower in 2005 compared with farm households in the unaffected districts. But, by 2010, it had decreased by BDT 31,000 and the mean difference was highly statistically significant. Similarly, yearly total income of the non-farm households in cycloneaffected districts was BDT 20,000 lower in 2005 compared with the non-farm households in unaffected districts. But, by 2010, it had decreased by the same amount (BDT 21,000), and the mean difference was highly statistically significant. This indicates the significant negative impacts of the cyclone

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

563

Table 4. Differences between households not affected by cyclone Aila and those affected using means of some variables of interest (per household at the district level) in sampled years Characteristics of the household

Farm household

Year Durable asset value (BDT 000) Crop income (BDT 000) Yearly total income (BDT 000) Persons engaged in non-farm sector Total food expenditures (BDT 000) Annual non-food expenditures (BDT 000) Health expenditures, females (BDT 000) Health expenditures, males (BDT 000) Education expenditures, males (BDT 000) Education expenditures, females (BDT 000) Years of schooling, household head Years of schooling, spouse % of households with a male child in high school or college % of households with a female child in high school or college

Non-farm household

2005

2010

2005

2010

4.12** (2.09) 2.18 (1.04) 8.42 (1.36) 0.08 (1.11) 0.33*** (6.43) 3.90 (1.41) 0.22** (2.17) 0.01 (0.04) 0.20 (0.65) 0.21 (1.10) 0.86** (2.52) 1.46*** (5.14) 4.86 (1.19) 7.35* (1.80)

4.31*** (4.19) 4.40** (2.07) 30.69*** (5.48) 0.07 (1.13) 0.22*** (4.23) 8.07*** (3.26) 0.59* (1.75) 0.07 (0.23) 0.30 (0.80) 0.17 (0.62) 0.85*** (2.52) 0.76*** (2.63) 4.14 (1.07) 1.13 (0.29)

12.91*** (4.89) 0.10 (1.49) 20.37 (1.59) 0.001 (0.12) 0.34*** (7.14) 17.33*** (6.62) 0.33 (1.22) 0.05 (0.18) 0.87*** (3.32) 0.65** (2.54) 0.14 (0.45) 0.56** (2.02) 0.57 (0.58) 0.90 (0.28)

8.23*** (3.70) 0.02 (0.66) 20.51*** (4.35) 0.01 (0.17) 0.22*** (5.51) 15.88*** (5.61) 0.06 (0.25) 0.23* (1.86) 1.54* (1.95 1.03*** (3.27) 0.32 (1.14) 0.36 (1.38) 1.70 (0.58) 0.1 (0.04)

Notes: Difference = mean (not affected)  mean (cyclone affected). H0: Diff = 0, Ha: Diff ! = 0 (WHAT IS THIS??). Numbers in parentheses are t-statistics. respectively.

on the income of both farm and non-farm households in the treatment group. Similarly, the crop income of farm households in cyclone-affected districts in 2005 was higher than that of the comparison group; however, by 2010, it too had reverted so the value of the total crop produced by farm households in cyclone-affected districts was lower than that of farm households in unaffected districts, by BDT 4,400. 4 A similar trend can also be seen in the case of expenditures on health and education, particularly expenditures on male family members’ education. The farm households in the affected districts in 2005 spent more on the education of both male and female family members but, in 2010, the opposite was observed: education expenditures on male family members by farm households in the cyclone-affected districts were lower than in the control group, by BDT 300. Table 4 thus lends support to the exogeneity of cyclone effects on the treatment groups. 3. EMPIRICAL METHOD AND MODEL SPECIFICATION The balancing test in Table 4 shows that households in all districts behaved almost the same way in the case of health and schooling expenditures for both genders. However, after a negative income shock caused by tropical cyclone Aila, households in the affected districts spent more on health for both male and female members and less on education of males than of females. To econometrically examine the issue, we applied the difference-in-difference (DID) estimation approach to examine household expenditure behavior on health and education of the male and female family members in the face of negative income shocks caused by Aila. Let us denote the cyclone-affected district as D2 and the unaffected district as D1. We used two sets of HIES data, 2005 and 2010. As Aila hit on May 25–27, 2009, HIES 2005 data are the “before the shock” information and the HIES 2010 data are the “after the shock” information. With this information, DID estimators can be specified as: YDT = (YD2(10)  YD1(05))  (YD2(05)  YD1(05)). where YDT is the difference-in-difference estimators that provide any changes in the expenditure behavior (YDT) on health and education conditional on gender (male and female),

*** **

,

, and * indicate 1%, 5%, and 10% levels of significance,

caused by Aila. Econometrically, to disaggregate the effects of the cyclone from the other effects on loss in farm income and expenditures on health and education, conditional on gender, we estimated the following equation: ;nf Y fDiT ¼ b0 þ b1 ðYear2010dummyÞ þ b2 D2 þ b3 ðD2  Year2010dummyÞ þ ðZ i Þhi þ nit

where YDT is a vector of dependent variables that include crop income, food expenditures, number of persons engaged in the non-farm sector, and expenditures on health and education for male and female family members by type of household (farm and non-farm) separately; D2 is the dummy variable for treatment districts that assumes a value of 1 if the district is affected by the cyclone and 0 otherwise; Z is a vector of variables that include annual total household non-food expenditure in real BDT, food price index, 5 total cropland (in acres), value of household durable assets, real daily wage rate at the subdistrict level, age of the household head, and years of schooling of the household head and spouse; sex dummy variable for household head that assumes a value of 1 if the household head is a female and 0 otherwise; the number of male and female family members in a household separately; the number of children in high school, and college or above; a tuition fee waiver dummy variable, which assumes a value of 1 if a student in high school or above receives a full tuition fee waiver and 0 otherwise; b0 is a scalar parameter; bi and hi are the parameters to be estimated; t indicates year (t = 2005, 2010); superscript f indicates a farm household and nf indicates a non-farm household; and n is the random error term with white-noise property. Note that, in the equation specification, the coefficient b3 = ((YD2(10)  YD1(10))  (YD2(05)  YD1(05))) is the DID estimator. To estimate the functions explaining crop income, persons engaged in the non-farm sector, and food and non-food expenditures, we applied the random effects Tobit estimation process, using subdistrict as our unit of panel and allowing intra-group correlation of the standard error at the household level. We applied a random effects Tobit model because a number of households did not report expenditures, for example, on health and education. 6 Importantly, we estimated functions separately, not only for farm and non-farm

564

WORLD DEVELOPMENT

households, but also for disaggregated households with at least one boy or one girl in high school or above. This is mainly to test our hypothesis that the cyclone increased the demand for male labor in post-disaster mitigation and recovery construction and other productive activities relative to the demand for female labor. And, this decreases education expenditures by households with at least one boy in high school or above compared with girls in the cyclone-affected areas. 4. RESULTS Table 5 presents the estimated functions that explain crop income, total food expenditures, and the number of persons engaged in the non-farm sector, by farm and non-farm households, and whether a household has a boy or girl in high school or above. The year dummy for year 2010 in all cases, which shows overall trends in 2010 compared with 2005 (the base year), was positive and significant in the estimated function explaining crop income (Table 5), indicating an increase in overall crop income of households in the coastal districts of Bangladesh over time. Interestingly, on average, crop income of households in the cyclone-affected districts was higher, by more than BDT 5,000, than for households in the unaffected districts. However, the interaction term, the dummy variable indicating cyclone-affected district multiplied by year 2010, was negative and statistically significant in the estimated function. Results reveal that cyclone Aila significantly reduced the crop income of farm households in the cyclone-affected coastal districts compared with that of farm households in the unaffected districts. Table 5 shows that, on average, a household in a cyclone-affected district lost more than BDT 6,000—more than 10% of the total income of a farm household in 2010. This is likely due to the timing of cyclone Aila, which caused a loss of Boro rice, the major rice crop in Bangladesh, and loss of Aman seedlings. The estimated function explaining crop income of farm households in the coastal districts of Bangladesh in Table 5 further shows that the total crop income of farm households is positive and significantly determined by the size of cropland and the number of male and female family members. Farming in Bangladesh is highly labor-intensive and, thus, a farm household with more family members can have a greater access to a free, high-quality, and timely supply of labor than a farm household with fewer family members. The significant and positive coefficients of the number of male and female family members reflect further support for our findings. Table 5 presents the functions explaining expenditures on food and the number of family members participating in the non-farm sector for income-generating activities, separately for farm and non-farm households and for households with male and female heads of households. The coefficient of the year 2005 dummy variable indicates that, overall, the sampled farm and non-farm households spent less on food in 2010 than in 2005. Interestingly, the estimated functions in Table 5 show that farm households in cyclone-affected districts, with at least one boy or one girl in high school or above, spent relatively more on food in 2010 compared with farm households in the unaffected districts. Recall that Aila significantly reduced the production of rice, a staple food for all Bangladeshis, of farm households in the cyclone-affected districts and therefore these households were compelled to spend more on food in 2010 compared with farm households in the unaffected districts. Importantly, the positive and significant multiplicative dummy variable (cyclone-affected district dummy multiplied

by year 2010 dummy variable) supports the hypothesis that farm households located in the cyclone-affected districts and with at least one boy in high school or above were likely to have more family members in non-farm income-generating activities in 2010 compared with households in the unaffected districts. This is because households were willing to exploit the benefits of increased demand for male labor in the post-disaster mitigation and reconstruction labor market. Severe crop failure followed by increased food expenditures along with increased demand for male labor in the post-disaster mitigation and reconstruction labor market gave farm households with more male labor force, particularly for those with at least one boy in high school or above, the impetus to engage in nonfarm income-generating activities. The positive and significant coefficient of the number of total male family members in the estimated function explaining participation in the non-farm sector by farm households with at least a boy in high school or above also supports this finding. Therefore, farm households located in cyclone-affected districts with at least one boy in high school or above were likely to pull out their boys from school to supply labor to meet the increased labor demand from non-farm activities. This pullout thus reduced farm households’ education expenditures (Table 6). In the case of education expenditures, Table 6 shows that, in general, farm households in cyclone-affected districts spent significantly less on education for both male and female family members, as indicated by the coefficient of the cyclone-affected district dummy compared with farm households in the unaffected districts. The coefficient of the cyclone-affected district dummy multiplied by year 2010 dummy variable, which is our coefficient of interest, was negative and statistically significant in the estimated function explaining the education expenditures for male family members of a farm household with at least one boy in high school or above; however, this coefficient was negative and insignificant in the case of education expenditures for female members of farm households in the cycloneaffected districts with at least one girl in high school or above. On the other hand, a farm household with at least one boy in high school or above in a cyclone-affected district reduced its education expenditures on the male family members by BDT 1,610 compared with farm households in unaffected districts. Interestingly, Table 6 also shows that farm and nonfarm households made a similar, though lesser (BDT 850), average reduction in education expenditures on female family members of farm households with at least one girl in high school or above in cyclone-affected districts compared with similar households located in unaffected districts, but this was not statistically significant. Finally, the findings in Table 6 support the presence of a gender-differentiated outcome of negative income shocks in Bangladesh, in which farm households spent less on the education of their male family members, while keeping female education expenditures relatively unchanged. Section 2(a) already mentioned that cyclonic disasters usually increase the demand for male labor in the male-oriented construction and repair sectors, which increases the opportunity cost of schooling for boys. Households pull boys from schools and colleges to engage them in construction and repair work within the household or let them be paid workers to earn extra income to compensate for the loss in crop income caused by disasters. The withdrawal of boys from school can also save on transportation cost and the cost of lunch at school, resulting in a lower schooling expenditure for male family members. Traditionally, households in Bangladesh do not let girls work outside the house, particularly on farms. As a result, even after a significant reduction in crop income because of tropical

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

565

Table 5. Estimated functions applying random effect Tobit estimation approach, explaining the impacts of cyclone Aila on crop income food expenditures and engagement in non-farm income-generating activities based on gender, by household Dependent variables Household characteristics

Crop income (BDT 000) of farm household

No. of persons engaged in non-farm sector

Farm household

Farm household

With male college student

With female college student

Non-farm household With male college student

With female college student

With male college student

With female college student

Non-farm household With male college student

With female college student

Constant

0.02 (0.00)

0.45*** 0.46*** 0.45*** 0.46*** 0.20 0.14 0.21** 0.23*** (6.42) (6.24) (8.38) (8.54) (1.09) (0.79) (2.35) (2.58) 0.05 0.04 0.04 0.03 0.07 0.15 0.10 0.10 (0.68) (0.49) (0.79) (0.61) (0.40) (0.89) (1.07) (1.08) 0.16** 0.15** 0.09 0.10* 0.35* 0.29 0.03 0.04 (2.14) (1.99) (1.56) (1.79) (1.68) (1.43) (0.23) (0.35) 0.003*** 0.003*** 0.003*** 0.003*** 0.003* 0.004** 0.002*** 0.002** (4.05) (4.24) (6.16) (6.53) (1.72) (2.03) (2.68) (2.15) 1.04*** 1.08*** 0.95*** 0.93*** (13.60) (13.41) (18.38) (18.21) 0.01*** 0.02*** 0.04*** 0.04*** (3.84) (4.01) (3.93) (4.16) 0.0002 0.0001 0.001* 0.002*** 0.0002 0.001 0.001 0.0002 (0.32) (0.09) (1.65) (3.64) (0.10) (0.45) (0.84) (0.22) 0.01*** 0.01*** 0.01*** 0.01** 0.004 0.004 0.01** 0.01** (4.08) (3.82) (2.61) (2.37) (0.60) (0.67) (2.07) (2.04) 0.04** 0.02** 0.03*** 0.01 0.01 0.001 0.002 0.04*** (1.04) (1.29) (0.18) (0.36) (2.82) (2.39) (2.22) (2.93) 0.001 0.0001 0.01 0.002 0.01 0.02 0.01 0.01 (0.15) (0.01) (1.11) (0.30) (0.57) (0.87) (1.49) (1.28) 0.002 0.001 0.004** 0.003* 0.003 0.002 0.01** 0.01** (1.21) (0.20) (2.52) (1.86) (0.66) (0.45) (2.46) (2.33) 0.06 0.06 0.06 0.09* 0.21 0.11 0.33*** 0.43*** (0.74) (0.71) (1.10) (1.76) (0.87) (0.47) (3.28) (4.60) 0.19*** 0.19*** 0.15*** 0.15*** 0.15*** 0.14*** 0.13*** 0.14*** (15.53) (14.78) (13.63) (13.13) (4.12) (3.99) (6.00) (6.81) 0.18*** 0.18*** 0.16*** 0.16*** 0.28*** 0.33*** 0.25*** 0.26*** (13.86) (13.03) (13.55) (13.69) (7.31) (8.69) (11.07) (11.84) 0.01 0.02 0.19** 0.21*** 0.17*** 0.17*** 0.05** 0.05* (2.04) (1.94) (0.40) (0.83) (2.44) (2.78) (3.67) (3.74) 0.04 0.05 0.06** 0.06** 0.10 0.11 0.15*** 0.15*** (1.20) (1.45) (2.43) (2.50) (1.10) (1.28) (3.10) (3.18) 0.01 0.02 0.03 0.08 (0.09) (0.18) (0.36) (0.94) 2.95*** 3.01*** 2.93*** 2.82*** 1.28** 1.48*** 0.51 0.54* (8.83) (8.65) (12.63) (12.34) (2.19) (2.60) (1.64) (1.77)

Sigma_u

5.10*** (5.57) 22.10*** (50.01)

0.22*** (8.85) 0.429*** (35.64)

0.24*** (8.90) 0.442*** (35.62)

0.13*** (6.54) 0.414*** (41.47)

0.12*** (6.27) 0.415*** (41.86)

0.34*** (4.62) 1.186*** (24.58)

0.35*** (4.86) 1.155*** (24.65)

0.21*** (5.24) 0.769*** (37.79)

0.23*** (5.62) 0.754*** (38.01)

1345 2 1 459.97 0.00 17.01 0.00

725 0 1 909.26 0.00 67.26 0.00

728 0 1 833.48 0.00 69.68 0.00

944 0 4 1423.84 0.00 31.39 0.00

957 0 4 1474.55 0.00 27.65 0.00

725 320 2 117.75 0.00 9.98 0.00

728 320 2 138.63 0.00 11.54 0.00

944 97 17 241.43 0.00 15.01 0.00

957 98 17 274.37 0.00 18.28 0.00

Year 2010 dummy (base 2005) Affected district (yes = 1) Affected district  year 2010 dummy Annual non-food expenditures (BDT 000)

5.90** (2.57) 0.84 (0.37) 5.84** (2.20) 0.07*** (2.97)

Total expenditures on food (BDT 000)

Food price index Cropland (acres) Durable asset value (BDT 000) Real wage rate at the subdistrict level (BDT) Years of schooling, household head Years of schooling, spouse Age, household head Female household head dummy (yes = 1) Total female members in the household Total male members in the household No. of children in high school No. of children in college or above

2.92*** (18.74) 0.001 (0.01) 0.10 (1.17) 0.20 (1.00) 0.09 (0.39) 0.06 (1.08) 3.88 (1.28) 1.27*** (2.66) 0.88* (1.78) 0.56 (0.68) 0.80 (0.62)

Tuition fee waiver dummy (yes = 1)

Sigma_e N Left censored Right censored Wald chi2 p-Value Likelihood ratio test of sigma_u = 0 p-Value

Note: Values computed in terms of real BDT using general price index 1995–96 = 100. Numbers in parentheses are z statistics based on standard errors that allow for intra-group correlation. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

566

WORLD DEVELOPMENT

Table 6. Estimated functions applying random effect Tobit estimation approach, explaining the impacts of cyclone Aila on annual education and health expenditures (BDT 000) based on gender, by household Dependent variables

Annual education expenditures (BDT 000)

Household characteristics

Farm household

Sex Year 2010 dummy Affected district (yes = 1) Affected district  year 2010 dummy Annual non-food expenditures (BDT 000) Cropland (acres) Durable asset value (BDT 000) Real wage rate at the subdistrict level (BDT) Household head’s years of schooling Spouse’s years of schooling Age of household head Female household head dummy (yes = 1) Total female members in the household Total male members in the household No. of children in high school No. of children in college or above Tuition fee waiver dummy (yes = 1) Constant Sigma_u Sigma_e N Left censored Right censored Wald chi2 p-Value Likelihood ratio test of sigma_u = 0 p-Value

Non-farm household

Annual health-related expenditures (BDT 000) Farm household

Non-farm household

Malea

Femaleb

Malea

Femaleb

Malea

Femaleb

Malea

Femaleb

0.85 (1.05) 2.10*** (2.79) 1.61* (1.70) 0.04*** (4.75) 0.08* (1.75) 0.03 (0.25) 0.05 (1.59) 0.05 (0.68) 0.07 (0.75) 0.001 (0.05) 0.17 (0.15) 1.17*** (6.75) 1.14*** (6.35) 0.94*** (2.64) 3.82*** (9.54) 0.98 (1.43) 7.37*** (2.86)

0.64 (0.99) 1.69*** (2.85) 0.85 (1.15) 0.04*** (5.05) 0.001 (0.02) 0.01 (0.74) 0.05** (2.38) 0.02 (0.27) 0.08 (1.10) 0.02 (1.30) 0.49 (0.56) 1.00*** (7.66) 0.76*** (5.42) 0.38 (1.38) 1.76*** (5.64) 1.59*** (3.30) 8.96*** (4.56)

1.86** (2.37) 0.04 (0.05) 0.63 (0.67) 0.06*** (8.66)

0.93 (1.22) 0.34 (0.48) 0.90 (0.96) 0.06*** (9.64)

0.13 (0.28) 0.63 (1.45) 0.26 (0.45) 0.03*** (7.55)

0.03*** (3.93) 0.04 (1.34) 0.04 (0.53) 0.23*** (2.76) 0.03 (1.36) 3.03*** (3.50) 1.77*** (9.46) 1.95*** (9.53) 2.50*** (6.31) 2.48*** (6.03) 2.01*** (2.88) 12.96*** (4.95)

1.48*** (2.90) 0.09 (0.12) 0.44 (0.74) 0.04*** (6.37) 0.004 (0.14) 0.002 (0.32) 0.03 (1.46) 0.05 (1.21) 0.08 (1.62) 0.01 (1.02) 0.29 (0.45) 0.07 (0.69) 0.14 (1.30) 0.02 (0.09) 0.44* (1.84) 0.11 (0.28) 1.80 (0.95)

0.64** (2.00) 0.60** (2.03) 0.64* (1.66) 0.03*** (9.92)

0.01 (1.22) 0.01 (0.45) 0.01 (0.10) 0.21** (2.48) 0.01 (0.24) 0.84 (0.93) 1.64*** (8.58) 0.92*** (4.54) 1.49*** (3.76) 3.43*** (8.24) 0.24 (0.30) 4.76* (1.81)

0.14 (0.31) 0.56 (1.37) 0.19 (0.36) 0.04*** (7.62) 0.02 (0.57) 0.002 (0.34) 0.001 (0.05) 0.02 (0.42) 0.05 (1.00) 0.01 (0.68) 0.27 (0.43) 0.16* (1.65) 0.09 (0.94) 0.12 (0.61) 0.02 (0.09) 0.21 (0.56) 1.02 (0.71)

0.01** (2.47) 0.01 (0.66) 0.02 (0.57) 0.04 (1.21) 0.01 (0.77) 0.17 (0.46) 0.03 (0.34) 0.20** (2.47) 0.36** (2.19) 0.23 (1.35) 0.45 (1.42) 0.54 (0.50)

0.01** (2.16) 0.02 (0.91) 0.02 (0.49) 0.03 (0.66) 0.02 (1.09) 0.29 (0.55) 0.30*** (2.60) 0.01 (0.06) 0.42* (1.66) 0.56** (2.14) 0.09 (0.18) 3.13** (1.97)

0.92* (1.90) 5.64*** (29.44)

0.77** (2.12) 4.32*** (27.75)

0.91** (2.11) 6.61*** (34.56)

0.64 (1.38) 6.40*** (32.79)

0.32 (0.85) 3.30*** (33.54)

3.03*** (6.77) 3.26*** (29.68)

0.29* (1.65) 2.79*** (39.04)

0.00 (0.00) 4.31*** (40.65)

725 216 1 315.39 0.00 1.13 0.14

728 251 1 213.59 0.00 1.48 0.11

944 264 1 443.43 0.00 1.50 0.11

957 348 1 434.34 0.00 0.59 0.22

725 87 1 96.80 0.00 0.20 0.32

728 67 1 88.56 0.00 12.94 0.00

944 139 1 159.27 0.00 0.85 0.17

957 116 1 103.72 0.00 0.0 1.00

Note: Values computed in terms of real BDT using general price index 1995–96 = 100. Numbers in parentheses are z statistics based on standard errors that allow for intra-group correlation. a Expenditures on male family members by households with at least one male child studying in college. b Expenditures on female family members by households with at least one female child studying in college. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

cyclone Aila, farm households with at least one girl in college or above in a cyclone-affected district did not reduce education expenditures for their female family members. This finding is

similar to that of Jensen (2000): in general, boys’ school enrollment fell more than girls’ in response to a negative income shock caused by drought in Coˆte d’Ivoire. The findings of this

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

567

Table 7. Estimated functions applying two-limit Tobit estimation approach, explaining the impacts of cyclone Aila on annual education and health expenditures (BDT 000) based on gender, by household Dependent variables Household characteristics Sex Year 2010 dummy Affected district (yes = 1) Affected district  year 2010 dummy Annual non-food expenditures (BDT 000) Cropland (acres) Durable asset value (BDT 000) Real wage rate at the subdistrict level (BDT) Household head’s years of schooling Spouse’s years of schooling Age of household head Female household head dummy (yes = 1) Total female members in the household Total male members in the household No. of children in high school No. of children in college or above Tuition fee waiver dummy (yes = 1) Constant Sigma N Left censored Right censored F-Value p-Value

Annual education expenditures (BDT 000) Farm household

Non-farm household

Annual health-related expenditures (BDT 000) Farm household

Non-farm household

Malea

Femaleb

Malea

Femaleb

Malea

Femaleb

Malea

Femaleb

0.86 (1.08) 2.10*** (2.86) 1.58* (1.71) 0.04*** (4.31) 0.08 (1.35) 0.01 (0.45) 0.05** (2.53) 0.05 (0.84) 0.06 (0.69) 0.0001 (0.00) 0.01 (0.01) 1.18*** (6.97) 1.15*** (7.36) 0.90** (2.54) 3.76*** (8.42) 1.01* (1.93) 7.52*** (3.83)

0.67 (1.14) 1.69*** (3.56) 0.902 (1.33) 0.04*** (3.80) 0.002 (0.06) 0.01 (0.50) 0.05*** (3.19) 0.01 (0.26) 0.06 (0.93) 0.02 (1.04) 0.44 (0.74) 1.01*** (7.46) 0.77*** (6.00) 0.36 (1.19) 1.74*** (5.22) 1.54*** (4.66) 8.80*** (5.05)

1.68 (1.38) 1.76 (1.56) 2.107 (1.41) 0.19*** (9.12)

0.90 (1.05) 0.26 (0.38) 0.919 (0.94) 0.06*** (4.16)

0.262 (0.62) 0.581 (0.96) 0.142 (0.19) 0.03*** (4.58)

0.03*** (2.78) 0.04* (1.78) 0.04 (0.58) 0.23*** (2.95) 0.03 (1.57) 3.06*** (3.63) 1.79*** (7.71) 1.99*** (6.87) 2.54*** (5.35) 2.53*** (4.94) 2.01*** (2.91) 12.83*** (5.91)

1.09** (2.01) 0.01 (0.02) 0.250 (0.45) 0.04*** (3.94) 0.01 (0.48) 0.004 (0.44) 0.01 (0.55) 0.07 (1.39) 0.13** (2.25) 0.01 (1.26) 0.15 (0.24) 0.09 (0.86) 0.13 (1.58) 0.10 (0.72) 0.58*** (3.24) 0.31 (1.07) 0.65 (0.68)

0.73** (2.15) 1.15*** (3.65) 1.149*** (3.00) 0.04*** (7.51)

0.09*** (4.59) 0.04 (1.19) 0.23** (2.14) 0.25** (2.02) 0.06 (1.46) 2.61* (1.76) 2.82*** (8.65) 0.98*** (2.60) 1.60** (2.45) 3.90*** (5.35) 1.57 (1.29) 8.90** (2.40)

0.13 (0.32) 0.61* (1.80) 0.161 (0.31) 0.04*** (4.39) 0.02 (1.16) 0.002 (0.22) 0.001 (0.10) 0.02 (0.35) 0.05 (0.72) 0.01 (0.91) 0.25 (0.42) 0.15** (2.20) 0.09 (0.85) 0.12 (0.68) 0.01 (0.06) 0.20 (0.47) 1.22 (1.21)

0.01*** (2.77) 0.01 (0.70) 0.07 (1.56) 0.06 (1.46) 0.01 (1.22) 0.36 (0.79) 0.10 (1.24) 0.32*** (2.92) 0.56*** (3.12) 0.30 (1.50) 0.61* (1.73) 0.92 (0.91)

0.01* (1.91) 0.02 (0.83) 0.03 (0.80) 0.03 (0.97) 0.02 (1.25) 0.29 (0.79) 0.31*** (3.15) 0.001 (0.01) 0.49** (2.36) 0.62*** (3.03) 0.08 (0.15) 3.18* (1.96)

5.72*** (51.35)

4.40*** (57.85)

15.01*** (37.37)

6.55*** (12.29)

3.32*** (53.98)

3.75*** (33.61)

4.92*** (172.09)

4.66*** (45.15)

725 216 1 26.98 0.00

728 251 1 15.72 0.00

944 264 1 28.17 0.00

957 348 1 10.21 0.00

725 87 1 6.28 0.00

728 67 1 9.55 0.00

944 139 1 8.94 0.00

957 116 1 7.70 0.00

Note: Values computed in terms of real BDT using general price index 1995–96 = 100. Numbers in parentheses are t statistics based on standard errors that allow for intra-group correlation. a Expenditures on male family members by households with at least one male child studying in college. b Expenditures on female family members by households with at least one female child studying in college. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

paper are also similar to the results obtained by Maluccio (2005) and Lo´pez Bo´o (2010), which demonstrate that, because of the reduction in opportunity cost caused by a reduction in coffee price in Nicaragua and the financial crisis in Argentina, boys’ school participation in rural areas actually increased. In the present case, however, the opposite case has been shown: male education expenditures decreased because of the increased opportunity cost of male labor in the construction and repair sectors in the aftermath of a cyclonic

disaster that substantially reduced the crop income of farm households. The estimated functions explaining health expenditures by farm households indicate that the negative income shock caused by cyclone Ala did not generate any gender-differentiated impacts on health expenditures in farm households in Bangladesh. However, in the case of non-farm household expenditures on health, the results in Table 6 indicate a reduction in health expenditures for male family members

568

WORLD DEVELOPMENT

of non-farm households, but this remained statistically unchanged for the female family members of non-farm households. These findings indicate that a negative income shock can generate a gender-differentiated impact on health expenditures, based on the source of livelihood of a household (farm versus non-farm). Tables 5 and 6 further reveal other interesting findings. Households that spent more on non-food items also tended to spend more for family members’ food, health, and education and, most importantly, to allocate more members to non-farm income-earning activities. Strikingly, the coefficient of the tuition fee waiver dummy variable was negative and insignificant in explaining education expenditures for males, but positive and significant for female members of farm households. This is difficult to explain, but it may be the case that a tuition fee waiver for a male student merely reduces the educational cost burden of a household, while a tuition fee waiver for a female student actually generates further incentives for parents to spend more on their daughters’ education since, historically, women have lagged behind men in educational achievement in Bangladesh. The last two rows in Tables 5 and 6 present the likelihood ratio test results of the relevance of the application of the random-effects Tobit model compared with the ordinary two-limit Tobit estimation approach. The likelihood ratio test statistics and corresponding p-values in Table 5 clearly support the application of the random-effects Tobit model for estimating functions explaining crop income, food expenditures, and the number of family members’ engagement in non-farm income-generating activities. The likelihood ratio test statistics and corresponding p-values in Table 6, however, suggest applying a simple two-limit Tobit model for estimating the functions explaining health and education expenditures of farm and non-farm households. Table 7 presents the estimated functions explaining health and education expenditures of farm and non-farm households using a simple two-limit Tobit estimation approach. The magnitude and the sign of most of the major coefficients remained unchanged. This indicates the robustness of our major findings, which strengthened the conclusion that a negative income shock can generate gender-differentiated impacts on households’ expenditure behavior, particularly for education expenditures. 5. CONCLUSIONS AND POLICY IMPLICATIONS Understanding intra-household resource allocation behavior conditional on gender in the face of negative income shocks is important to ensure not only gender parity (as per the Millennium Development Goals of the United Nations) but also the long-run economic growth of a country. Further, many important decisions regarding allocation of resources and income-generating activities that significantly affect development and growth are managed by households. Farm households across the world often face negative income shocks because of natural disasters and development economists and policymakers have grown increasingly concerned that negative income shocks may impact intra-household resource allocation to health and education according to gender. Health and education are a principal concern as they are the major ingredients of human capital development of a country. As only a few empirical studies address this issue and, ironically, the findings tend to contradict each other, an attempt has been made to examine farm households’ choices relating

to expenditures on health and education conditional on gender in the face of negative income shocks. In this article, farm and non-farm households’ expenditure behavior on food, education, and health and allocation of family members to non-farm income-generating activities, conditional on gender, in the presence of negative income shocks are examined, using Bangladesh as a case study. Using the HIES data collected by the BBS in 2005 and 2010 and analyzing the effect of a natural disaster, cyclone Aila, that hit the coastal area of Bangladesh on May 27–27, 2009, as a case, this paper examines households’ loss in crop income and farm and non-farm households’ expenditure behavior on health and education, conditional on gender, under negative crop income shocks. The findings show that the loss in crop income of farm households in the cyclone-affected districts was substantial compared with that of households in the unaffected districts. However, under negative income shocks, farm households in the cyclone-affected districts did not differentiate between male and female family members with respect to their health expenditures. It is found that, in the cyclone-affected districts, the health expenditures of nonfarm households actually decreased in 2010 for male members but remained unchanged for female members. The gender-differentiated outcome, however, emerged in the case of education expenditures. In particular, farm households in cyclone-affected districts spent less on the education of male family members than on female family members compared with those in unaffected districts. We conjecture that disasters usually raise the demand for male labor in construction and repair work, which creates a greater opportunity cost of education for boys, and increased wages for male workers. A drastic reduction in education expenditures for the male members of a household due to negative income shocks, however, can affect human capital formation in the long run. To ensure human capital development in agriculture-dependent developing countries in the long run (where negative income shocks are frequent because of crop failure caused by natural disasters), disaster loans, temporary subsidies for basic needs, and conditional cash transfers to households with school-age boys and girls can be extended and strengthened. These services can be particularly designed to provide for farm households in coastal areas, where both the frequency and severity of natural disasters are high and where catastrophic events are projected to occur more in the future because of global climate change. In addition, to minimize losses in crop income brought about by natural disasters, international agricultural research institutes, in collaboration with national partners, could develop and disseminate more climate-change-resilient, short-duration, and high-yielding crop varieties. Finally, alternative livelihood strategies other than farming can be introduced in the coastal areas. Extension of general education, training in modern agriculture, and marketing linkages can contribute to the expansion of income diversification strategies of farm households and the development of off-farm incomeearning opportunities. Thus, policymakers and donors are strongly encouraged to fund programs that deal with the extension of general education, training on modern agriculture, and development of and dissemination of climate-smart agricultural techniques to ensure stable income of farm households. This will reduce gender-differentiated impacts on education expenditures brought about by negative income shocks.

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

569

NOTES 1. In Bangladesh a fiscal year starts in July and ends in June. Thus, HIES 2005 data were actually collected during 2004–05 and HIES 2010 data were collected during 2009–10. Therefore the cyclone had direct impact on income of the households. 2. We strictly consider only households that cultivate any crop in a parcel of their farmland as a farm household. 3. The decision to be a farm/non-farm household may not be strictly exogenous after income shock. In particular, after the shock, many households might choose to become non-farm households (at least for a short period of time). The problem though is that the data are not longitudinal in nature so we do not know whether the same household moved from being a farm household to being a non-farm household. So it is difficult to directly address this concern. Nonetheless, we divided the sampled households into farm and non farm households to examine the differential impacts of negative income shocks on different households. 4. US$1 = BDT 80 in 2010.

5. We calculated the geometric price index P suggested by Stone (1954), using the following formula: price index = iwi ln Pi where wi is the share of expenditure on commodity i and Pi is the price of the ith commodity. 6. In the case of censored data an application of a tobit model is suggested for estimation purposes (Gujarati, 1995), as the use of an OLS model may provide biased and inefficient estimates. The bias and inefficiency of the OLS estimates, however, depend on the ratio of zero observations in a dependent variable related to total observations in a data set. The greater the ratio, the greater the bias and inefficiency of the OLS estimates and vice versa (e.g., Willson & Tisdell, 2002). In cases where the ratio of zero observations in a dependent variable is significantly low, the difference between OLS and tobit estimates also tends to be low. Considering this, we also report OLS estimation results in the Appendix. The OLS estimation results show that the effect of cyclone on education and health expenditures does not generate any significant difference between affected and unaffected households irrespective of whether a household has a male or female kid in high school or above. This might be because of the biasness in OLS estimation due to the presence of a number of zeros in the dependent variables.

REFERENCES BBS (Bangladesh Bureau of Statistics). (1999). 1998 Statistical yearbook of Bangladesh. Dhaka: Statistics Division, Ministry of Planning. BBS (Bangladesh Bureau of Statistics). (2011). Statistical yearbook of Bangladesh-2010. Dhaka: Statistics Division, Ministry of Planning. Cameron, L. A., & Worswick, C. (2001). Education expenditure responses to crop loss in Indonesia: A gender bias. Economic Development and Cultural Change, 49(2), 351–363. del Ninno, C., Dorosh, P. A., & Smith, L. C. (2003). Public policy, food markets, and household-coping strategies in Bangladesh: Lessons from the 1998 floods. Washington, DC.: Food Consumption and Nutrition Division, International Food Policy Research Institute. FCND Discussion Paper No. 156. Online: Accessed on June 25, 2012. GOB. (2014). Bangladesh economic review 2014. Dhaka: Ministry of Finance. Online: http://mof.gov.bd/en/budget/14_15/ber/bn/ Index_&_Indicators_bn_2014.pdf Accessed on September 03, 2014. Gujarati, D. N. (1995). Basic econometrics (International 3rd ed.). New York: McGraw-Hill Inc. IFRCRC (International Federation of Red Cross and Red Crescent Societies). 2010. Operation update, Bangladesh: Cyclone “Aila”. Online: Accessed August 30, 2013. IRRI (International Rice Research Institute). (2010). Scuba rice: Breeding flood tolerance into Asia’s local mega rice varieties. Online: Accessed on November 20, 2011. ITHACA (Information Technology for Humanitarian Assistance Corporation and Action). (2009). Bangladesh-tropical cyclone Aila. Online: Accessed on June 10, 2012. Jensen, R. (2000). Agricultural volatility and investments in children. American Economic Review, 90(2), 399–404. Khandker, S. R. (2007). Coping with flood: role of institutions in Bangladesh. Journal of Agricultural Economics, 36, 169–180. Khandker, S., Pitt, M., & Fuwa, N. (2003). Subsidy to promote girls’ secondary education: The female stipend program in Bangladesh. Munich Personal RePEc Archive. Online: Accessed August 30, 2013. Lo´pez Bo´o, F. (2010). In school or at work? Evidence from a crisis. IZA Discussion Paper Series 4692. Bonn: Institute for Labor Studies. Online: Accessed August 30, 2013.

Maluccio, J. A. (2005). Coping with the ‘coffee crisis’ in Central America: The role of the Nicaraguan Red de Proteccio´n Social. FCND Discussion Paper 188. Washington, DC: International Food Policy Research Institute, Food Consumption and Nutrition Division. Online: Accessed August 30, 2013. Paul, B. K. (1998). Coping mechanisms practiced by drought victims (1994/5) in North Bengal, Bangladesh. Journal of Applied Geography, 18(4), 355–373. Paul, B. K., & Rashid, H. (1993). Flood damages to rice crop in Bangladesh. Journal of Geographical Review, 83(2), 150–159. Raynor, J., & Wesson, K. (2006). The girls’ stipend program in Bangladesh. Journal of Education for International Development, 2(2), Online: Accessed August 30, 2013. Rose, P., & Al-Samarrai, S. (2001). Household constraints on schooling by gender: Empirical evidence from Ethiopia. Comparative Education Review, 45(1), 36–63. Sarker, M. R., Alam, A. K. ., & Gow, J. (2012). Exploring the relationship between climate change and rice yield in Bangladesh: An analysis of time-series data. Agricultural Systems, 112, 11–16. Schiermeier, Q. (2014). Floods: Holding back the tide: With the Ganges– Brahmaputra delta sinking, the race is on to protect millions of people from future flooding. Nature, Online: Accessed April 20, 2014. Stone, R. (1954). Linear expenditure system and demand analysis: An application to the pattern of British demand. The Economic Journal, 64(255), 511–527. Tansel, A. (2002). Determinants of school attainment of boys and girls in Turkey: Individual, household, and community factors. Economics of Education Review, 21(5), 455–470. Willson, C., & Tisdell, C. (2002). OLS and Tobit estimates: When is substitution defensible operationally? Working Paper No. 15. School of Economics, The University of Queensland, Brisbane 4072 Australia. Online: Accessed on September 04, 2014. World Bank. (2012). Gender equality and development. Washington, DC: The International Bank for Reconstruction and Development.

570

WORLD DEVELOPMENT

APPENDIX

Table 8. Estimated functions applying OLS, explaining the impacts of cyclone Aila on crop income food expenditures and engagement in non-farm incomegenerating activities based on gender, by household Dependent variables Household characteristics

Year 2010 dummy (base 2005) Affected district (yes = 1) Affected district  year 2010 dummy Annual non-food expenditure (000, BDT)

Crop income (000, BDT) of farm household

5.85** (2.57) 0.62 (0.31) 4.76* (1.81) 0.07*** (3.04)

Food price index Cropland (acres) Durable asset value (000, BDT) Real wage rate at the sub district level (BDT) Years of schooling, household head Years of schooling, spouse Age, household head Female household head dummy (yes = 1) Total female members in the household Total male members in the household No. of children in high school No. of children in college and above

2.99*** (18.88) 0.001 (0.01) 0.14* (1.79) 0.18 (0.90) 0.05 (0.20) 0.05 (0.86) 4.62 (1.48) 1.09** (2.24) 0.85* (1.65) 0.18 (0.21) 1.17 (0.88)

Tuition fee waived dummy (yes = 1) Constant

3.25 (0.48)

N

1345

Total expenditure on food (000 BDT) per household in two weeks

No. of persons engaged into non-farm sector

Farm household

Farm household

With male college student

With female college student

Non-farm household With male college student

With female college student

With male college student

With female college student

Non-farm household With male college student

With female college student

0.40*** 0.41*** 0.46*** 0.46*** 0.16 0.09 0.23*** 0.26*** (5.69) (5.51) (8.70) (8.74) (1.55) (0.89) (2.79) (3.19) 0.17** 0.06 0.06 0.05 0.05 0.03 0.08 0.16** (1.03) (0.99) (1.14) (1.06) (0.35) (0.91) (2.16) (2.42) 0.15** 0.16** 0.10* 0.11* 0.21* 0.14 0.06 0.01 (2.03) (2.01) (1.67) (1.80) (1.80) (1.25) (0.61) (0.06) 0.004*** 0.004*** 0.003*** 0.003*** 0.002 0.002* 0.002** 0.001* (5.36) (4.98) (6.61) (6.91) (1.43) (1.83) (2.34) (1.66) 1.02*** 1.03*** 0.96*** 0.93*** (13.30) (12.88) (18.56) (18.31) 0.02*** 0.02*** 0.03*** 0.03*** (4.31) (4.42) (4.35) (4.50) 0.0001 0.0001 0.001 0.002*** 0.00003 0.001 0.001 0.0001 (0.13) (0.07) (1.48) (3.62) (0.02) (0.53) (0.52) (0.13) 0.01*** 0.01*** 0.10*** 0.01** 0.002 0.002 0.01*** 0.01*** (4.17) (3.65) (2.64) (2.48) (0.61) (0.72) (2.75) (2.84) 0.01 0.01 0.001 0.001 0.02*** 0.02** 0.02** 0.02*** (1.41) (1.56) (0.11) (0.15) (2.64) (2.26) (2.35) (3.03) 0.01 0.003 0.004 0.0001 0.01 0.01 0.01 0.01 (0.65) (0.37) (0.73) (0.02) (0.53) (0.71) (1.48) (1.17) 0.002 0.001 0.003** 0.003* 0.002 0.001 0.01** 0.01** (1.09) (0.41) (2.28) (1.75) (0.75) (0.39) (2.58) (2.47) 0.05 0.05 0.07 0.10* 0.19 0.12 0.32*** 0.40*** (0.55) (0.60) (1.31) (1.81) (1.31) (0.88) (3.38) (4.55) 0.19*** 0.18*** 0.15*** 0.15*** 0.12*** 0.11*** 0.13*** 0.15*** (14.13) (13.26) (13.17) (12.71) (5.45) (5.08) (6.84) (7.69) 0.18*** 0.18*** 0.16*** 0.16*** 0.19*** 0.23*** 0.26*** 0.26*** (13.00) (12.09) (12.90) (12.97) (8.62) (10.14) (12.04) (12.84) 0.01 0.03 0.14*** 0.15*** 0.19*** 0.19*** 0.06** 0.06* (2.20) (1.91) (0.50) (1.03) (3.01) (3.37) (4.39) (4.56) 0.05 0.06* 0.06** 0.06** 0.06 0.08 0.15*** 0.15*** (1.52) (1.67) (2.25) (2.41) (1.20) (1.48) (3.37) (3.54) 0.01 0.03 0.02 0.05 (0.13) (0.31) (0.23) (0.67) 3.00*** 3.02*** 2.95*** 2.81*** 0.22 0.34 0.62** 0.65** (9.47) (9.25) (13.23) (12.90) (0.71) (1.12) (2.31) (2.52) 725

728

944

957

725

728

Note: Values computed in terms of real BDT using general price index 1995–96 = 100. Numbers in parentheses are t statistics. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

944

957

INTRA-HOUSEHOLD RESOURCE ALLOCATION UNDER NEGATIVE INCOME SHOCK: A NATURAL EXPERIMENT

571

Table 9. Estimated functions applying two-limit Tobit estimation approach, explaining the impacts of cyclone Aila on annual education and health expenditures (BDT 000) based on gender, by household Dependent variables

Annual education expenditure (000 BDT)

Household characteristics

Farm household

Sex Year 2010 dummy Affected district (yes = 1) Affected district  year 2010 dummy Annual nonfood expenditure (000, BDT) Cropland (acres) Durable asset value (000, BDT) Real wage rate at the sub district level (BDT) Household head’s years of schooling Spouse’s years of schooling Age of household head Female household head dummy (yes = 1) Total female members in the household Total male members in the household No. of children in high school No. of children in college and above Tuition fee waived dummy (yes = 1) Constant N

Non-farm household

Annual health-related expenditure (000, BDT) Farm household

Non-farm household

Male1

Female2

Male1

Female2

Male1

Female2

Male1

Female2

0.27 (0.44) 0.55 (1.02) 0.40 (0.57) 0.04*** (5.62) 0.08** (2.16) 0.001 (0.11) 0.02 (0.76) 0.05 (0.96) 0.03 (0.41) 0.01 (0.40) 0.27 (0.32) 0.73*** (5.66) 0.54*** (3.96) 0.64** (2.36) 3.27*** (10.50) 0.92* (1.81) 2.23 (1.20)

0.19 (0.42) 0.61 (1.56) 0.40 (0.78) 0.03*** (5.77) 0.02 (0.67) 0.004 (0.63) 0.04** (2.48) 0.01 (0.33) 0.05 (1.13) 0.02 (1.60) 0.33 (0.54) 0.50*** (5.44) 0.49*** (4.92) 0.19 (0.96) 1.40*** (6.16) 0.63* (1.74) 4.21*** (3.17)

0.60 (0.43) 0.96 (0.78) 1.84 (1.10) 0.18*** (14.28)

0.68 (1.24) 0.13 (0.26) 0.60 (0.91) 0.05*** (11.40)

0.52 (1.10) 0.35 (0.82) 0.07 (0.12) 0.03*** (7.74)

0.02*** (4.29) 0.01 (0.38) 0.01 (0.20) 0.19*** (3.25) 0.03 (1.50) 1.53** (2.53) 0.80*** (6.11) 1.08*** (7.74) 1.68*** (5.85) 1.90*** (6.38) 0.88* (1.67) 4.81*** (2.71)

1.03** (2.18) 0.26 (0.64) 0.24 (0.44) 0.04*** (7.27) 0.01 (0.49) 0.003 (0.47) 0.01 (0.49) 0.06 (1.56) 0.11** (2.13) 0.01 (0.83) 0.16 (0.24) 0.08 (0.77) 0.13 (1.20) 0.11 (0.53) 0.61** (2.56) 0.11 (0.28) 0.18 (0.13)

0.13 (0.26) 0.79* (1.76) 0.64 (1.07) 0.04*** (7.79)

0.08*** (5.52) 0.01 (0.20) 0.18 (1.28) 0.16 (1.07) 0.02 (0.52) 1.75 (1.11) 1.45*** (4.38) 0.29 (0.81) 0.59 (0.81) 2.67*** (3.54) 0.95 (0.68) 0.25 (0.05)

0.25 (0.62) 0.42 (1.17) 0.01 (0.02) 0.04*** (7.88) 0.02 (0.91) 0.002 (0.34) 0.01 (0.40) 0.02 (0.56) 0.05 (1.07) 0.01 (0.61) 0.11 (0.19) 0.13 (1.56) 0.05 (0.49) 0.04 (0.23) 0.01 (0.06) 0.27 (0.79) 0.01 (0.00)

0.01** (2.46) 0.01 (0.36) 0.06 (1.13) 0.05 (0.91) 0.001 (0.06) 0.22 (0.39) 0.08 (0.66) 0.28** (2.16) 0.59** (2.28) 0.25 (0.93) 0.54 (1.08) 0.56 (0.34)

0.01*** (2.67) 0.01 (0.67) 0.02 (0.39) 0.04 (0.69) 0.01 (0.95) 0.36 (0.69) 0.32*** (2.81) 0.01 (0.06) 0.57** (2.28) 0.55** (2.13) 0.12 (0.25) 1.89 (1.22)

725

728

944

957

725

728

944

957

Note: Values computed in terms of real BDT using general price index 1995–96 = 100. Numbers in parentheses are t statistics. a Expenditure on male family members by households with at least one male kid studying in college. b Expenditure on female family members by households with at least one female kid studying in college. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

Available online at www.sciencedirect.com

ScienceDirect