Background Te National Statistics Centre (NSC) of te Lao Peoples Democratic Republic as conducted tree expenditure and consumption surveys in te last decade. Te first Lao Expenditure and Consumption Survey (LECS-I) was conducted in 1992/93. Te second, LECS-II was conducted in1997/98. A trd survey, LECS-III, was conducted in 2002/03. Ts tecnical report concerns te LECS-III survey. LECS is a multi- purpose survey. Suc surveys can be designed in various ways wit empasis on particular issues. In 1992/93 te LECS was combined wit a large module of social indicators, Lao Social Indicator Survey (LSIS). Te 1997/98 and 2002/03 versions focused on economic activities of te ouseolds. Te objectives of te LECS are basically to provide, macro estimates for te National Accounts, bot private consumption and ouseold investments and income from agriculture and businesses te consumption structure (weigng system) for te Consumer Price Index (CPI) estimates of labor force statistics on access to services statistics on nutrition statistics on poverty and income distribution Te survey is conducted troug interviews wit ouseolds and village cairmen. Te interviews are evenly spread over one full year. Te survey also as a primary scool module administered to scool officials and teacers. Sample design Te LECS-III is designed essentially as LECS-II wit regard to sample design. Experiences from LECS-II ave been used to fine-tune te sample design. Te number of primary sampling units (PSU) as been increased and te sample size witn PSU reduced, giving a somewat smaller sample in terms of ouseolds. Te sample consists of 8100 ouseolds selected troug a two-stage sample design. Villages serve as primary sampling units (PSU). Te villages are stratified on eigteen provinces and witn provinces on urban/rural sector. Te rural villages are furter stratified on villages wit access to road and no access to road. In all, te village population was divided into 18 (provinces) x 3 (urban/rural classes) = 54 sampling strata. Te total first-stage sample consists of 540 villages. Te sample is allocated to provinces approximately proportionally to te square root of population size according to population census. Te PSUs are selected wit a systematic probability proportionate to size (PPS) procedure in eac province. Table 1 sows te sample allocation over strata.
Table 1: Number of sample villages in eac stratum. Province Urban villages Rural villages wit access to road Rural villages witout access to road Total Vientiane C. 36 12. 48 Pongsaly 3 6 15 24 Luangnamta 4 11 9 24 Oudomxay 4 8 12 24 Bokeo 3 18 3 24 Luangprabang 4 19 13 36 Huapan 2 16 18 36 Xayabury 6 25 5 36 Xiengkuang 5 8 11 24 Vientiane 4 29 3 36 Borikamxay 5 11 8 24 Kammuane 6 26 4 36 Savannaket 9 27 12 48 Saravane 3 12 21 36 Sekong 4 7 7 18 Campasack 6 27 3 36 Attapeu 2 9 7 18 Xaysomboun SR 1 11. 12 Total 107 282 151 540 Te ouseolds in te selected villages were listed prior to te survey. 15 ouseolds were selected wit systematic sampling in eac village, giving a sample of 8100 ouseolds. Data collection LECS III contains six modules: a diary to record ouseold transactions, a ouseold questionnaire a time use diary, a price collection form, a village questionnaire, a primary scool questionnaire. A large part of te ouseold questionnaire remains te same as in previous surveys, except for some modifications in questions tat didn t work well in LECS-2. Some topics, mainly ousing, ealt, agriculture and time use, ave been expanded. Te ouseold interview is consequently somewat longer in LECS-3. Data on expenditure and consumption are collected for a wole mont based on daily recording of all transactions. At te end of te mont te ouseold is asked about purcases of durable goods during te preceding 12 monts. During te mont eac member of te ouseold
sould record te time use during a 24-our period. Te rice consumption of eac member of te ouseold is measured for one yesterday to get a more precise measure of intake at eac meal for eac person. Te village questionnaire is administered to te ead of te village. Te questionnaire covers roads and transport, water, electricity, ealt facilities, local markets, scools etc. Te interviewers also conduct a primary scool survey in te area. Interviews are made wit te scool principal and te teacers. Te survey collects data on pupils, scool facilities, management of te scool, finances and sources of support. Te principal and te teacers are also asked questions about te decision-making in te scool. Te price collection form is used by te interviewers to collect local prices on 121 commodities. Table 2: Overview of te content of te LECS-III survey odule Contents Data specified for: Diary All ouseold transactions during sampled monts. Houseold Transactions coded to consumption/expenditure, ouseold business, agriculture and investment outlays Houseold questionnaire Time use Prices Village questionnaire Primary scool module Houseold composition Parents Education Labour force participation Victimization Nutrition Healt ceck, measurements of eigts and weigts Possession of durables and assets values Housing conditions - ouseold Construction activities - ouseold Houseold business Agriculture - ouseold Healt evaluation of ealt, use of ealt services, ealt seeking beavior, ealt costs Purcases and selling of durables during te last 12 monts Income and transfers by all members of ouseold Borrowing and lending by ouseold Time spent recorded for a period of 24 ours in a sampled day for 22 activities Prices for 92 basic goods and services recorded in nearest local market Data provided by village eads on situations in te village concerning: - demograpy - access to services - agriculture - general economic conditions - wages and prices Covers 470 primary scools in or close to te selected villages All ouseold members Non-ouseold member 6 years and above 10 years and above Houseold All ouseold members Cldren 4 years and below Houseold Houseold Houseold By business Houseold All ouseold members, costs for ouseold Houseold All ouseold members Houseold 10 years and above Village Primary scool
Field work Te statistical provincial offices were in carge of te field operations wit supervision from NSC. Data on expenditure and income were collected for a wole mont based on daily notation of all transaction divided into consumption, agriculture production and ouseold businesses. Interviews wit ouseold eads or oter ouseold members were eld during various parts of te mont. At te end of te mont ouseolds were asked about purcases of durable goods, e.g. furniture, TV, cars, motorcycles, etc., during te preceding 12 monts. In te middle of te mont one 24 our period was selected to record data on time use for all persons in te ouseold ten years and above. Te measurement of daily consumption troug a diary kept by te ouseold puts a eavy burden not only on te ouseolds but also on te field interviewers. any ouseolds, especially in te rural areas, need frequent support in te task of keeping te diary. In order to secure an acceptable quality in te data it as been deemed necessary to keep te interviewers in te village for te wole mont rater tan aving te interviewers traveling to te villages for repeated interviews and follow-up. Ts decision is also supported by te fact tat many villages, especially in te mountainous areas, are difficult to access (some villages require travel by foot for several days). In LECS-1 and LECS-2 te fieldwork was done by teams of two interviewers in eac village. For LECS-3 a single-interviewer design was considered. However, in te final analysis factors related to interviewers security and well-being weiged in favor of aving two interviewers in te village. Te field staff consisted of 180 interviewers organized in 90 two-member teams. 36 supervisors from te provincial statistical offices and 10 central supervisors from ead office supervised te teams. Calculation of sampling weigts Houseold survey Te process of calculating weigts is indicated in file final weigts.xls. Te weigts for ouseold j were calculated as: w j = n m were: m = number of ouseolds in stratum according to village register = number of ouseolds in villagei in stratum according to village register = number of ouseolds in villagei in stratum according to survey listing = number of ouseolds in te sample from villagei in stratum
m is usually = 15 but in a few cases it is less tan 15 due to nonresponse. To adjust for ouseold nonresponse te weigt w j is multiplied by 15/ m were m = number of responding ouseolds in te village. A calibration factor is calculated for eac stratum.. K Kˆ were: K = mid-survey population in stratum according to official statistics Kˆ = mid-survey population in stratum estimated from te survey Summary: w j = n 15 15 m K Kˆ weigt = First stage sampling weigt Second Stage Sampling weigt Adjustment for HH nonresponse Calibration to known population totals Te final weigts are stored in SQL-table final weigts. Day correction factors (daycorr) ave been calculated. Te factors are =1 for most ouseolds and > 1 for ouseolds wc ave not reported for te wole mont. If a ouseold as diary records during a number of days tat is less tan te total number of days in te mont, te ouseold is cecked. If te last day of recording is te last or next to last day of te mont it is considered OK and te ouseold get te factor = 1. If te last day of recording is at least tree days before te end of te mont, te remaining days of te mont are considered as days wit missing data and te factor assigned to te ouseold is te total number of days in te mont divided by te number of days aving diary records. A new variable is created: diarywgt=daycorrweigt. Ts is te ouseold weigt to be used for estimations of diary.
In te master data set a second set of weigts popweigt - are included. Te popweigt is equal to te ouseold weigt (weigt) multiplied by te number of ouseold members. Variation in ouseold sampling weigts Due to poor measures of size Due to calibration Village survey Te village weigts ave been calculated as: w j = n In te master data set a second set of weigts popweigt - are included. Te popweigt is equal to te village weigt multiplied by te total population in te village. Correction for unit non-response in te village survey Tree villages are missing in te village survey. Ts was taken care of by re-weigting. Te sampling weigts in province 8, 10 and 11 ave been adjusted to compensate for te non-response. Data entry cecks, editing, imputation Range and consistency cecks Te data entry program ad some built-in range cecks but most of te cecks were done in batc-mode. A large number of range cecks and some consistency cecks were designed in a Visual Basic program. A special problem was te cecks of te values of expenditures and consumption recorded in te diary. Te values in Kip are often large numbers wit many trailing zeroes. A common mistake during te data entry is to enter one zero too muc or too less. Some obvious errors of ts type can be detected in te range cecks, especially wen te Kip value is very g. In some cases it was also possible to detect errors by relating te values to te recorded quantity. However, also te quantities contain errors tat make tem difficult to use as cecks.
Te Time Use Diary contained a lot of errors due to te use of te 12-our time notation used in te diary. Te data entry staff was supposed to convert te data to te 24-our time notation at te data entry stage but tat was not done correctly in many cases. A rater tedious and time-consuming consistency ceck and correction of a large number of records ad to be done. Correcting food consumption for ouseolds wit no rice consumption in te diary Tere are 50 ouseolds tat ave not reported any consumption of rice in te diary but ave reported daily consumption of rice in te nutrition module. Ts could of course be true if te ouseold as made a bulk purcase in te mont preceding te survey mont. It was, owever, decided to impute rice consumption for tese ouseolds for te poverty calculations (as was made in te LECS-II). Correction for double-counting of rice Some ouseolds ave misunderstood te instructions and recorded bot purcases and daily consumption of purcased rice. Corrections of te most obvious cases ave been made in te database. Te following diary entries are deleted in te database: All entries were itemid = 1 and kind = 2 and kip< 15,000 All entries were itemid = 2 and kind = 2 and kip< 20,000 All entries were itemid = 127 and kind = 2 in ouseolds were te total number of suc entries is >26 and te total value of itemid 1 is > 80,000 Housing: odel for imputing values on rent Approximately 1000 ouseolds ave non-response or unreasonably low values on estimated annual rent (p9s1q2). A regression model as been developed relating rent to a number of quality caracteristics. Te data file is ousing_regression3.sav located in folder Data processing/ oter modules. Syntax is in ousing_regression.sps and output in ousing_regression3.spo. Te result of te regression is presented in table 1. Table 3: Regression results ousing. Dependent variable: Log(rent) Factor Regr coefficient t-value Probability Est value on factor (Constant) 11.27971 296.48 0 79198 Size in square m 0.001036 5.28 < 0.0001 Area Urban 0.533263 13.71 < 0.0001 1.70 Rural 1.00
Wall Bricks, concrete 0.994476 18.49 < 0.0001 2.70 Unburnt bricks, wood 0.340541 10.74 < 0.0001 1.41 Oter 1.00 Roof RODERN 0.50235 13.74 < 0.0001 1.65 Oter 1.00 Floor FODERN 0.413085 10.61 < 0.0001 1.51 Oter 1.00 Toilet TODERN 0.439998 3.79 0.000152 1.55 TNORAL 0.522449 16.32 < 0.0001 1.69 Oter 1.00 Region Vientiane 0.228002 4.13 < 0.0001 1.26 Nort -0.12956-3.98 < 0.0001 0.88 Sout -0.16189-4.38 < 0.0001 0.85 Central 1.00 Sampling errors Standard errors and confidence interval ave been calculated for te most important estimates. Design effects and rates of omogeneity (ro) easurement errors Te interviewers spent a lot of time in te ouseolds assisting te respondents in teir task of recording all transactions relating to te ouseold as well as ouseold businesses and agricultural operations. Tere are reasons to believe tat ts tedious and time-consuming work improved te quality of te responses. Tere is anecdotal evidence tat te frequent visits to te ouseold by te interviewer in many cases establised a relaxed and trustful relation between te parties. It also gave te interviewers ample time to sort out te often-complicated relations between ouseold consumption and ouseold production in agriculture or ouseold businesses.
Cecks against external information A few cecks of quality were made. Te estimates of rice consumption from te survey was cecked against external agricultural production data and found to agree reasonably well. Livestock Planted areas Respondent fatigue A ceck on consumption levels between te first and te second two-week diary period was also made. Table 4 sows te proportion of entries over te two alf parts of te mont. Te expected proportions are 49.3 and 50.7 if we assume an entirely even pattern over all days in te mont (te proportions differ from 50% because te second period is sligtly longer). Table 4: Percent distribution of number of entries over first and second alf of te mont Item group Days 1-15 Days 16 - end of mont Food expenditure 50.4 49.6 Consumption of own produced food 49.6 50.4 Clotng and footwear 51.4 48.6 Housing 49.7 50.3 Houseold utensils and operations 50.4 49.6 edical care 50.6 49.4 Transport and communications 51.5 48.5 Education 50.5 49.5 Personal care 51.2 48.8 Recreation 50.6 49.4 Alcool and tobacco 50.0 50.0 Oters 50.1 49.9 Total 49.9 50.1 Tree item groups - clotng and footwear, transport and communications and personal care - seem to ave a somewat ger proportion of entries during te first alf mont. Te oter item groups ave proportions fairly close to te expected proportion. Te fact tat tere were very small differences in consumption on aggregate level between te first and te second two-week diary period raises te question weter a sorter diary period migt be sufficient to capture te consumption.
Te average number of entries in te diary per ouseold was 150 in LECS-III and 145 in LECS-II. Table 5: Average number of entries of food items in te diary Itemgroup LECS2 LECS3 Rice 2.2 2.2 Oter cerals and bread 4.7 3.5 eat 6.5 6.1 Fis 4.7 4.3 ilk, ceese and eggs 1.2 1.3 Oils and fats 0.5 0.4 Vegetables and potatoes 11.9 11.1 Fruits 2.5 2.2 Sugar and sweets 2.7 2.4 Non-alcoolic beverage coffee & tea 2.6 1.9 Oter food 4.2 3.5 eals 5.6 4.2 Own produced rice 21.5 24.3 Own produced oter grains 0.4 0.5 Own produced meat 3.8 7.5 Own produced fis 12.7 9.7 Own produced fruits 1.3 1.6 Own produced vegetables 27.2 27.0 Oter own produced 3.4 3.2 Total number of food entries 119.6 117.0 Interviewer effects Riceballs in Kammuan and Savannaket provinces. Brief comments on measurement problems in some variables Wages, salaries