Consumption and income seasonality and food deficiency in rural Ethiopia

Nigussie Tefera

EDRI, Young Lives: An International Study on childhood poverty project

e-mail: nigtefera@yahoo.com

First draft

April 2006

Abstract

Rural households in Ethiopian rely on seasonal agriculture for their income. The paper explores the implication of this income seasonality for household consumption. I use household-level data from four Ethiopia villages to document seasonal patterns in income and consumption, and to test whether income seasonality produces seasonal consumption variation. While the findings substantiate seasonal patterns in income and non-food expenditure, the null hypothesis of no seasonal pattern in household consumption can not be rejected i.e., there is no well-built seasonal variation in food consumption. This in turn implies that seasonal food consumption doesn’t tracks seasonal income.

Usually household retain output harvested during harvesting seasons and draw down throughout the course of the year to smooth consumption. However, the paper found out that smoothing consumption doesn’t necessary mean that output retained for consumption during harvesting is sufficient for the entire year. Using net calorie retained for consumption, the paper has shown that households are in transitory food deficient for at least three months.

1. Introduction

Rural household in many agrarian economies, where rain-dependent crop cultivation is the primary source of household income, have income that vary seasonally (Paxson and Chaunduri, 2001). The seasonal variation in income is quite large. For instance three of the four rural villages of Ethiopia examined in this paper receive, on average, more than 50% of household income in the top three months.

What are the consequences of this pronounced income seasonality for consumption? Empirical evidence suggests that although household consumption also varies and leading in many cases to seasonal variation in nutritional status and health (see Chambers et al.,(1981) and Sahn (1989) and Walker and Ryan (1990)), income seasonality need not be the sources of consumption seasonality. For instance, Paxson (1992) using rural household in Thailand found little evidence that consumption trucks income seasonality. She indicated that the observed seasonal variations are the result of seasonal variation in preferences or prices that are common to all households rather than seasonal income. Likewise, using rural households in India Paxson and Chauduri (2001) indicated that while there does appear to be some seasonality in consumption patterns, it is much less pronounced than in the case of income, and more surprisingly, they have shown that the consumption patterns are quite similar for households with very different seasonal income patterns and suggested that borrowing constraints caused by credit market imperfections can indeed combine with income seasonality to produce consumption seasonality. If households cannot borrow against future income during the slack season, then consumption levels may well be lower during the period before the harvest.

Using three years panel data of Ethiopian Rural Household Survey (ERHS) Dercon and Krishanen (2001) has also shown high seasonal variability in consumption due to common climatic and other crop shocks, as well as idiosyncratic shocks to livestock rather than income. The other case studies conducted in Arsi zone, Oromia, regional state of Ethiopia by Degefa (2000) also has also shown seasonal food deficient at least for four months. Nevertheless, these surveys typically construct monthly data from either short recall periods, by using data based very long recall periods, such total yearly consumption from own production, or based on questions related to a ‘usual month’. Such recalls is subjected to errors that would cause either over or under estimation of monthly expenditure and income. It pertains especially to income ('flow') data, as income is more volatile, and presents one of the main challenges to survey research, even in the context of detailed questionnaires.

The first objective of this paper is, therefore, to investigate whether households are able to smooth consumption across seasons, despite seasonal variation in income using very short recalls of income and expenditure (only at 15 days interval, unlike that of “usual month”). The second objective of the paper is to explore whether no seasonality in food consumption necessary mean that retained output for food consumption is sufficient for entire year.

The paper is organized as follows: in the next section theoretical framework and empirical model to test impact of income seasonality on consumption is presented. While section 3 gives the data used and discuses income and consumption seasonality, section 4 present empirical finding and calorie retained from own consumption. Section 5 presents concluding remarks of the research results.

2. Theoretical framework and empirical model

In order to analyse the impact of income seasonality on consumption let assume that preferences are quadratic, separable across periods, that the rate of time preference is equal to a fixed interest rate r, and preferences do not vary across time periods. Then, the Permanent Income Hypothesis (PIH) model state that the change in consumption from period to period can be expressed as:

… 1

Where C_jt denotes consumption of household j in time t and Y_jt is non-asset income of household j in time t. E_jt is the expectations operator, conditional on information known at time t. i.e., the change in consumption between t−1 and t is equal to the revision in permanent income between the two periods, where permanent income equals the annuity value of discounted future earnings plus the value of current assets. Equation (1) does not explicitly incorporate income seasonality, but since the equation is valid for any income process, it applies equally to situations with seasonal income variation. One can let “t” denote a month m rather than a year, and allow average income levels (and higher moments of income) to vary systematically across months as:

… 2

Where C_jm denotes consumption of household j in month m and Y_jm is non-asset income of household j in month m. E_jm is the expectations operator, conditional on information known in month m. Equation (2) implies that since consumption responds only to unexpected innovations in permanent income, deterministic seasonal patterns in income should have no effect on consumption. The expected value of the change in consumption between two months, conditional on information known in the earlier month, is zero. Therefore, (2) implies that on average there will be no seasonal consumption variation. More general models of consumption, for instance, those that allow for seasonal variation in preferences, prices or interest rates, or seasonal patterns in the resolution of income uncertainty do yield systematic seasonal patterns in consumption (see Chaudhuri, 1999). However, even in these more general models, seasonality in income levels will not directly translate into consumption seasonality and consumption will still not respond to anticipated income changes.

Paxson (1992) and Chaundhuri and Paxson (2001) modified equation (2) and express as:

... 3

Where ln(Cjm) is the logarithm of expenditure of household j in month m (see Paxson (1993) for detail derivation of equation 3). This variable is regressed on logarithm of average monthly income of household j (ln(Y_j)), set of dummy variables for each months ( ), a set of variables that are constant within a year, including the numbers of males, females, and children in the household at the beginning of the month and a constant (X_j) and a set of month dummies interacted with an indictor of households status (), which equal to 1 if the household head is male and equal 0 otherwise.

Measurement error, missing values of the data set, differences in unit of observation etc., usually causes bias in estimation. All possible measures were taken to reduce such errors during both data collection and data entry period. However, there may be possible bias due to gender of household head. The triple burden of household chores, childcare and agricultural works. Rural female household head has triple burden of household chores, childcare and agricultural works as compared to its male counterparts. Thus, inclusion of a set of month dummies interacted with and indictor of household status () is used to control such bias in specific months. If h _jm is 1 for male household head, the parameter measure month effects in expenditure for female household head. The month effect in expenditure for male household head are measured as + (see Wooldrige, 2003). The parameters represent the difference in male and female household head. An intercept is included in the model, so first month effect , is normalized to zero¹.

Likely, seasonality in income equation can be derived as a function of months and month-head interaction as:

… 4

Where A_jm is the ratio of household j’s total income in month m to the household’s average annual income (i.e. annual income divided by 12)². This variable is regressed on a set of dummy variables for each month and a set of month dummies interacted with households head gender status. In the same approach to equation (3) while measures the average monthly income share of female household head in respective villages, + measures addition month effects in income share for of male household head. And also represent the different in month effects between male and female household head.

3. Seasonal pattern in income and consumption

This section presents descriptive information on seasonal patterns in income and expenditure for households in four villages of Ethiopia. The data used come from 5^th round longitudinal monitoring survey for Deberebrehan and Yetmen villages in Amhara, Eteya village in Oromia, and Azedebo (Durame) village in Southern Nations and Nationality People (SNNP) regional states and carried out by Economics Department of Addis Ababa University (AAU) and United State Agency for International Development (USAID). In each village, a sample of sixty-two households was selected (about 16% are female headed) and hence total sample consists of 247 households³. These households were then interviewed frequently, typically every two weeks for a year, from April 2000 to July 2001 and asked question about socio-demographic characteristics, main activity and labor use, shocks and events, market information, crop output, food and non-food consumption expenditure etc. In this study information from May 2000 to May 2001 was used where complete information is found for almost all sampled households⁴.

The details of the survey allow me to construct income and consumption aggregate. Monthly expenditure is measured as the sum of expenditure on food (both purchased and value of consumption from own production and gifts) and non-food (clothes/shoes/fabric, household utensils, ceremonial expenses⁵ and others). Monthly income is measured as the sum of value of crop output during the month (crop cultivation), representing the value of food production that is retained for own consumption, gross income from sales of livestock/livestock products, cereals, coffee, chat, pulses, oilseeds etc., gross income from petty trade, off-farm labor for wage and any transfers in the form of remittance and gifts. All money variables are expressed in May 2000 Birr.

These villages represent different agro-ecological zones where crop cultivation is the primary sources of household income and profoundly depend on meher rainfall, usually lasting from June to the beginning of September. Crop cultivation is the dominant activity and source of livelihood. Table 1, presents major crop growing, percent growing and quantity produced (in kg/ha). The main crops grown include barely, wheat and beans in Debrebrehan, teff and maize in Yetmen and wheat in Eteya. More than 50% of the households have been reporting growing these crops in the respective villages. On average, household produced 13, 10 and 12qt/ha of barely, wheat and beans in Debrebrehan, 10 and 6qt/ha of teff and maize in Yetmen and 19qt/ha of wheat in Eteya, respectively. In Azedebo about 9qt/ha of wheat and maize were obtained although less than 40% of the households have been reporting growing these crops⁶.

The average yields were very low implying that household net income, mostly derived from these crops, were also low provide that price is usually unsatisfactory in Ethiopia. For instance, household monthly income is merely greater than 1000.00 Birr in Eteya (the potential cereal crops producing village) followed by almost Birr 650.00 in Debrebrehan and Birr 400.00 in Azedebo. In Yetmen average monthly income is limited to nearly 250.00 Birr (see Table 2). However, the average monthly income is less representatives for all households in a village as its standard deviation is very high (twice of average monthly income) (see Table 2). Such variability in income is expected as rural household income in Ethiopia is derived typically from crops cultivated during harvesting seasons and only little income is generating during slack seasons. This is an indication for seasonality in income of rural households.

Table 1: Percentage of households Growing Selected Crops and average yield produced by those households growing

Crop	Debrebrehan			Yetmen					Eteya		Azedebo
% Growing			Quantity produced (in kg/ha)		% Growing		Quantity produced (in kg/ha)	% Growing		Quantity produced (in kg/ha)		% Growing		Quantity produced (in kg/ha)
Teff	0.00	0.00		93.44		921.52 (335.21)			0.00	0.00	30.65		454.24 (322.70)
Barely	100.00	1336.72 (584.68)		0.00		0.00			22.58	1356.89 (950.88)	0.00		0.00
Wheat	83.87	952.62 (535.98)		29.51		782.70 (307.55)			95.16	1836.45 (730.21)	30.65		826.66 (440.39)
Maize	0.00	0.00		47.54		535.54 (329.61)			14.52	840.89 (329.61)	37.10		819.46 (703.11)
Sorghum	8.06	1049.99 (455.66)		0.00		0.00			20.97	1263.07 (567.72)	0.00		0.00
Bean	74.19	1140.81 (816.16)		0.00		0.00			37.10	1157.48 (254.90)	6.45		383.15 (161.29)
Vetch	0.00	0.00		40.98		805.64 (555.75)			0.00	0.00	0.00		0.00

Source: Own computation from survey data

An important feature of these villages is that income derived from crops has pronounced seasonal patterns. This is not surprising, given that farming in Ethiopia is rain-fed where agricultural cultivation depends heavily on meher rains. As indicated in Table 2, crop income accounts for a large share of total annual income ranging from 47.5% in Debrebrehan to 85.3% in Eteya. Although Ethiopia has high water potential, irrigation, which could potentially result in less seasonality in profits from crops, is fairly uncommon. Very recently, government has implemented water-harvesting scheme, its impact is not yet analyzed, however.

The combination of meher dependent crops and lack of irrigation results in flows of income from crops that are very unevenly distributed over the year, much more so than flows of income from other sources. Figure 1 traces out monthly averages of income from crops, and income from other sources (trade and off-farm, livestock etc.) The seasonal patterns in crop income are consistent with the timing of main harvesting seasons in each of the four villages. In Debrebrehan, Eteya and Azedebo harvesting on crop fields begins in mid October and usually lasts for about three to four months. In Yetmen harvesting commence a bit earlier, a mid of September. As shown in Table 2, the fraction of income received by households in the three highest-income months ranges from 38% in Azedebo to 60% in Eteya. The corresponding fraction of income is 53% in Yetmen and 49% in Debrebrehan. There is a second harvesting seasons between March and May. In this season Debrebrehan and Azedebo households are merely increasing their crops income.

Meanwhile, Azedebo is one of cash crop growing village in Ethiopia. At the village level coffee is the main cash crop, followed by tef (see Data et al., 1996). Coffee harvesting begins in June. This exceptionally favors households in the village to raise crops income from June through July. Household members also try as far as possible to supplement their farm income by other means. Small-scale trade is the first to be mentioned. People trade coffee, clothes and beverages such as local areki, tella, borde (see Data et al., 1996).

The other seasons are either pre-harvest or far post-harvest seasons know as slack seasons where crops income is either almost stagnant or far below the average crops income. For instance, in Yetmen crops income between August and September is significantly declines and it is usually the hunger season for most of households (see Tassew et al., 1996).

Table 2: Sample mean of income, consumption expenditure and related variables

	Debrebrehan	Yetmen	Eteya	Azedebo
Monthly income (in Birr)	649.43 (1027.981)	242.21 (437.25)	1092.38 (4353.8)	397.08 (767.95)
%age of annual income from crop cultivation	47.4	75.3	85.2	67.2
%age of income received in top 3 months	49.3	53.4	60.2	38.3
Monthly expenditure on food	60.34 (24.5)	36.91 (25.4)	99.11 (44.2)	60.8 (33.5)
Monthly expenditure on non-food	69.49 (67.10)	35.00 (28.7)	284.49 (872)	36.72 (22.8)
%age of annual expenditure on food	49.6	51.4	37.6	64.1
%age of food expenditure incurred in top 3 months	31.3	29.4	32.6	40.2
%age of non-food expenditure incurred in top 3 months	36.1	37.2	59.4	46.3

Source: Own computation from survey data.

Notes: 1. Average monthly incomes and expenditures are in May 2000 prices. The %age of income/expenditure in top 3 months is calculated as the sum of monthly incomes/expenditures in the 3-highest income/expenditure months of a year, divided by annual income/expenditure.

2. Figures under parentheses are standard deviations.

Fig 2 has shown that the higher degree of income seasonality in the household doesn’t translate into higher seasonality in consumption expenditure. Total, food and non-food expenditures are less erratic among months (seasons) from their respective annual average. In relative speaking, slight income seasonality is observed in Eteya. In this village all expenditures declines June through July and September through October, and increase July through September (due to main harvesting season Onion) and during main harvesting seasons (between October and February). In Azedebo extraordinary raises in all expenditure observed between July and October following coffee harvesting season in the villages. However, all expenditure seriously declines following month of October indicating low saving behaviors of households. January through May is the hunger seasons when there is only enset for consumption (see Data et al. 1996). Very slight expenditure seasonality is also observed in Debrebrehan proceeding main harvesting season and slack season of August through September (this is main harvesting seasons of potato in the village).

Moreover, closely examining of all expenditure has shown that the deviation of non-food expenditure is greater than the deviation of total and food expenditure from their respective annual mean expenditure. While households spend less on non-food items (clothes/shoes/furniture, unties etc.,) during pre-harvest and extreme post-harvest period (slack seasons), they spend more on it during main harvesting seasons.

Further exploring the nature of expenditure has shown that total food and non-food expenditures are less concentrated in the top three months as compared to income (see Table 2). Total food expenditure incurred in the top three months ranges from 29% in Yetmen to 32% in Eteya. In Azedebo it accounts up to 40% of the total food expenditure. In this village food expenditure incurred in the top three months is slightly greater than total agricultural income derived in the top three months (38.2%). This is not surprising as Data et al. (1996) indicated the village is a deficit production area and most people buy additional food for consumption. In the village households are more likely looking for additional wage employment including food-for-work program to mitigate income shocks. Nigussie (2006) indicated that more than 90% of households in the village were seeking to engage in additional food-for-work program.

The figure for non-food expenditure incurred in the top three months ranges from 36% in Debrebrehan to 46% in Azedebo. In Eteya it accounts for 59%, which is almost comparable with the total income received in the top three months, implying that most of income received in the top three months spends on non-food expenditure. This in turn implies that households in the villages are self sufficient in production for consumption and spent less on food items rather they spend on non-food items.

Food and non-food expenditure incurred in the top three months are very low provided that rural households in Ethiopia are mainly characterized by complete absence of credit market for consumption. Recently microfinance institutions are involved in providing rural credit (loan) with only group base collateral (no need of asset collateral). However, most of them engaged in provision of credit for investment (productive) purpose as one means of poverty alleviation rather than for short run consumption purpose. It is obvious that long-term is more benefiting them rather than short term.

Fig. 1: Income seasonality

Source: own computation from survey data.

Fig.2: Expenditure seasonality

3. Regression results

In the previous section we have seen that income, particularly crops income, have seasonal patterns than expenditure in all villages. In another words expenditure is relatively flat across seasons than income. This section uses an econometric tool and further explores seasonality in income and expenditure. Table 3 presents the results of F-tests of the significance of the month effects and the month-head gender interactions. While Tests 1 and Test 2 tests the null hypothesis of no seasonal effects in income and expenditure across male and female household head, Test 3 measures whether month-head gender interactions is jointly significant.

The F-tests presents in Table (3) consistent with the descriptive results indicated in Figure 1 and Table 2. First, the null hypothesis of no seasonal effect in income strongly rejected across male and female except for female household head in Yetmen i.e., in three villages both male and female household headed have seasonal pattern in income. Although Figure 2 depicts less erratic patterns of non-food expenditure, the regression result confirms seasonality in non-food expenditure in all villages except for female-headed household in Yetmen. In contrast to income and nonfood expenditure, the null hypothesis of no month effect cannot be rejected for food expenditure across both male and female-headed households. Thus, food consumption doesn’t track income seasonality. Thus, household not only draw down crops output retain food for consumption from their own harvest but also use different copying strategies to buffer food consumption seasonality (see Nigussie, 2006).

Secondly, the null hypotheses of identical month effects among male and female headed households can not be rejected in all villages i.e., month-gender of the family head interaction is jointly insignificant for income, food and non-food expenditure. This implies that there is no estimation is bias for any attribute resulting from male or female headed household members.

Table 3: F-statistics test for month effects in income and expenditure (F-statistics, P-value in parentheses)

	D/Brehan	Yetmen	Eteya	Durame
Observation (household- month)	713	781	698	664
Income
Test 1: no month effect: Female head	7.76(0.000)	0.38(0.970)	9.73(0.000)	2.33(0.0063)
Test 2: no moth effect: Male head	35.02(0.000)	10.11(0.000)	19.25(0.000)	3.05(0.0005)
Test 3: month effect identical	0.47(0.9317)	0.46(0.939)	0.71(0.745)	0.61(0.834)
R-square	0.442	0.090	0.079	0.144
Food Expenditure
Test 1: no month effect: Female head	1.56(0.0972)	1.56(0.0982)	1.02(0.424)	0.59(0.8491)
Test 2: no moth effect: Male head	1.07(0.3827)	1.54(0.1126)	1.13(0.3322)	2.61(0.0030)
Test 3: month effect identical	1.09(0.3653)	1.38(0.1687)	0.50(0.9146)	0.56(0.8773)
R-square	0.6520	0.5953	0.4574	0.496
Non-food Expenditure
Test 1: no month effect: Female head	3.75(0.000)	1.53(0.109)	2.90(0.0006)	3.16(0.0002)
Test 2: no moth effect: Male head	10.35(0.000)	4.60(0.0000)	3.41(0.0001)	11.38(0.000)
Test 3: month effect identical	0.91(0.5399)	0.71(0.7413)	1.51(0.1175)	0.78(0.6679)
R-square	0.4602	0.5301	0.4614	0.3146

Does rejection of seasonality in food consumption necessary mean that harvested output retained for consumption and draw down throughout the course of the year is sufficient to buffer seasonal consumption for the entire year? This insists on an assessment of food retained for consumption in terms of net calorie preserved for consumption. Table 4 presents net calorie available for consumption per adult equivalent per day (in Kilocalorie) by land holding tercile and sample average. Average net calories retained for consumption per adult equivalent during a year varies and ranges from 356.12 Kcal in Azedebo to 1276.66 Kcal in Eteya. Households of highest tercile have relatively better net calorie preserved for consumption than lowest and medium terciles in all villages. While net calorie preserved was 1036.07, 898.70, 1518.01 and 409.65Kcal/ae/day for highest terciles, it was only 490.66 and 640.55, 687.99 and 522.64, 1346.09 and 1518.01 and, 399.15 and 268.60 Kcal/ae/day for lowest and medium terciles in Debrebrehan, Yetmen, Eteya and Azedebo, respectively.

Given that the major source of calorie retained is derived from own harvest ((food preserved for consumption) see Annex I) mainly obtained during harvesting seasons (January-March) and the nuclear family size of the households (usually 4 to 6 persons) per households, the available calorie can sufficiently feed household families only for additional five months in Deberebrehan, six months in Yetmen and Eteya and two months in Azedebo. Likewise, the net calories available for households in highest tercile can feed families for a maximum of additional seven, eight and nine months in Debrebrehan, Yetmen and Eteya, respectively. For lowest and medium terciles households the preserved calorie can feed ranging from three to nine additional months (Table 4). However, in Azedebo, it couldn’t feed household families more than two additional months in either of tericle.

In general except households in the highest tercile of Eteya, households in other villages including the lowest and medium terciles in Eteya are in temporary declines in the households’ access to enough food for at least three months (i.e., in transitory food shortage) although the variability in consumption is less seasonal. This indicates that most farmers could not produce enough to meet the annual requirements. This clearly indicates that rejection of seasonality in consumption doesn’t necessarily mean retained output for food consumption is sufficient for the entire year.

Table4: Household adult equivalent and calorie preserved/ae/day (Kcal) by area of survey and land Tercile

Survey area/ Indicator		Land area per adult equivalent Tercile					Sample average
Lowest			Medium		Highest
Debrebrehan Cultivated land in ha/ae Adult equivalent Calorie available7/ae/day (Kcal)		0.09 - 0.31 5.02(1.32) 490.66(130.03)		0.31 - 0.44 5.46(1.46) 640.55 (313.95)		0.46 - 1.26 5.12(1.43) 1036.07 (1146.90)	5.19(1.39) 723.75(722.54)
Yetmen Cultivated land in ha/ae Adult equivalent Calorie available/ae/day (Kcal)		0.15 - 0.34 5.01(1.94) 687.99 (1154.16)		0.34 - 0.51 4.63(1.56) 522.64(168.80)		0.52 - 1.67 3.46(2.23) 898.70 (394.35)	4.37(2.00) 700.15 (710.48)
Eteya Cultivated land in ha/ae Adult equivalent Calorie available/ae/day (Kcal)		0.02 - 0.33 8.00(2.86) 1346.09 (2669.42)		0.34 - 0.44 6.58 2(2.28) 950.33(286.04)		0.44 - 0.98 6.03(2.82) 1518.01(622.43)	6.87(2.76) 1276.66(1595.45)
Azedebo Cultivated land in ha/ae Adult equivalent Calorie available/ae/day (Kcal)		0.01 - 0.07 6.87(1.97) 399.15(572.65)		0.07 - 0.14 5.76(1.68) 268.60(167.96)		0.14 - 0.49 5.56(1.71) 409.65(301.14)	6.06(1.85) 356.21(380.01)

Source: Own computation from survey data. Figures in parenthesis are standard deviation

Note: Figures under parentheses are standard deviation.

4. Conclusion

The paper has explored whether rural household in Ethiopia who experience seasonal income variation also experience seasonal consumption, particularly food consumption variation. The major finding of this paper is that, at household level, seasonal consumption and income patterns are largely unrelated i.e., food consumption patterns at household level are similar despite dramatic difference in the timing of income flows. Furthermore, although there is seasonality in income and non-food expenditure, the paper has shown identical income, food and non-food expenditure between male and female household head.

Although rural household in Ethiopia usually characterized by complete absence of credit market for consumption, they characterized by accumulating crops output for food consumption during main harvesting season and draw down the entire year to smooth consumption. However, the finding has shown that retained food for consumption (in kilocalorie) is insufficient to provide food throughout the course of the year. Households in the study villages are in a transitory food shortage at least for three months. It is more severe in Azedebo where households are in this shortage for at least five months.

As retained calorie derived from crops harvested is not sufficient, new strategies should be designed to boost production more than one. Very recently the government of Ethiopia launched water harvesting scheme to overcome not only rainfall shortage but also to produce more than one even in a year with good rainfall condition in high potential areas. This scheme should be widely practiced provided that it is properly managed. Whenever the land is repeatedly tilled its fertility is deteriorated hence, proper land management for sustainable development is also essential.

References:

Chambers, R., Richard, R. and Arnold, C., 1981. Seasonal Dimensions to rural povery. London: Frances Pinter Limited.

Chaudhuri, S., 1999. Forward-looking behaviors, precautionary saving, and borrowing constraints in a poor agrarian economy: tests using rainfall data, Columbia University Department of Economics discussing paper #9899.

Dercon, S. and Krishnan, P., 2000. Vulnerability, seasonality, and poverty in Ethiopia. Journal of Development Studies 36(6): 25-53.

Data, D., Berihun, D. and Alemu, S., 1996. Ethiopian Village studies: A qualitative study. Phillippa Bevan and Alula Pankhust (eds). Azedebo, Kembata.

Paxson, C. and Chaunduri, S., 2001. Smoothing consumption under income seasonality: buffer stocks vs. credit. Photocopy. Colombia and Princeton Universities Discussion paper.

Paxson, C., 1993. Consumption and Income seasonality in Thailand: Journal of Political Economy, 101(1): 39- 72.

Nigussie, T., 2006. Consumption smoothing and vulnerability in rural villages of Ethiopia. A paper presented on EEA, 3^rd International Annual conference.

Sahn,.D (ed.), 1989. Seasonal variability in third world Agriculture: the consequences for food security, Baltimore: Johns Hopkins University Press.

Tassew, S., Berihun, M. and Gebeie, B., 1996. Ethiopian Village studies: A qualitative study. Phillippa Bevan and Alula Pankhust (eds). Yetmen, Enmay, Gojjam.

Walker, T. and James G., 1990. Villages household Economies in India’s Semi-Arid tropics, Baltimore: Johns Hopkins University Press.

Wooldridge, J., 2003. Introduction to Econometrics. A modern Approach (2eds). Thomoson South-western University Press.

Annex I: Sources of calorie

Source	Debrebrehan	Yetmen	Eteya	Azedebo
Food retained for consumption Purchased food Received as remittance etc	92.39 3.40 4.21	94.95 4.51 0.54	94.03 5.28 0.69	63.09 32.41 4.50