A European workshop to introduce the EU SILC and the EU LFS data Practical Session Exploring EU SILC. Heike Wirth & Pierre Walthery

Transcription

1 A European workshop to introduce the EU SILC and the EU LFS data Practical Session Exploring EU SILC Heike Wirth & Pierre Walthery Exercise 1: Severe material deprivation rate by age, sex, at risk of poverty status and country Exercise 2: Very low work intensity by country BASICS IMPORTANT TO KNOW WHEN WORKING WITH EU SILC DATA EU SILC data are provided in four files, containing two different level of responses (household; individual): (a) household level (1) Household register file (D File): contains basic household data and information regarding the selection of the sample. (2) Household data file (H File) contains basic household data, household income, social exclusion, housing. (b) individual level (3) Personal register file (R File): contains a record with basic information such as sex, date of birth, year of immigration for every person currently living in the household (4) Personal data file (P File) contains a record with personal information on education, labour market, health, income etc. for all household members aged 16+ for whom the information could be completed. Linking EU SILC datasets To use the individual information (data or register) with the household information (data or register) you must combine the files. The matching variables (or key variables) are: Year: DB010, HB010, RB010, and PB010 Country: DB020, HB020, RB020, and PB020 Household ID: DB030, HB030, RX030 and PX030 Personal ID: RB030 and PB030. => the 1st letter of the variable name corresponds to the file it is included, e.g. HS010 (= H File) 1

2 Extract: DESCRIPTION OF TARGET VARIABLES: Cross sectional and Longitudinal, 2008 operation (Version January 2010) HS010 HS010_F Flag Variables Nearly all variables in EU SILC do have a so called flag variable [= variable name e.g. HS010 with the suffix "_F" e.g. HS010_F]. The flag variable gives additional information to the value of the main variable. E.g. a negative flag variable specifies the reason why the main variable is blank. => See Appendix 1 for the list of variables incl. in the EU SILC User data base , flag variables are not included in the variable list. 1 Source: 1_ pdf/_EN_1.0_&a=d 2

3 Exercise 1 Severe material deprivation rate by age, sex, at risk of poverty status and country Step 1: Compute the indicator material deprivation rate Definition: Material deprivation refers to the inability for individuals or households to afford those consumption goods and activities that are typical in a society at a given point in time, irrespective of people s preferences with respect to these items. Definition EU SILC material deprivation: severe material deprivation: household is deprived of at least 3 of 9 items (see list of items below) household is not able to afford any 4 of the 9 items Material deprivation items are HS010; HS020, HS030: ARREARS (mortgage or rent (HS010), utility bills (HS020) or hire purchase instalments (HS030)) HS040: CAPACITY TO AFFORD PAYING FOR ONE WEEK ANNUAL HOLIDAY AWAY FROM HOME HS050: CAPACITY TO AFFORD A MEAL WITH MEAT,CHICKEN,FISH (OR VEGETARIAN EQUIVALENT) Every 2 nd DAY HS060: CAPACITY TO FACE UNEXPECTED FINANCIAL EXPENSES HS070: DO YOU HAVE A TELEPHONE (INCLUDING MOBILE PHONE)? HS080: DO YOU HAVE A COLOUR TV? HS100: DO YOU HAVE A WASHING MACHINE? HS110: DO YOU HAVE A CAR? HH050: ABILITY TO KEEP HOME ADEQUATELY WARM You need to: Create a variable that flags household with missing data for any of the above variables (a) The material deprivation items refer to the household level => H File (h_household_data.dta) (b) Households with missing values (flag variables equal to 1) in at least one of the material deprivation items are excluded from the calculation. gen cases_excl = 1 replace cases_excl =0 if hs010_f!=-1 & /// hs020_f!=-1 & hs030_f!= -1 & /// hs040_f!= -1 & hs050_f!= -1 & hs060_f!= -1 & /// hs070_f!= -1 & hs080_f!= -1 & hs100_f!= -1 & /// hs110_f!= -1 & hh050_f!= -1 3

4 (c) Compute 9 dichotomous material deprivation items (ma_de_it1 to ma_de_it9) with 1 = deprived, 0 = not deprived. Note: in the case of arrears (HS010, HS020, HS030) the dichotomous variable will take the value 1 for any type of arrear; in the case of mobile phones, computers and colour TV, values 2 and 3 are recoded 0. gen ma_de_it1 = 1 if (hs010 == 1 hs020 == 1 hs030 == 1) replace ma_de_it1 = 0 if (hs010 == 2 hs010_f == -2) & /// (hs020 == 2 hs020_f == -2) & /// (hs030 == 2 hs030_f == -2) *** Note: The indicators can be created individually as show below... recode hs040 1=0 2=1, gen(ma_de_it2) recode hs050 1=0 2=1, gen(ma_de_it3) recode hs060 1=0 2=1, gen(ma_de_it4) recode hh050 1=0 2=1, gen(ma_de_it9) Or by using the programming commands of Stata: We need first to rename the variables below so that we can automatize the recoding (be careful hs090 is not used as indicator!) rename hs070 newhs070 rename hs080 newhs080 rename hs100 newhs090 rename hs110 newhs0100 We can now type in the program, which consists of a loop that will create 4 dummy variables. Note: N is a system counter variable that is used to name the indicators MA_DE_IT5...8 whereas the local command creates another system variable that matches the name of the original hs variables. forval n=5/8 { local n2=2+`n' recode newhs0`n2'0 (2=1) (1 3 =0), gen(ma_de_it`n') } (d) Compute a new variable (ma_de_sum) which sums up ma_de_it1 to ma_de_it9. Use the variable created in (b) to recode cases with missing values as system missing (.). egen ma_de_sum = rowtotal(ma_de_it*),missing 4

5 replace ma_de_sum=. if cases_excl==1 The same result can also be achieved with a simple addition of all variables. We can now inspect the results: tab ma_de_sum. tab ma_de_sum, m ma_de_sum Freq. Percent Cum. 0 7, , , Total 12, (e) Compute a material deprivation indicator (mat_dep) = 1 (deprived) if ma_de_sum greater equal 3; = 0 (not deprived) if ma_de_sum less 3. Label it gen mat_dep=. replace mat_dep= 1 if (ma_de_sum >=3 & ma_de_sum ~=.) replace mat_dep= 0 if (ma_de_sum < 3 label define lmat_dep 0 Deprived 1 "Not deprived" label values mat_dep lmat_dep tab mat_dep. tab mat_dep mat_dep Freq. Percent Cum. Deprived 10, Not deprived 1, Total 11, (f) Compute a severe material deprivation indicator (s_mat_dep) = 1 (deprived) if ma_de_sum greater 3; = 0 (not deprived) if ma_de_sum less equal 3. Label it accordingly gen s_mat_dep = 1 if (ma_de_sum > 3 & ma_de_sum ~=.) /*deprived */ replace s_mat_dep = 0 if (ma_de_sum <= 3) /* not deprived */ label define ls_mat_dep 0 "Not deprived" 1 "Depr. at least 4 items" 5

6 label values s_mat_dep ls_mat_dep ta s_mat_dep. tab s_mat_dep s_mat_dep Freq. Percent Cum. not deprived 11, deprived (g) Rename HB010, HB020 and HB030 to year, country, hid (= variables needed for linking household & personal data files!) rename hb010 year rename hb020 country rename hb030 hid Total 11, (h) Sort data by year country and hid sort year country hid (i) Save the file (e.g. h_depr.dta) save h_hhld_data.dta, replace (j) SAVE FILE (e.g. h_depr.dta) save h_hhld_data.dta, replace Step 2: Compute the share of households in (severe) material deprivation (as well as the number of deprivation items) by country. What are the main findings?. tab ma_de_sum country, col nofreq country ma_de_sum AT UK Total Total

7 . tab mat_de country, col nofreq country mat_dep AT UK Total not deprived deprived Total tab s_mat_de country, col nofreq country s_mat_dep AT UK Total not deprived deprived Total Step 3: Calculation of the breakdowns (age, sex, at risk of poverty status and country) The new computed material deprivation indicators refer to the household level, but the interest is on breakdowns on the individual level The household data and person register file each contain a year, country and household ID variables [HB010, HB020, HB030 and PB010; PB020, PB030]. These variables (or key variables) enable us to relate the household data records to the persons in the personal register file. In this way, we can make a rectangular data file that contains data from both files by distributing information from the household data to the individual data. 2 Step 3a: Merging of personal information (R FILE) with the household data (H FILE) in such a way as to keep all rows from the R File Open the personal register file (R FILE) use r_personal_register.dta Rename RB010, RB020 and RB030 to year, country, hid: when merging files the key variables must have identical names. (Optional: You can also rename RB090 SEX to avoid confusion later). rename rb010 year rename rb020 country rename rb030 pid rename rx030 hid ta _merge Sort the dataset by RB010 RB020 RB030 and save it 2 See also the guide provided by ESDS Government: Working with survey files: using hierarchical data, matching files and pooling data (as of 7/20/2011) 7

8 sort year country hid pid save r_pers_register.dta, replace Use the merge command by year country hid to merge the R FILE with h_hhld.dta merge m:1 year country hid using h_hhld_data.dta Save the file (e.g. h_r_matched.dta) save h_r_matched.dta, replace Step 3b: Compute and label a new variable for age group using RX020 with the following categories 1 ' 0 17' 2 '18 24' 3 '25 54' 4 '55 64' 5 '65+' recode rx020 (0/17 = 1) (18/24 = 2) (25/54 = 3) /// (55/64 = 4) (65/max = 5), gen(age_group) Step 3c: Compute the share of population being (severe) material deprived by age group; sex (RB090), at risk of poverty threshold (HX080) and country. What are the main findings? Note: There is no explicit command to create three or four ways tables in Stata. However, with the bysort prefix, you can produce nested two way tables for each combination for the categories specified after bysort which is almost the same. The full command bysort country age_group sex : ta hx080 mat_dep, row will generate 20 such two way tables, which is to much to show in this document. Below is a sample output: -> country = UK, age_group = 25-54, sex = Male Key frequency row percentage mat_dep poverty indicator Deprived Not depri Total income>=60% median eq 2, , income <60% median eq Total 3, ,

9 -> country = UK, age_group = 25-54, sex = Female Key frequency row percentage mat_dep poverty indicator Deprived Not depri Total income>=60% median eq 3, , income <60% median eq Total 3, , > country = AT, age_group = 25-54, sex = Male Key frequency row percentage mat_dep poverty indicator Deprived Not depri Total income>=60% median eq 1, , income <60% median eq Total 1, , > country = AT, age_group = 25-54, sex = Female Key frequency row percentage mat_dep poverty indicator Deprived Not depri Total income>=60% median eq 1, , income <60% median eq Total 1, ,

10 Exercise 2: Very low work intensity by country Step 1: Compute the indicator persons living in households with very low work intensity by country The work intensity of the household refers to the ratio between the number of months that all working age household members have been working during the income reference year and the total number of months that could theoretically have been worked by the same household members. Definitions: Working Age Person = person aged 18 64, not being a dependent child 3. persons living in households with very low work intensity = the share of persons with the work intensity of the household below the threshold set at 0.20 The information needed to compute the work intensity indicator are included in the personal data set (P FILE) Step 1a: Merging of personal data (R FILE) with our own data set r_h_hhld.dta (a) Load the personal data file (P FILE) (b) Sort the data set by PB010 PB020 PX030 PB030 [PB030 = personal id. Needed because the data are now matched at the individual level] (c) Rename PB010, PB020, PX030 and PB030 to year, country, hid, pid (d) Use the merging procedure by year country hid pid to merge the P FILE with r_h_depr.dta (e) SAVE FILE (e.g. p_r_h_depr.dta) Step 1b: Computing two new variables (NWA & NW) for each working age person 4 NWA: Number of workable months = The number of months during the income reference period for which information on the person s activity status is available NW: Number of worked months The number of months during the income reference period for which the person has been classified as worker List of variables needed see next page 3 'Dependent children' include all persons aged below 18 as well as persons aged 18 to 24 years, living with at least one parent and economically inactive. Which will be ignored here for practical reasons 4 Source: UDB variables description ver from doc, p

11 11

12 Step 1b (continued): Computing two new variables (NWA & NW) for each working age person 5 (a) Before summing the variables, set their value to 0 if the information about the activity status is missing (flag variable = 1) (b) compute the variable nw = PL070 + PL072 + PL080 + PL085 + PL087 + PL090 (c) compute the variable nwa = PL070 + PL072 (d) Referring to working age persons: (AGE = PX020) If PX NWA = NW = 0 If PX NWA = NW = 0 (e) Calculate for each household then the the total sum of workable months (sum_nwa) the total sum of worked months (sum_nw) (f) Compute the work intensity of the household (WI) that is the ratio between the total sum of worked months and the total sum of workable months WI = sum _ NW sum _ NWAm (g) Define a new variable low_work WI = 0 low_work = 0 0 < WI< 0.2 low_work = < =WI< 0.5 low_work = W< 1 low_work = 3 W = 1 low_work = 4 where Step 1c: 0 'jobless household' 1 'very low work intensity' 2 'low work intensity' 3 'medium work intensity' 4 'high work intensity'. Compare the distribution of the low work indicator across countries. What are the main findings? 5 Source: UDB variables description ver from doc, p

13 Appendix 1: Variables EU SILC 2007 Source: 7/2007 1_ pdf/_EN_1.0_&a=d 13

14 14

15 15

16 16

17 17

18 References Income poverty and material deprivation in European countries RA /EN/KS RA EN.PDF 18