Dia 1

1
MOBILITY AND HETEROGAMY EFFECTS SOBELS DIAGONAL
REFERENCE MODELS Wout Ultee Radboud University
Nijmegen Presentation at the Observatoire
Sociologique du Changement, Sciences-Po,
Paris June 15 2007
2
WHAT ?! SOBELS DIAGONAL REFERENCE MODELS ?!
THE UMPTEENTH TECHNIQUE OF DATA ANALYSIS
?! YES, BUT A TECHNIQUE TO ANSWER QUESTIONS
WHICH ARE BETTER THAN QUESTIONS OF OLDER
GENERATIONS
3
A SUCCESSION OF FOUR TECHNIQUES OF DATA ANALYSIS
IN SOCIAL MOBILITY RESEARCH LIPSET
GENERATION PERCENTAGES DUNCAN GENERATION PATH
MODELS GOLDTHORPE GENERATION LOGLINEAR
MODELS MAYER GENERATION EVENT HISTORY MODELS
THESE TECHNIQUES ARE APPLICABLE IN HETEROGAMY
RESEARCH TOO
4
QUESTIONS OF THE LIPSET GENERATION IF A
PERSON HAS MORE EDUCATION THAN ANOTHER PERSON ,
IS THIS PERSON MORE LIKELY TO MOVE UP ? IF
PERSONS HAVE MORE EDUCATION THAN THEIR PARENTS ,
ARE THEY MORE LIKELY TO BE UPWARDLY MOBILE ?
HOW MUCH OF TOTAL MOBILITY IS STRUCTURAL MOBILITY
AND HOW MUCH CIRCULATION MOBILITY ? ACCORDING TO
LATER GENERATIONS , THESE QUESTIONS ARE POOR OR
MEANINGLESS
5
STYLIZED LIPSET GENERATION FACTS FOR RAMDOM
SAMPLES TAKEN FROM A COUNTRYS POPULATION THE
CORRELATION OF EDUCATION WITH UPWARD AND DOWNWARD
MOBILITY IS DISAPPOINTINGLY LOW AND SO IS THE
CORRELATION OF THE DIFFERENCE BETWEEN FATHERS
AND SONS YEARS OF EDUCATION WITH THE DIFFERENCE
BETWEEN FATHERS AND SONS OCCUPATIONAL PRESTIGE
6
DUNCAN GENERATION QUESTIONS ABOUT MOBILITY AS
ONE VARIABLE AND ABOUT DIFFERENCE SCORES IN
GENERAL POSE THE ISSUE POORLY THESE QUESTIONS
UNWITTINGLY INVOLVE FLOOR AND CEILING EFFECTS
AND THEREFORE REGRESSION TOWARDS THE MEAN IS
BEING DISCOVERED QUESTIONS SHOULD NOT BE ABOUT
DIFFERENCE SCORES BUT ABOUT THEIR COMPONENTS
7
THE RIGHT QUESTION ACCORDING TO THE DUNCAN
GENERATION TO WHAT EXTENT DOES A PERSONS
OCCUPATIONAL PRESTIGE DEPEND NOT ONLY UPON THIS
PERSONS EDUCATION , BUT ALSO UPON THE EDUCATION
AND OCCUPATIONAL PRESTIGE OF THIS PERSONS FATHER
? TECHNIQUE TO ANSWER THESE QUESTIONS PATH
MODELS
8
GOLDTHORPE GENERATION IF THE DISTINCTION
BETWEEN STRUCTURAL AND CIRCULATION MOBILITY HAS
ANY MEANING , IT HAS SO AT THE MACRO LEVEL ONLY
NO SOUND ANALYTICAL TECHNIQUE UNTIL NOW HAS
BEEN FOUND TO GIVE MEANING TO THE EQUATION TOTAL
MOBILITY STRUCTURAL MOBILITY CIRCULATION
MOBILITY
9
THE RIGHT QUESTIONS ACCORDING TO THE GOLDTHORPE
GENERATION HOW HIGH ARE ABSOLUTE MOBILITY
RATES ? HOW HIGH ARE RELATIVE MOBILITY CHANCES ?
HOW STRONG IS SOCIAL FLUIDITY AND HOW UNEQUAL
ARE COMPETITIVE OUTCOMES ? HOW LARGE ARE
MARGINAL DIFFERENCES ? HOW UNBALANCED IS THE
COMPETITIVE STRUCTURE ? AND TO WHAT EXTENT ARE
COUNTRY DIFFERENCES IN ABSOLUTE
MOBILITY EXPLAINED BY DIFFERENCES IN RELATIVE
MOBILITY CHANCES COMPARED WITH DIFFERENCES IN
COMPETITIVE STRUCTURES ?
10
THE RIGHT EQUATION ACCORDING TO THE GOLDTHORPE
GENERATION COMPETITIVE OUTCOMES PLUS
COMPETITIVE STRUCTURES RESULT IN ABSOLUTE
MOBILITY QUESTIONS ABOUT COMPETITIVE OUTCOMES
ARE ANSWERED BY LOGLINEAR MODELS
11
MAYER GENERATION QUESTIONS LIKE HOW MUCH
SOCIAL FLUIDITY IS THERE IN THE U.K. IN 1972
? ARE TOO VAGUE FLUIDITY REFERS TO THE MOVEMENT
OF PERSONS BETWEEN TWO POINTS IN TIME THIS SHOWS
THAT THE REFERENCE TO 1972 IS INCOMPLETE
12
MAYER GENERATION QUESTIONS LIKE IS THERE MORE
SOCIAL FLUIDITY IN THE U.K. IN 2000 THAN IN 1980
? ARE ALMOST SUPERFLUOUS QUESTIONS SHOULD NOT
BE ABOUT OVERLAPPING POPULATIONS BUT ABOUT
SUCCESSIVE COHORTS
13
THE RIGHT QUESTION ACCORDING TO THE MAYER
GENERATION ARE FOR SUCCESSIVE COHORTS THE
CHANCES FOR MOVING UP OR DOWN IN SOME ALWAYS
EQUALLY LONG TIME SPAN NOT ONLY DEPENDENT UPON
INDIVIDUAL CHARACTERISTICS LIKE AGE AND EDUCATION
BUT ALSO UPON COHORT AND PERIOD CHARACTERISTICS
? THESE QUESTIONS ARE TO BE ANSWERED WITH EVENT
HISTORY MODELS
14
IN THE BEGINNING THE APPLICATION OF MODELS THAT
DID NOT TAKE INTO ACCOUNT THE CONTEXTUAL NATURE
OF PART OF THE ORGINALLY INDIVIDUAL DATA
CONTEXTUAL THE UNEMPLOYMENT RATE AT THE MOMENT
OF A POSSIBLE JOB SHIFT CONTEXTUAL JOB
STRUCTURE AT THE END OF COMPULSORY
SCHOOLING NOWADAYS APPLY MULTI-LEVEL EVENT
HISTORY MODELS
15
ARE MAYER GENERATION QUESTIONS FLAWLESS ? MAYER
QUESTIONS EXPLAIN DIFFERENCE SCORES ! IT IS TRUE
THAT IF THE START SCORE FOR OCCUPATIONAL PRESTIGE
IS INCLUDED AS A PREDICTOR IN EVENT HISTORY
MODELS , ESTIMATES OF PARAMETERS FOR OTHER
EFFECTS DO NOT DIFFER FROM THOSE MODELLING END
SCORES BUT STILL , WHY EXPLAIN DIFFERENCE SCORES
RATHER THAN END SCORES ?
16
FIFTH GENERATION QUESTIONS ? THE MAYER
GENERATION COLLECTED OCCUPATIONAL HISTORIES ,
BUT MODELLED SINGLE EVENTS FIFTH GENERATION
QUESTIONS SHOULD BE ABOUT CAREERS IS AN
OCCUPATIONAL CAREER A SELF-REINFORCING UPWARD
PROCESS ? WILL A PERSON WHO ONCE MOVED DOWN BE
PERMANENTLY SCARRED ? MERTONS MATTHEW EFFECTS
GO BEYOND BECK, RISK SOCIETY , BAUMAN , LIFE IN
FRAGMENTS , SENNETTS BOOKS
17
(No Transcript)
18
MOBILITY QUESTIONS ARE NOT ABOUT MOBILITY BUT
ABOUT CONDITIONS THAT CAUSE PERSONS TO ATTAIN
HIGHER AND LOWER CLASSES OR BETTER AND WORSE JOBS
19
HOW ABOUT QUESTIONS ON THE EFFECTS OF MOBILITY
? ARE THERE FALSE QUESTIONS ABOUT EFFECTS OF
MOBILITY ? ARE THERE GENERATIONS OF QUESTIONS ON
MOBILITY EFFECTS AND HETEROGAMY EFFECTS ? YES ,
AND THERE ARE NICE PARALLELS
20
THREE GENERATIONS IN RESEARCH ON MOBILITY AND
HETEROGAMY EFFECTS
LIPSET GENERATION PERCENTAGES OF PHENOMENON
X FOR CLIMBERS, STAYERS, AND
SKIDDERS DUNCAN GENERATION LINEAR REGRESSION
MODELS WITH INTERACTION TERMS SOBEL
GENERATION NON - LINEAR REGRESSION
MODELS AGAIN TECHNIQUES NO GOALS IN
THEMSELVES NEW TECHNIQUES ANSWER BETTER QUESTIONS
21
LINEAR REGRESSION MODELS WITH INTERACTION TERMS
ARE MATHEMATICALLY EQUIVALENT TO NON - LINEAR
REGRESSION MODELS BUT IN THE DUNCAN GENERATION
INTERACTION PARAMETERS ARE DIFFICULT TO
INTERPRET WHEREAS IN THE SOBEL GENERATION
PARAMETERS STAND FOR INTERESTING SUBSTANTIVE
HYPOTHESES
22
EFFECTS OF MOBILITY AND HETEROGAMY ON WHAT
? VOTING MARITAL FERTILITY DIVISION OF
HOUSEHOLD LABOUR FEMALE LABOUR MARKET
PARTICIPATION VISITING A MUSEUM , CONCERT HALL ,
THEATER OTHER CULTURAL ACTIVITIES LIKE READING
BOOKS
23
THE LIPSET GENERATION QUESTION ARE CLIMBERS
MORE LIKELY THAN STAYERS TO VOTE FOR THE
RIGHT AND SKIDDERS FOR THE LEFT ?
24
LIPSET GENERATION RESULTS WERE MIXED THIS LED TO
A POLISHING OF FINDINGS NEXT TO MOBILITY ,
CLASS OF ORIGIN OR CLASS OF DESTINATION WAS
INTRODUCED AS A CONTROL THIS COMPOUNDED THE
DIFFICULTIES FLOOR AND CEILING EFFECTS BECAME
EASY TO SPOT
25
LIPSETS TABLE FOR THE USA IN 1948 PERCENT VOTING
REPUBLICAN AFTER OWN CLASS AND CLASS OF ORIGIN
MANUAL NON-MANUAL ORIGIN MANUAL 20
70 NON-MANUAL 40 60
70 IS MORE THAN 60 UPWARDLY MOBILE PERSONS
OVER-CONFORM LIPSET COMPARES AN OFF-DIAGONAL WITH
A MAIN-DIAGONAL PERCENTAGE
26
DUNCAN , WHEN EXPLAINING MARITAL FERTILITY , HELD
LIPSET GENERATION QUESTIONS TO BE PREMATURE THE
RIGHT QUESTION ACCORDING TO THE DUNCAN GENERATION
IS AFTER TAKING INTO ACCOUNT EFFECTS OF
PRESENT OCCUPATION AND EFFECTS OF ORIGIN , ARE
THERE EFFECTS OF MOBILITY AS SUCH ?
27
DUNCAN ANSWERED THIS QUESTION WITH LINEAR
REGRESSION MODELS INCLUDING ON TOP OF MAIN
EFFECTS INTERACTION EFFECTS
28
THE SOBEL GENERATION QUESTION WHAT DO PERSONS
WEIGH MORE HEAVILY IN CHOOSING BETWEEN
ACTIVITIES THEIR ORIGIN OR THEIR
DESTINATION? THEIR OWN GROUP OR THAT OF THEIR
PARTNER?
29
SOBEL HOLDS THAT DUNCAN GENERATION QUESTIONS ARE
IMPRECISE THEY PERTAIN TO INTERACTIONS ,
WITHOUT SPECIFYING IN ADVANCE THE TYPE OF
INTERACTION
30
SOBEL REJECTS LIPSET GENERATION QUESTIONS THEY
POOL ALL STAYERS WHEREAS QUESTIONS ABOUT
CONFORMITY TO THE NORMS IN A PERSONS DESTINATION
GROUP AND ABOUT REJECTION OF NORMS IN A PERSONS
ORIGIN GROUP IMPLY A COMPARISON OF CLIMBERS AND
SKIDDERS WITH TWO SPECIFIC TYPES OF STAYERS
31
(No Transcript)
32
PERCENT OF WOMEN VOTING LABOUR AFTER OWN CLASS
AND CLASS HUSBAND IN BRITAIN 1974 1987
ACCORDING TO HEATH DE GRAAF 1992
OWN HIGH 2 3 4 LOW TOTAL HUSBAND HIGH
8 23 14 41 48 20 2 27 32 26 0 22 31
3 33 8 18 25 48 26
4 35 27 57 65 69 61 LOW 46 44 45 77 68 60
TOTAL 20 28 33 60 65 43
DUNCAN READS THIS TABLE FROM THE TWO MARGINALS
AND FINDS INDEPENDENT EFFECTS OF OWN CLASS AND
PARTNERS CLASS SOBEL BEGINS WITH MAIN DIAGONAL
PERCENTGAES AND ASKS WHETHER OFF DIAGONAL
PERCENTAGES ARE IN BETWEEN
33
DUNCANS MARGINAL REFERENCE MODELS VERSUS
SOBELS DIAGONAL REFERENCE MODELS SOBELS MOST
SIMPLE MODEL WHATEVER TWO DIAGONAL CELLS ARE
CHOSEN A PERSON ALWAYS WEIGHS OWN CLASS - OR
ORIGIN - WITH THE SAME WEIGHT AND CLASS PARTNER -
OR DESTINATION - WITH ANOTHER SOME WEIGHT WITH
BOTH WEIGHTS ADDING UP TO ONE
34
SOME OLDER ANSWERS TO SOBEL GENERATION QUESTIONS
ABOUT MOBILITY AND HETEROGAMY EFFECTS IN
BRITAIN , FOR FEMALE VOTING THE CLASS OF THE
HUSBAND CARRIES MORE WEIGHT THAN THE CLASS OF THE
WIFE IS FOR MALE VOTING WITH RESPECT TO FATHER -
SON MOBILITY FOR THE NETHERLANDS IT WAS FOUND
THAT FATHERS CLASS CARRIED MORE WEIGHT THAN OWN
CLASS
35
(No Transcript)
36
ONE VARIATION ON SOBEL GENERATION QUESTIONS
PEOPLE ARE QUICK TO ADAPT TO THE NORMS OF THEIR
DESTINATION GROUP - IF THEY MOVE UP BUT ARE THEY
LESS LIKELY TO ADMIT TO THEMSELVES THAT THEY
MOVED DOWN ? ASSIGN WEIGHTS FOR ABOVE MAIN
DIAGONAL CELLS WHICH ARE DIFFERENT FROM WEIGHTS
FOR BELOW MAIN DIAGONAL CELLS
37
ANOTHER VARIATION ON SOBEL GENERATION QUESTIONS
OVER - CONFORMITY CAN BE MODELLED BY INCLUDING
A SPECIAL PARAMETER FOR THE CELLS WHERE OVER -
CONFORMITY IS PREDICTED AFTER USA 1948 NO NEW
STRONG CASES OF OVER - CONFORMITY IN VORING HAVE
BEEN FOUND
38
(No Transcript)
39
A SOBEL GENERATION QUESTION IS THE DIVISION OF
HOUSEHOLD LABOUR LEAST UNEQUAL WHEN BOTH
PARTNERS HAVE THE HIGHEST LEVEL OF EDUCATION
, MOST UNEQUAL WHEN THEY BOTH HAVE THE LOWEST
LEVEL OF EDUCATION ? AND DO MIXED COUPLES FALL IN
BETWEEN , OR ARE THERE MACHO MAN EFFECTS
? TABLES FROM VAN BERKELS PH.D. WHO DOMINATES
WHEN ? USING THE FAMILY SURVEY DUTCH POPULATION
1993
40
FEMALE SHARE IN DOING THE DIRTY LAUNDRY AFTER
EDUCATION WOMAN AND EDUCATION PARTNER IN THE
NETHERLANDS IN 1993 MAN LOW MIDDLE
HIGH TOTAL WOMAN LOW 90 89
93 90 MIDDLE 88 87 87 87 HIGH 85 79
80 80 TOTAL 85 87 85 87
HIGHLY EDUCATED MEN OF LOWLY EDUCATED WOMEN DO
LESS THAN LOWLY EDUCATED MEN OF LOWLY EDUCATED
WOMEN WOMEN WITH HIGH EDUCATION BRING HIGHLY
EDUCATED MEN TO DOING THE DIRTY LAUNDRY
41
FEMALE SHARE IN CLEANING AFTER EDUCATION WOMAN
AND EDUCATION PARTNER IN THE NETHERLANDS IN
1993 MAN LOW MIDDLE HIGH TOTAL WOMAN
LOW 84 83 88 84 MIDDLE 79 79
80 79 HIGH 84 72 70 72 TOTAL 83 80
77 81
HIGHLY EDUCATED MEN OF LOWLY EDUCATED WOMEN DO
LESS THAN LOWLY EDUCATED MEN OF LOWLY EDUCATED
WOMEN WOMEN WITH HIGH EDUCATION BRING HIGHLY
EDUCATED MEN TO CLEANING
42
FEMALE SHARE IN COOKING AFTER EDUCATION WOMAN AND
EDUCATION PARTNER IN THE NETHERLANDS IN
1993 MAN LOW MIDDLE HIGH TOTAL WOMAN
LOW 74 73 73 74 MIDDLE 70 69
69 69 HIGH 70 66 63 64 TOTAL 73 71
67 71
WOMEN WITH HIGH EDUCATION BRING HIGHLY EDUCATED
MEN TO COOKING IN DIAGONAL REFERENCE MODELS THE
WEIGHT FOR WOMENS EDUCATION IS ONE THERE ARE
SPECIAL EFFECTS
43
NOTE THAT THE MEASURES FOR THE DIVISION OF
HOUSEHOLD LABOUR REFER TO A COMPOSITE SCORE IS
THIS FIRST GENERATION ERROR OK HERE ? ONLY IF THE
NUMBER OF TIMES AN ACTIVITY IS PERFORMED DOES NOT
DIFFER MUCH BETWEEN COUPLES !
44
(No Transcript)
45
ANOTHER SOBEL GENERATION QUESTION IS HAVING A
PAID JOB FOR WOMEN WITH A PARTNER MOST LIKELY
WHEN BOTH HAVE THE HIGHEST LEVEL OF EDUCATION ,
LEAST LIKELY WHEN BOTH HAVE THE LOWEST LEVEL OF
EDUCATION , AND IN BETWEEN FOR MIXED COUPLES ?
46
PERCENT OF WOMEN WITH A PAID JOB DUTCH LABOUR
FORCE SURVEY PARTNERS EDUCATION HIGH 2 3 L
OW TOTAL OWN HIGH 79 83 85 83 80 EDUCATION 2
63 72 67 64 68 3 44 49 47 41 47 LOW 28 37 35
27 32 TOTAL 68 63 55 41 60 FOR WOMEN WITH
THE HIGHEST EDUCATION , HAVING A PARTNER WITH THE
HIGHEST EDUCATION IS A DISINCENTIVE TO A PAID JOB
47
THE FINAL SOBEL GENERATION QUESTION WHEN
EXPLAING HIGH CULTURE ACTIVITIES , WHAT WEIGHS
HEAVIER ? RULE OF THUMB FOR THE NETHERLANDS
FROM DIAGONAL REFERENCE MODELS A PERSONS OWN
EDUCATION WEIGHS FOR ONE HALF PARTNERS EDUCATION
FOR ONE THIRD FATHERS EDUCATION FOR ONE
SIXTH THE 3 - 2 - 1 RULE
48
(No Transcript)
49
COMBINING THE FIRST SERIES OF MACRO QUESTIONS
WITH THE SECOND SERIES OF MICRO QUESTIONS INTO
AN NEW MACRO QUESTION IS IT ABSOLUTE MOBILITY
, SOCIAL FLUIDITY , OR THE COMPETITIVE BALANCE
THAT TIPS ELECTORAL OUTCOMES ? BACK TO SOMBART
WHY NO SOCIALIST PARTY IN THE USA ? BECAUSE OF
HIGH SOCIAL MOBILITY ! BUT WHAT KIND OF SOCIAL
MOBILITY ?!
50
IF ALL VOTERS WEIGH CURRENT AND PAST ECONOMIC
INTERESTS , THEN A MORE FAVOURABLE COMPETITIVE
BALANCE MAKES FOR A HIGHER PERCENTAGE OF RIGHT -
WING VOTES IF ALL VOTERS WEIGH CURRENT AND PAST
STATUS MOTIVES , THEN A HIGHER PERCENTAGE OF
ABSOLUTE MOBILITY - WHETHER UPWARD OR DOWNWARD -
MAKES FOR A HIGHER PERCENTAGE OF RIGHT - WING
VOTES
51
DESTINATION WEIGHT CLIMBERS AND SKIDDERS ONE
DESTINATION WEIGHT CLIMBERS ONE ORIGIN WEIGHT
SKIDDERS ONE
52
DESTINATION WEIGHT CLIMBERS AND SKIDDERS 0,8
DESTINATION WEIGHT CLIMBERS 0,8 ORIGIN WEIGHT
SKIDDERS 0,8
53
FROM ELECTORAL OUTCOMES TO OVERALL BOOK READING
AS THE MACRO PHENOMENON TO BE EXPLAINED A
COMPARISON OF EUROPEAN UNION COUNTRIES ? DOES
BOOK READING DECLINE? AND TO WHAT EXTEND DOES IT
DO SO BECAUSE OF MORE SOCIAL MOBILITY AND MORE
MIXED MARRIAGES?
54
SEE DE GRAAF ULTEE ELECTORAL STUDIES 1990 SEE
EQUALSOC WEBSITE FOR DIAGONAL REFERENCE MODELS
SEE SOBEL AMERICAL SOCIOLOGICAL REVIEW 1981 AND
1985