Gustavo De Santis gustavo.desantis@unifi.it Mauro Maltagliati mauro.maltagliati@unifi.it How close? Silvana Salvini silvana.salvini@unifi.it An attempt at measuring the cultural distance between countries Comparing families: does international perspective help? Warsaw, 17 18 December 2013
Database: World Value Survey http://www.worldvaluessurvey.org/
21 European Countries Wave Label Country 1981-1984 1989-1993 1994-1999 1999-2004 2005-2007 Total BG Bulgaria - - 1 072-1 001 2 073 HR Croatia - - 1 196 - - 1 196 CZ Czech Rep. - 924 1 147 - - 2 071 EE Estonia - - 1 021 - - 1 021 FI Finland 1 003-987 - 1 014 3 004 FR France - - - - 1 001 1 001 DE Germany - - 2 026-2 064 4 090 HU Hungary 1 464-650 - - 2 114 IT Italy - - - - 1 012 1 012 LV Latvia - - 1 200 - - 1 200 LT Lithuania - - 1 009 - - 1 009 NL Netherlands - - - - 1 050 1 050 NO Norway - - 1 127-1 025 2 152 PL Poland - 938 1 153-1 000 3 091 RO Romania - - 1 239-1 776 3 015 SK Slovakia - 466 1 095 - - 1 561 SI Slovenia - - 1 007-1 037 2 044 ES Spain - 1 510 1 211 1 209 1 200 5 130 SE Sweden - - 1 009 1 015 1 003 3 027 CH Switzerland - 1 400 1 212-1 241 3 853 GB Great Britain - - 1 093-1 041 2 134 ALL 2 467 5 238 20 454 2 224 16 465 46 848
About a thousand questions... But not all are asked in all (21, EU) countries or waves. In all cases, too many. Therefore we group them in 9 domains (latent dimensions) No. of Domain Variables Variables 1. Religion 7 A006, A040, A098, E069_01, F028, F034, F063 2. Family (traditional) 5 A001, A042, D023, D054, E019 3. Friends/Associations 4 A002, D055, MEMBER*, ACTIVE* 4. Happiness/Health 4 A008, A009, A170, C006 5. Self 9 A003, A005, A029, A030, A032, A034, A039, A173, E039 6. Gender 5 C001, D057, D059, D060, E069_15 7. Ethics 4 F114, F115, F116, F117 8. Work 4 A005, C009, C010, E040 9. Politics 17 A004, A165, C002, E001, E002, E003, E004, E005, E006, E012, E023, E025, E033, E035, E041, E069*, Y001
Purpose: to assess as objectively as possible how close (or how far) any two countries are (Italy and Poland, notably: is their inclusion in the FAMCHIP project justified?) How do we proceed? (next, animated PPT, please)
Domain (or Dimension) 1: Religion 7 elementary variables 1) How important is Religion is in your life? 1 'Very important' 2 'Rather important' 3 'Not very important' 4 'Not at all important 2) List of qualities that children can be encouraged to learn at home. Religious faith? 1 'Mentioned' 2 'Not mentioned'
3) Are you a member of Church or religious organizations? 0 'Not a member' 1 'Inactive member' 2 'Active member 4) How much confidence do you have in the Church? 1 'A great deal' 2 'Quite a lot' 3 'Not very much' 4 'None at all' Domain 1: Religion 5) How often do you attend religious services? (Apart from weddings and funerals) 1 'more than once a week' 2 'once a week' 3 'once a month' 4 'only holy days' 5 'once a year' 6 'less often' 7 'never'
Domain 1: Religion 6) Would you say you are: 1 'A religious person' 2 'Not a religious person' 3 'A convinced atheist' 7) How important is God in your life? 1 'Not at all'... 10 'Very much
Clusters We let the (SAS) software classify the 39 thousand odds respondents in clusters, each cluster encompassing homogeneous persons (from the point of view of religion) How many clusters? 3, 5, 10, 15, 20, 30, 50 What clustering method? Ward s (or minimum variance) method Then we compare the distribution of nationals among clusters, assuming that countries with similar distributions (of individuals among clusters) are similar to each other.
Religion / 10 clusters DiSIA
A synthesis of distances for Italy and Poland Religion (10 clusters) Sum of squared differences in proportions, for all pair of countries. Then, for all countries, we have a distribution of distances. For instance...
The number of cluster matters! RELIGION High/Low= Average, plus or minus one std. dev. for each country How to make results independent of the number of clusters? Standardization
Standardized results Italy and Poland are close to each other (Note: objective distances are the same. But relative distance represented here are not)
Other «clear» domains: IT and POL are close
In some domains: IT and POL are «average»
Unclear domains (results not robust) Family = traditional vision of Happiness and health: subjective assessment of
All in all... DiSIA
Conclusions (Within the limits of our data and our approach: method, cluster criterion, measure of distance, domains retained, etc.) IT and PL seem to be relatively close to each other, within the European context. The approach is relatively new: it starts from individual data but produces an output that makes sense at the aggregate level (distribution). Assumptions: clusters create homogeneous groups, and the distribution of nationals among groups is a valid indicator of distance between countries. Limitations of the data: small national samples, and observed over sometimes long periods of time. (Are answers stable? Do they make sense?) Results generally (but not always) robust to a change in the number of clusters.
Gustavo De Santis gustavo.desantis@unifi.it Mauro Maltagliati mauro.maltagliati@unifi.it How close? Silvana Salvini silvana.salvini@unifi.it An attempt at measuring the cultural distance between countries Comparing families: does international perspective help? Warsaw, 17 18 December 2013