Longitudinal Analysis Using Panel Data for Assessing Seasonality Effects on the Food Security Situation in Tajikistan 2005 Hbs
Issue: Longitudinal analysis using panel data for assessing seasonality effects on the food security situation in Tajikistan 2005 HBS Tajikistan: Longitudinal analysis using panel data for assessing seasonality effects on the food security situation in Tajikistan 2005 HBS RAMASAWMY, Seevalingum FAO Statistics Division, Rome. Italy Household income and expenditure survey (HIES) usually collects food data from households at only one period which may refer to one week, two weeks or one month.
Most HIES extend the field work over the entire period of one year to account for any seasonal effects of household expenditure particularly food consumption.
or any similar topic only for you
The survey estimates assume seasonal effects cancelled out in large groups of households but not at the level of the individual household. Thus, the inter-household variation estimated on the basis of such data would tend to include the seasonal effects. However, the Tajikistan Household Budget Survey (HBS) presents a particular characteristic that it collects expenditure and income data from the same household over a long period of time.
The longitudinal design survey accounts for all variations including the seasonal effects when analysed over the months of the yearly period. This paper presents some trend analysis of food security statistics derived from the 2005 Tajikistan household panel monthly food consumption data collected from the sample of 925 households and evaluate the impact of the variability of the distribution of the food consumption in the food security statistics estimates. Keywords: Food consumption data, Food security statistics, Panel data, Dietary energy consumption, Food Deprivation, Critical food poverty, Coefficient of variation.
Acknowledgements: FAO (Statistics Division and Food Security Information for Action Programme) for technical assistance and the European Community for financial support. 1. BACKGROUND Tajikistan is a landlocked country, largely mountainous and sparsely inhabited, 90 percent is mountainous and the total area splits the country into four regions (Oblasts) and one independent city, the national capital Dushanbe. Only seven percent of the land area is arable; cotton and wheat are the main important crops.
Aluminium is the major country resource together with other limited mineral resources such as silver, gold, uranium, and tungsten. With abundant water resources, it possesses much hydropower facilities which are however not well distributed among its population. The civil war (1992-97) severely damaged the already weak economic infrastructure and caused a sharp decline in industrial and agricultural production. While Tajikistan has experienced steady economic growth since 1997, nearly two-thirds of the population continues to live in poverty. Economic growth reached 10. percent in 2004 but dropped to eight percent in 2005 and to 7 percent in 2006. Tajikistan’s economic situation, however, remains fragile due to uneven implementation of structural reforms, weak governance, widespread unemployment, and the external debt burden. Unemployment is officially estimated at 30 percent, while the figure is likely to be much higher. Lack of alternative sources of livelihoods continue to exacerbate household food insecurity and results in under-employment in the agricultural sector, while a large number of young men seasonally or definitely migrate for employment in other CIS countries.
There is a high mobility of the working population to Russia, where more than half a million of the population are currently working. The 2007 Tajikistan population was about seven million; nearly 70 percent live in rural areas. The annual population growth is about 2 percent. 2. OBJECTIVES, METHODS AND DATA The paper analyses the trends of food security statistics derived from the monthly and quarterly food consumption data of the Tajikistan 2005 HBS. It also evaluates the trend variations of inequality measures of dietary energy onsumption due to other factors such as income and area of residence and their effects on the measurement of food deprivation. The Tajikistan State Committee of Statistics has been conducting household budget survey (HBS) based on the Soviet methodology collecting household consumption expenditure from a fixed sample of households over time. A nationally representative sample of 925 households was selected from the 2000 population census data frame using the multi-stage stratification. Rural and urban areas together with criteria of mountains, valley, uplands, lowlands and country borders on the north and south were accounted for.
The households were selected at the last stage using the available administrative data with regard to the composition of the household. Household detailed expenditure including food and income data are collected using daily records from the same 925 households over years since January 2000. Each household receives a monthly incentive equivalent to one dollar in local currency. The Tajikistan HBS collected consumption and expenditure data from 925 households over the year using eight different types of questionnaires which enable the collection of complementary expenditure data on a daily, monthly and quarterly basis.
Food data are recorded in detail, – stock at the start of the month for each food item, purchases, own production, transfers, (aid, gifts, etc. ) during the month on a daily basis, and closing stock at the end of the month. Income is also collected by sources on a daily and monthly basis. SCS uses a detailed nutrient conversion table covering dietary energy, protein, fat and carbohydrate values for computing nutrient values. The Tajikistan 2005 HBS monthly food consumption data together with the household income were analysed using the FAO statistical procedures of the food security statistics module (FSSM).
The paper compares the food security statistics (FSS) estimates from the two sets of data namely the twelve sets of monthly food consumption data and the quarterly aggregated data of the 925 households. The food consumption in terms of dietary energy and expenditure are examined together with the dietary energy unit value at the national level and by the four main regions, Dushanbe, RPR, Sogd and Chatlon and by income quintiles. The inequality measure of food assess is studied in much details to evaluate the variation of area of residence and income over the months of 2005.
Measures of prevalence of hunger, food deprivation and critical food poverty are also discussed.Lastly the food expenditure share of total consumption and the diet diversity are compared for the two sets of data. 3. FOOD SECURITY STATISTICS DERIVED USING THE LONGITUDINAL APPROACH. a. Dietary energy consumption The average daily dietary energy consumption (DEC) of the Tajik was 2150 kcal in 2005. The DEC by regions and income levels showed wide fluctuations over the months of the year 2005 as illustrated in Figures 1 and 2 below.
The population of the capital city Dushanbe and RPR regions had lower DEC levels than the national level during all the months of the year. These two regions had low food production as Dushanbe is the capital city and RPR is the region of aluminium ores and had to rely on food imports from other local regions or imports from neighbouring countries. However, Sogd, the industrial region and Chatlon, the cotton and wheat growing region had DEC higher than the national level almost all the months of the year.
These two regions have good food availability as they contain the largest cropping areas cultivating crops such as potatoes, barley, melons, etc. Figure 1: Trends of DEC by Regions Figure 2: Trends of DEC by Income levels |[pic] |[pic] | Analysing the DEC by daily per person income quintiles showed a gradual increasing in the overall monthly level of dietary energy consumption from the lowest to the highest income population groups.
The population of the three lowest income groups had DEC below the national minimum dietary energy requirement (MDER) of 1880 kcal/person/day during all the months of the year. Those of the two highest income quintiles had DEC well above the national average DEC. The fluctuations in the DEC of the four lowest income groups over the months were small and less irregular than those observed among regions. However, the highest income population group had more pronounced fluctuations which kept increasing over the months with peaks in March and October.
The dietary energy consumption among the Tajik population related more to income levels than place of residence. The levels of DEC for the population of the first four income quintiles did not differ greatly in magnitude. The average daily per person income ranged from 0. 91 Somoni for the lowest quintile to 1. 91 Somoni for the fourth one while the highest quintile had a much higher average of 3. 20 Somoni. This group of high income population which were present in all the four regions may probably influence those observed fluctuations.
March and October 2005 were the two months registering high peaks in dietary energy consumption, probably linked to national socio-cultural or religious events. More than 90 percent of Tajiks are Sunnis and were most probably fasting in October 2005 which was the month of Ramadan in Tajikistan. It is a well known fact that during that special religious month, there is a high acquisition and consumption of food in terms of both quality and quantities particularly among the high income levels households.
In addition, there is much sharing of food among the community with a large part of food given away by households and at the same time received by other households. However, the recording of such data on food transfers did not take place, thus leaving its effect unknown. b. Food expenditure The national average daily per person monetary values of the food expenditure fluctuated over the months of the year with the lowest value (1. 12 Somoni) in February and the highest value (1. 68) in October when there was that overall high level of consumption.
The analysis by regions showed that the population of Dushanbe had a relatively low level of dietary energy consumption, but food expenditure higher than the national level indicating that prices in the capital city were higher than in other parts of Tajikistan probably due to a high importation of food products from other regions or countries. The industrial region of Sogd showed a high level of food expenditure ranging from 1. 17 to 1. 61 Somoni slightly higher than the average food spending in Dushanbe. The population of RPR had the overall lowest food expenditure (Figures 3).
Figure 4 shows the monthly trends of food expenditure by income levels. Again the hierarchical differences from highest to lowest income quintile are clearly observed. Food expenditures for the population of the three lowest income quintiles were lower than the national level for all the months of the year. The amount of money spent on food remained at almost the same levels, but with an increase in October followed by a decrease in November and another increase in December probably due with the end of the year celebrations events.
The population of the highest income quintile had increasingly high food expenditures with peaks in the three last months of the year of 2005. Figure 3: Trends of food expenses by Regions Figure 4: Trends of food expenses by Income levels |[pic] |[pic] | Figure 4 shows the monthly trends of food expenditure by income levels. Again the hierarchical differences from highest to lowest income quintile are clearly observed.
Food expenditures for the population of the three lowest income quintiles were lower than the national level for all the months of the year. The amount of money spent on food remained at almost the same levels, but with an increase in October followed by a decrease in November and another increase in December probably due with the end of the year celebrations events. The population of the highest income quintile had increasingly high food expenditures with peaks in the three last months of the year of 2005. c. Dietary unit value
The national average dietary unit value was 0. 57 Somoni per 1000 kcal. This value varied from 0. 55 Somoni for the months of February and March to 0. 63 Somoni in December. The population of the capital city Dushanbe had the highest dietary energy unit value over all months of the year of 2005, paying abnormal high values in the months of January (0. 71 Somoni) and October (0. 75 Somoni). Population of Sogd had also a high overall dietary energy unit value which increased slowly over the months of 2005 (see Figure 5).
It is surprising to note that the dietary unit value fell in all regions in the month of November before going up again in December. This could probably be due to a fall in food prices resulting in a surplus of food items on the market due to the end of the religious month of October. Figure 5: Dietary energy unit value by Regions Figure 6: Dietary energy unit value Income levels |[pic] |[pic] |
The dietary energy unit value showed marked increasing patterns over the months of the year when analysed by income levels with again a drop in the values in November followed by an increase in December. The lowest quintile population had an overall yearly dietary energy unit value of 0. 49 Somoni compared to a value of 0. 66 Somoni for the highest income quintile. d. Diet Diversity The share of total calories of nutrients in total calories of dietary energy showed a protein deficiency diet when compared to the WHO norms (Figure 7). The share contribution of protein was around nine percent, while the |Figure 7: Share (%) of nutrients in total calories and WHO guidelines | |WHO minimum and maximum values are 10 and 15 percent respectively. The| | |share contribution of fats was within the WHO norms, but the share | | |contribution of carbohydrates (70 percent) was more towards the | | |maximum value of 75 percent.
Consumption of protein food sources such| | |as pulses, fish, meat or dairy products were very low. | | | |[pic] | The regional analysis of the share of protein calorie contribution to total calories is given in Figure 8 and showed large and uneven variations were observed among the regions over the months of the year.
The population of Dushanbe were more protein deficient (almost below 9 percent in all months except July to September) than other regions while the population of RPR had relatively, though still deficient, higher protein consumption. There were two months (July and September) when there was increasing protein consumption in all regions probably due to availability of protein rich food products coming from the harvest seasons.
There was no clear difference in the level of protein consumption among population groups of different income levels over the months of the year (Figure 9), except that all income groups showed the same increasing patterns over the months of July and September, before falling down in October. This situation could be the due to the scarcity of high protein food products on the Tajikistan markets or highly prohibitive selling prices if available.
Figure 8: Share of protein calorie by Regions Figure 9: Share of protein calorie by Income levels |[pic] |[pic] | e. Inequality. The distribution of dietary energy consumption is assumed to be lognormal and its variance is a function of the Coefficient of Variation (CVx). This CVx includes income and biological sources of variations of dietary energy consumption and is a measure of access to food.
The biological variation (CVr) accounts for factors such as sex age composition, body weight and physical activity of household members. The CVr is estimated as a value of 20 percent. |Figure 10: CV of dietary energy consumption due to income by 2005 quarters and |Figure 10 gives the inequality of food access due to income | |months |computed for groups of households classified according to the | | |income deciles.
The national CVx of dietary energy consumption| | |(DEC) on yearly basis had a value of 30 percent which included| | |a value of 22 percent corresponding to the CV of DEC due to | | |income.
The CVx values estimated for quarterly data differed | | |marginally from the yearly CVx. The estimated monthly CVx was | | |less than the yearly CVx, except for the months of October and| | |November. |[pic] | | A striking observation is that the monthly inequality measures of DEC due to income were in most cases (except the three last months) less than the quarterly or yearly values which are inflated with other variations due to inter household, seasonal (within quarter) and other non random factors.
Those variations were analysed using the available 2005 longitudinal food data with between household factors of area of residence and income using a linear model of the log of dietary energy consumption with repeated measures (months). The analysis of variance results are shown below. |[pic] |The variation estimates in the rightmost column have been | | |converted to the original dietary energy consumption scale for| | |better understanding.
As expected area of residence (rural and| | |urban) and income levels (deciles) were significant sources of| | |the between household variation (standard deviation of 2062 | | |Kcal/ person/ day).
This variation reflects sources of | | |variations not included in the model, the random variation and| | |the undesirable variation due to sampling design and | | |instrumental errors.
The within sources of variation were | | |significant in time (months) and time within area of residence| | |and time within income levels. | The within household variation (standard deviation of 608 Kcal/person/day) was smaller than the between household variation. However there is a significant source of variation due to seasonality. In this study one should take into account that the between household variation estimates may be over-estimated as result of the sampling design.
This study does not address on this design effect on the variation between households. In commonly used household survey design where the household reference period is of one month or less and households are allocated over a one-year survey period the sources of the within variation (month, month within area, month within income and error) is added to the survey estimates. This means that in NHS considering random allocation throughout the year add variation to the between household CV and hence over-estimate the prevalence of food deprivation. f. Food deprivation.
The longitudinal data of Tajikistan 2005 HBS was also analysed by comparing the food deprivation over the four quarters to study the food consumption distribution and any improvement in the level of undernourishment over the one year period. Four percent of the population had moved out from the food deprived population over the year due to a 2 percent increase in the average daily dietary energy consumption and a three percent point decrease in the coefficient of variation (CV) due to income from 31 to 28 percent. The MDER of 1880 kcal/person/day was the same for both quarters.
Movements of population from food deprived population were observed in both urban and rural areas by two and six percent respectively. There were marginal increases in DEC were noted in both urban and rural areas, but the later registered a significant 5 percentage point decrease in the CV due to income. While there were significant improvements among the population of the three lowest income quintiles, there were small set backs for the populations of the two highest quintiles groups due mainly to an around 2 percent fall in their dietary energy consumption (Figures 11 and 12).
Figure 11: Food deprivation (%) by RegionsFigure 12: Food deprivation (%) by Income levels |[pic] |[pic] | The prevalence of food critical poverty which measures food income deprivation showed a marginal fall of 1 percentage point at national level from quarter one to quarter 4 of the year 2005 (Figure 13).
Figure 13: Food critical poverty (%) by quarters of 2005 national and sub national levels |[pic] | | | |The prevalence of critical food poverty fluctuated over the four | | |quarters of the year.
There was a high prevalence of food poverty | | |in the second quarter at the national and sub national levels. It | | |then fell in the following quarters. However, food critical | | |poverty in urban areas was higher than rural areas due to the food| | |availability at lower prices. g. Depth or Intensity of Food Inadequacy Figures 14 and 15 below show the depth of food poverty with relation to the MDER over the four quarters of the year, at the national and sub national levels and by income levels respectively. The Figures show that urban areas which had a low DEC had a higher food deficit than the national and rural areas. The food deficit was less in the fourth quarter at the national level and in urban and rural areas. This was also true for the income levels due to the high level of DEC observed as from October 2005.
There was a general high food deficit during the third quarter in almost all the regional or economic population groupings and the high income group witnessed a food deficit of about 150 kcal/person/day. Figure 14: Food deficit (Kcal) to MDER Regions Figure 15: Food deficit (Kcal) to MDER Income levels |[pic] |[pic] | h. Food share
The share of food expenditure to total consumption expenditure estimated from the annual household data at national level was 69 percent. The share of food expenditure at national level showed an erratic trend over the four quarters of the year 2005. From a level of 68. 4 percent in the first quarter, it increased to 69. 6 percent, fell down to 65 percent in the third quarter and rose to 65. 3 percent in the last quarter. The same patterns occurred in urban and rural areas, but with higher magnitude in rural areas and lower values for urban areas (figure 18).
However, the food share showed a decreasing trend over the four quarters with increasing income levels with a high food share (80 percent) among the population of the low income group to about 50 percent for those of the highest income group. The second quarter had the maximum food share in most of the population groupings (Figure 19) which could be a period of harvest of some food crops. Figure 16: Food expenditure share by Region Figure 17: Food expenditure share by Income levels |[pic] |[pic] |
Figures 18 and 19 illustrate the share of dietary energy consumption by food sources at national and sub-national levels and by income levels respectively for the four quarters of 2005. Purchase was almost the only source of dietary energy consumption for the population of the urban regions, while own production food constituted a significant share of DEC to the order of about 40 percent in most of the other population groupings. There was little variation in own production contribution between the quarters, apart from some high share in quarters one and four for the high income level groups, probably due to the harvesting season.
Figure 19: Share of DEC by food sources & Figure 18: Share of DEC by food sources & Regions Income levels |[pic] |[pic] | 4. CONCLUSION The analysis of the longitudinal food consumption data of Tajikistan 2005 HBS provides some useful and pertinent characteristics of food security statistics: • Food deprivation differed by seasons and by income levels. • Food consumption is seasonal and is influenced by national ocio-religious events. • Food demand was high in high income levels during specific periods. • Dietary energy unit value differed with seasons and income levels. • Diet consumption of nutrients was affected over the seasons. • Food consumption from purchases were not affected by seasons while that from own production varied over the months of the year. • There was a seasonal affect on the diet consumption of nutrients • Food inequality or access measures were low when estimated with monthly data and the use of more aggregated data caused overestimation. The intensity of hunger differed by season and income levels • Food share varied with seasons and level of income. REFERENCES 1. FAO (2003). Methodology for the measurement of food deprivation. Statistics Division, Food Security Statistics. Rome. Available at the Metadata of the Food Security Statistics webpage http://www. fao. org/faostat/foodsecurity/index_en. htm 2. FAO (2006). Food Security Statistics Module, Step 1 – Processing User Manual, Step 2 – Analysis User Manual and Step 3 – Reports User Manual, FAO July 2006. . Sibrian R Ramasawmy S and Mernies J (2007). Measuring hunger at sub national levels from household surveys using the FAO approach: MANUAL. FAO Statistics Division Working Paper Series No. ESS/ESSA/005e. Available at the webpage. http://www. fao. org/es/ess/faostat/foodsecurity/Papers_en. htm . 4. Tajikistan Food Insecurity Assessment report derived from the food consumption data of Tajikistan 2005 HBS, Dushanbe August 2007. http://www. stat. tj/english/home. htm[pic]