Implicit Discrimination in Hiring: Real World Evidence Dan-Olof ... - IZA
Implicit Discrimination in Hiring: Real World Evidence Dan-Olof ... - IZA
Implicit Discrimination in Hiring: Real World Evidence Dan-Olof ... - IZA
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
DISCUSSION PAPER SERIES<br />
<strong>IZA</strong> DP No. 2764<br />
<strong>Implicit</strong> <strong>Discrim<strong>in</strong>ation</strong> <strong>in</strong> Hir<strong>in</strong>g: <strong>Real</strong> <strong>World</strong> <strong>Evidence</strong><br />
<strong>Dan</strong>-<strong>Olof</strong> Rooth<br />
April 2007<br />
Forschungs<strong>in</strong>stitut<br />
zur Zukunft der Arbeit<br />
Institute for the Study<br />
of Labor
<strong>Implicit</strong> <strong>Discrim<strong>in</strong>ation</strong> <strong>in</strong> Hir<strong>in</strong>g:<br />
<strong>Real</strong> <strong>World</strong> <strong>Evidence</strong><br />
<strong>Dan</strong>-<strong>Olof</strong> Rooth<br />
Kalmar University,<br />
CReAM and <strong>IZA</strong><br />
Discussion Paper No. 2764<br />
April 2007<br />
<strong>IZA</strong><br />
P.O. Box 7240<br />
53072 Bonn<br />
Germany<br />
Phone: +49-228-3894-0<br />
Fax: +49-228-3894-180<br />
E-mail: iza@iza.org<br />
Any op<strong>in</strong>ions expressed here are those of the author(s) and not those of the <strong>in</strong>stitute. Research<br />
dissem<strong>in</strong>ated by <strong>IZA</strong> may <strong>in</strong>clude views on policy, but the <strong>in</strong>stitute itself takes no <strong>in</strong>stitutional policy<br />
positions.<br />
The Institute for the Study of Labor (<strong>IZA</strong>) <strong>in</strong> Bonn is a local and virtual <strong>in</strong>ternational research center<br />
and a place of communication between science, politics and bus<strong>in</strong>ess. <strong>IZA</strong> is an <strong>in</strong>dependent nonprofit<br />
company supported by Deutsche Post <strong>World</strong> Net. The center is associated with the University of Bonn<br />
and offers a stimulat<strong>in</strong>g research environment through its research networks, research support, and<br />
visitors and doctoral programs. <strong>IZA</strong> engages <strong>in</strong> (i) orig<strong>in</strong>al and <strong>in</strong>ternationally competitive research <strong>in</strong><br />
all fields of labor economics, (ii) development of policy concepts, and (iii) dissem<strong>in</strong>ation of research<br />
results and concepts to the <strong>in</strong>terested public.<br />
<strong>IZA</strong> Discussion Papers often represent prelim<strong>in</strong>ary work and are circulated to encourage discussion.<br />
Citation of such a paper should account for its provisional character. A revised version may be<br />
available directly from the author.
<strong>IZA</strong> Discussion Paper No. 2764<br />
April 2007<br />
ABSTRACT<br />
<strong>Implicit</strong> <strong>Discrim<strong>in</strong>ation</strong> <strong>in</strong> Hir<strong>in</strong>g: <strong>Real</strong> <strong>World</strong> <strong>Evidence</strong> *<br />
This is the first study provid<strong>in</strong>g evidence of a new form of discrim<strong>in</strong>ation, implicit<br />
discrim<strong>in</strong>ation, act<strong>in</strong>g <strong>in</strong> real economic life. In a two-stage field experiment we first measure<br />
the difference <strong>in</strong> callbacks for <strong>in</strong>terview for applicants with Arab/Muslim sound<strong>in</strong>g names<br />
compared to applicants with Swedish sound<strong>in</strong>g names us<strong>in</strong>g the correspondence test<strong>in</strong>g<br />
methodology. In the second stage of the experiment we measure, for a sample of the<br />
recruiters <strong>in</strong>volved, their explicit and implicit attitudes/performance stereotypes by the means<br />
of explicit questions and the implicit association test (IAT). We f<strong>in</strong>d (i) only weak correlations<br />
between explicit attitudes/performance stereotypes and implicit performance stereotypes but<br />
(ii) a strong and statistically significant negative correlation between the implicit performance<br />
stereotypes and the callback rate for an <strong>in</strong>terview for applicants with Arab/Muslim sound<strong>in</strong>g<br />
names, but not for applicants with Swedish sound<strong>in</strong>g names. These results <strong>in</strong>dicate that<br />
implicit discrim<strong>in</strong>ation acts differently compared to explicit discrim<strong>in</strong>ation and that it is an<br />
important determ<strong>in</strong>ant of the hir<strong>in</strong>g process.<br />
JEL Classification: J64, J71<br />
Keywords: implicit attitudes and stereotypes, discrim<strong>in</strong>ation, situation test<strong>in</strong>g,<br />
exit from unemployment<br />
Correspond<strong>in</strong>g author:<br />
<strong>Dan</strong>-<strong>Olof</strong> Rooth<br />
Kalmar University College<br />
391 82 Kalmar<br />
Sweden<br />
E-mail: <strong>Dan</strong>-<strong>Olof</strong>.Rooth@hik.se<br />
* I thank Jens Agerström, Per Johansson, <strong>Olof</strong> Åslund, participants at sem<strong>in</strong>ars <strong>in</strong> Kalmar and at IFAU<br />
(Uppsala) fpr valuable comments and helpful suggestions. Magnus Carlsson, Rickard Carlsson, Klara<br />
Johansson, and Terese Johansson provided excellent research assistance. A research grant from the<br />
Swedish Council for Work<strong>in</strong>g Life and Social Research and another research grant from the Institute<br />
for Labour Market Policy Evaluation (IFAU) are gratefully acknowledged.
*<br />
3 ) /<br />
3 8 * +<br />
/ 1 2<br />
3<br />
; <<br />
1 = + 3 >??(23<br />
* +<br />
+ @ A @ A 3 > @=<br />
B A<br />
; *<br />
* *<br />
+ 1 C >??D23 4<br />
; 3<br />
4 * 1 2<br />
3<br />
3 / $ 1>??D2<br />
* 3<br />
+ 3<br />
E<br />
F <<br />
) 5 C 1)GG'2 + 1>??)2<br />
+ 3<br />
> 5 " E C +1)GGG23<br />
+<br />
*<br />
*<br />
1
! 1>??(2<br />
<<br />
E 3<br />
1 >??>23 4<br />
E + E<br />
3<br />
5 " ??(2 ) DD> E<br />
5 +<br />
4 1 "42 )GG?<br />
I 1)GGJ23 K " , 1>??(2 "4 5<br />
;<br />
* " ??D2 1 2<br />
" <br />
+ * <<br />
K " ; "4 3<br />
* 3 6 E<br />
; * + 3<br />
' E<br />
3<br />
B<br />
2
3 5 K <<br />
' 3 5 D<br />
1 "4 2 *<br />
3 " * +<br />
3<br />
H * 3 4<br />
2. A model of implicit and explicit discrim<strong>in</strong>ation <strong>in</strong> hir<strong>in</strong>g<br />
"
ethnic group j, while is a measure of the degree of ethnic discrim<strong>in</strong>ation aga<strong>in</strong>st group<br />
j, expected to be zero aga<strong>in</strong>st native Swedes.<br />
In economics several forms of ethnic discrim<strong>in</strong>ation have been modelled (see the<br />
survey by Altonji and Blank, 1999). For our purposes we focus on two of those,<br />
preference based discrim<strong>in</strong>ation and statistical discrim<strong>in</strong>ation, both of which can be<br />
regarded as explicit forms of discrim<strong>in</strong>ation. 5 As mentioned <strong>in</strong> the <strong>in</strong>troduction the aim of<br />
this study is to test whether also implicit forms of discrim<strong>in</strong>ation are important <strong>in</strong> the<br />
hir<strong>in</strong>g situation. Hence, we can rewrite Equation 1 as:<br />
(2) - 1 = 1 2 = β + δ + δ + δ<br />
where the three discrim<strong>in</strong>ation terms express explicit preference discrim<strong>in</strong>ation, explicit<br />
statistical discrim<strong>in</strong>ation and implicit discrim<strong>in</strong>ation, respectively.<br />
One might wonder if it is possible to divide also the implicit discrim<strong>in</strong>ation measure<br />
<strong>in</strong>to preferences/attitudes and stereotypes as is done for the explicit discrim<strong>in</strong>ation<br />
measure. 6 Agerström et al (2007) implicit<br />
attitude score and the implicit stereotype score have a correlation of above 0.5 <strong>in</strong>dicat<strong>in</strong>g<br />
that they measure, if not exactly the same, but then at least highly related constructs. &<br />
3<br />
H<br />
4 * "<br />
D /<br />
# >??D / >??& E<br />
3 5 + 1)GGJ2<br />
3<br />
& 4 "4 "4 *<br />
3 "4 H +<br />
@ A @ A3<br />
"4 +<br />
3<br />
4
5 + I<br />
+ 3 4 + < +<br />
1 <<br />
2 +<br />
+ 3<br />
4 " # 5<br />
+ 1 7 >??? 8+ >??K23 5<br />
/ 5<br />
" ??(2 ) DD> E<br />
5 8 " 4 ") *3 ) ?K?<br />
D>><br />
3 C >KG<br />
5 >)(<br />
"
!" #<br />
4 "4 " >??& % >??(3<br />
+ "4 +<br />
3 7 1>??(2 ; "4<br />
1 M?3D23<br />
4 ; *<br />
* 3 4<br />
* E<br />
E 3 4 * +<br />
* 3 6<br />
H<br />
E<br />
E 3<br />
+ @ A<br />
< "
*<br />
3 (<br />
$% &<br />
4 )<br />
5 + "4<br />
3 =<br />
+ 3<br />
< + 3 6<br />
* 1 2<br />
E 3<br />
+ "4 1 F 2<br />
3 J + "4<br />
3<br />
8 3 "4<br />
3<br />
F " , 1>??(23 8<br />
1 2<br />
"> *3 4 +<br />
( / * @ A J(<br />
E D + "43 4<br />
J(
#<br />
A @ A +<br />
5 1 23<br />
4 + *<br />
H 3 / Feel<strong>in</strong>g thermometer they were asked to rate their positive or<br />
negative feel<strong>in</strong>gs on a ten-po<strong>in</strong>t scale (1 = very negative feel<strong>in</strong>gs, 10 = very positive<br />
feel<strong>in</strong>gs) toward Arab-Muslim m<strong>in</strong>ority men and native Swedish men, and then a<br />
difference between the two scales was calculated. This question is identical to the one<br />
used by Greenwald et al. (1998) and Nosek et al. (2005).<br />
Second, <strong>in</strong> the Hir<strong>in</strong>g preference rat<strong>in</strong>g participants had to choose which group they<br />
prefer when hir<strong>in</strong>g people. Rather than be<strong>in</strong>g directly related to the IAT, hir<strong>in</strong>g<br />
preferences are supposed to measure actual discrim<strong>in</strong>ation. The employers/recruiters had<br />
to choose one of the follow<strong>in</strong>g five alternatives: “When hir<strong>in</strong>g staff, I strongly prefer<br />
Arab-Muslim men (<strong>in</strong> Sweden) to native Swedish men”, “When hir<strong>in</strong>g staff, I moderately<br />
prefer Arab-Muslim men (<strong>in</strong> Sweden) to native Swedish men”, “When hir<strong>in</strong>g staff, I<br />
prefer Arab-Muslim men (<strong>in</strong> Sweden) and native Swedish men equally much”, “When<br />
hir<strong>in</strong>g staff, I moderately prefer native Swedish men to Arab-Muslim men (<strong>in</strong> Sweden)”,<br />
and “When hir<strong>in</strong>g staff, I strongly prefer native Swedish men to Arab-Muslim men (<strong>in</strong><br />
Sweden)”. The participants’ responses were coded from -2 to +2, with 0 as an<br />
<strong>in</strong>termediate po<strong>in</strong>t reflect<strong>in</strong>g no preference when hir<strong>in</strong>g staff.<br />
Third, <strong>in</strong> the Performance stereotype rat<strong>in</strong>g participants choose which of the two<br />
groups <strong>in</strong> question they consider to be more productive at work. The response alternatives<br />
were “Arab-Muslim men (<strong>in</strong> Sweden) are much more productive at work than native<br />
Swedish men”,” Arab-Muslim men (<strong>in</strong> Sweden) are slightly more productive at work<br />
than native Swedish men”, “Arab-Muslim men (<strong>in</strong> Sweden) and Swedish men are equally<br />
productive at work”, “Swedish men are slightly more productive at work than Arab-<br />
Muslim men”, “native Swedish men are much more productive at work than Arab-<br />
Muslim men”. Aga<strong>in</strong>, the participants’ responses were coded from -2 to +2, with 0 as an<br />
3<br />
@<br />
8
<strong>in</strong>termediate po<strong>in</strong>t reflect<strong>in</strong>g neutrality. 4 H<br />
5 3<br />
) @ A<br />
* 1D'O2<br />
1'DO2 E 5 " 3<br />
H * "<br />
' " !"<br />
* 3 5<br />
3<br />
4 ><br />
4 " 4 1 "42<br />
"
andom. The <strong>in</strong>tuitive idea is that it will be easier, and hence, go faster, to classify names<br />
and words that are compatible than those that are <strong>in</strong>compatible. The IAT measures every<br />
latency <strong>in</strong> response to the presented stimulus. For example, when two categories are<br />
“easily” associated <strong>in</strong> terms of their nom<strong>in</strong>al features (Arab names + low work<br />
productivity and Swedish names + high work productivity) the participant classifies the<br />
stimuli much faster and with fewer errors than when they are not associated. A total of<br />
sixty stimuli are presented for the compatible and compatible part, respectively. The<br />
difference <strong>in</strong> response latencies, or rather a recalculation of this difference called<br />
Greenwald’s D (as opposed to Cohen’s d), between the compatible and <strong>in</strong>compatible<br />
parts is known as the IAT effect or the IAT score.<br />
I 1>??K >??&2 "4<br />
1 / )23<br />
?3)D * ?3)D ?3KD ?3KD ?3&?<br />
"4 ?3&?<br />
" 3 )? 4 "4 )GK<br />
E<br />
@ F A @ A "
"4 * *<br />
@ A 3 4 *<br />
?3K<br />
" # "4<br />
4 K<br />
* 3 )><br />
4 * F * *<br />
3 3<br />
5 >3 4 < *<br />
+<br />
5 "
! 1>??(2 *<br />
+ >?<br />
* 3 4 @ A<br />
+<br />
3 )K # C 8<br />
3 4 3<br />
# C +<br />
3<br />
# !<br />
# 8 8H ><br />
23 4<br />
> * <<br />
3 4<br />
?3?K # ! 83<br />
F 3 /<br />
* 1 # !<br />
< 1<br />
< 2 K<br />
" ??D2 ?3KD ?3& < 3<br />
)D *<br />
3<br />
12
3 6<br />
& )?<br />
3 4<br />
3 )& H<br />
3<br />
@ A < * +<br />
1 2<br />
5 "
<strong>in</strong>terpreted with the “cont<strong>in</strong>uous” explicit and implicit discrim<strong>in</strong>ation measures.<br />
Therefore we have constructed b<strong>in</strong>ary explicit and implicit “equivalents”.<br />
4 4 '"<br />
< * 4 'C3 4<br />
"4 ?3' F<br />
3 4 * <<br />
"
"
+<br />
+<br />
1# C !2<br />
5 3 "<br />
* 3 6<br />
1# 2<br />
3 *<br />
* H +<br />
3 *<br />
>(& 3<br />
/ #<br />
3 4 *<br />
4 D<br />
S<strong>in</strong>ce the explicit discrim<strong>in</strong>ation measures are collected by <strong>in</strong>terview<strong>in</strong>g recruiters it is<br />
likely that some are reluctant to reveal their true attitudes, for <strong>in</strong>stance, for political<br />
correctness reasons. Such measurement errors will bias the estimates of the explicit<br />
discrim<strong>in</strong>ation measures. Also, the empirical analysis reveals that the estimates for the<br />
explicit measures are unstable across specifications, <strong>in</strong>dicat<strong>in</strong>g measurement problems. A<br />
well known strategy is then to f<strong>in</strong>d an <strong>in</strong>strumental variable and estimate IV 2SLS, which<br />
helps <strong>in</strong> delet<strong>in</strong>g the measurement error. This is achieved by hav<strong>in</strong>g a second measure of<br />
explicit preferences that is not perfectly correlated with the first one. Such an<br />
<strong>in</strong>strumental variable exists for the explicit attitude as well as for the explicit stereotype<br />
measure. In the case of the explicit attitude measures there are two questions be<strong>in</strong>g asked.<br />
Hence, the feel<strong>in</strong>g thermometer will be used as an <strong>in</strong>strument for the hir<strong>in</strong>g preference.<br />
For the explicit stereotype measure the IAT measure will be used as an <strong>in</strong>strument.<br />
6 Q >575<br />
* 4 &3 I<br />
4 ' D * 3<br />
3<br />
*<br />
16
(<br />
4 + *<br />
3 6<br />
"4 <<br />
"
strategy as ours are needed before fully accept<strong>in</strong>g implicit discrim<strong>in</strong>ation as a new form<br />
of labor market discrim<strong>in</strong>ation.<br />
18
)<br />
" , %3 ! 3 3 1>??(2 @8 B 8<br />
" 6 +- 5 5 A 3<br />
" E %3 C + 3 1)GGG2 I 7 # + 3<br />
" ! 1 2 3 8 3 " 3<br />
K)'KBK>DG3<br />
. 3 3 1>??D2 ”Shifts <strong>in</strong> Attitudes and Labor Market <strong>Discrim<strong>in</strong>ation</strong>:<br />
Swedish Experiences after 9-11”, Journal of Population Economics 18 (4), pp. 602-629.<br />
C # 3 ! 3 # 53 1>??D2 @= "<br />
9 A " 8 GD 1>2 3 G'<br />
GJ3<br />
! # 3 3 1>??&2 @8 8<br />
5 7 # + 0 8* A $" -R>>J) $"3<br />
! # 3 3 1>??(2 @8 8<br />
5 7 # + 0 8* A 3<br />
I "3 # I 3 5 F %3 1)GGJ23 #<br />
9 4 " 4 3<br />
! )'&'B)'J?3<br />
I "3 = + C3 C E # 3 1>??K2 @0 0<br />
" 4 9 3 " 5 " A<br />
JD 3 )G( >)&<br />
I "3 = + C3 5 =3 1>??&2 @! H<br />
" 4 9 ! C % A3<br />
0 3<br />
3 C %3 1)GG'2 @C 7 # + A "<br />
# $ % J' 1D29))(' ))G'3<br />
+ %3 1)GGJ2 @ A $ &<br />
)> 3 )?) )&3<br />
+ %3 S3 1>??)2 @ = 5+ 9 7<br />
I8 4 - A " # $ % 3 3<br />
7 3 C E # 3 = + C3 I " 1>??(2 @0<br />
9 Q3 6 + 15 / 2 # A C3<br />
6 + 5 F =3 18 2 ' " (<br />
$ 3 DG )?>3 = S +9 I - 3<br />
19
7 "3 1>???2 ) & 4<br />
C 1 + 2 = +, 3<br />
Nosek, B. (2005) ”Moderators of the relationship between implicit and explicit<br />
evaluation”, Journal of Experimental Psychology, vol 134, pp. 565-584.<br />
Nosek, B., Greenwald, A. and Banaji, M. (2007) ”The <strong>Implicit</strong> Association test at Age 7:<br />
A Methodological and Conceptual Review”, %3 C 18 32 "<br />
$ 3 >&D >G>3 - - 3<br />
-3 "3 %3 1>??>2 @/ * +<br />
A * ))> 3 /'J? /D)J3<br />
3 8+ %3 1>??K2 T0<br />
5 3 8 + T<br />
)& 1'2 3 (J(BJ)'3<br />
$ %3 -3 1>??D2 @8 9 4<br />
" # ! C A % "<br />
- G?1K2 3 DDK D&>3<br />
20
0<br />
/ )3 4 "4 1 2 "<br />
5 3<br />
40<br />
30<br />
20<br />
10<br />
0<br />
-0,50 0,00 0,50 1,00<br />
IAT Effect (D measure)<br />
Mean = 0,3841<br />
Std. Dev. = 0,34083<br />
N = 193<br />
/ >3 = "4 3 / )GK 3<br />
0,0<br />
-0,1<br />
-0,2<br />
-0,3<br />
-0,4<br />
.00 .50 1.00 1.50 2.00 2.50 3.00 3.50 4.00<br />
Standardised IAT effect<br />
L<strong>in</strong>ear<br />
Cubic<br />
+ 9 4 F "4<br />
3 " F "4 > H<br />
1 "4 M?3&23 4 + 4 '"<br />
?3?>& ?3?K) 3 4<br />
H 3<br />
21
K3 = "4 3 ! )K& 3<br />
0,00<br />
-0,20<br />
-0,40<br />
-0,60<br />
-0,80<br />
-1,00<br />
.00 .50 1.00 1.50 2.00 2.50 3.00 3.50 4.00<br />
Standardised IAT effect<br />
L<strong>in</strong>ear<br />
Cubic<br />
+ 9 4 F "4<br />
3 " F "4 > H<br />
1 "4 M?3&23 4 + 4 '"<br />
?3??) ?3?D) 3 4<br />
H 3<br />
22
" -<br />
4 )3 - * 3<br />
& !"<br />
1<br />
-<br />
15 2<br />
1 O2<br />
5 >?' )J' D& 1K?2 K3& 1>3?2<br />
" DG DG )J 1K)2 >3G 1)3K2<br />
C >>) )G> &> 1K>2 K3> 1)3D2<br />
= ) ?)' >G' D( 1)G2 )(3K 1)(3'2<br />
" ) 'GJ (>G )GK 1>&2<br />
+ / * < 1) 'GJ2 5 5 ; 3<br />
4 1 23 5 1K?O2<br />
@ A + "4 3 4<br />
3 / @ 5 A<br />
@ A3 3<br />
4 >3 8* 3 )GK 3<br />
! "<br />
"<br />
! "<br />
!<br />
+ 9 4 4 ) 3<br />
23
4 K3 ! * " # "4<br />
* 3 )GK 3<br />
# "<br />
$ % & ' ()* + ", + , + , + , + + ,<br />
$ % - ! . + ", + , + , + , + +<br />
$ % - ! . $ / 0 % + , + , + , + , + +<br />
$ % 1 ! . / ! + , + , + , + , + ", + ,<br />
$"% 1 ! . / ! $ / 0 % + , + , + , + , + , + ,<br />
$ % 2 / ! ' + + + + ", + , + ,<br />
$ % 2 / ! ' $ / 0 % + , + + + , + , + ,<br />
Note: * p
Table 4A. The correlation between the callback rate for <strong>in</strong>terview and the implicit and explicit attitude and stereotype measures.<br />
Percentage po<strong>in</strong>ts. Full data of 193 observations.<br />
3 . ! // ! # 4 5 ! ! . ! ) # !<br />
& ! 7 7 7 +<br />
$ + %<br />
3 7<br />
) 8 6<br />
+ ,<br />
$ + %<br />
- 6 ! 7 !<br />
+<br />
$ + %<br />
+ ,<br />
$ + %<br />
+<br />
$ + %<br />
- ! . + ,<br />
$ + %<br />
1 ! . / ! +<br />
$ + %<br />
2 / ! ' + " +<br />
$ + "% $ + %<br />
3 + ,,,<br />
$ + %<br />
). / +<br />
$ + %<br />
9 # / ' 5 4 0 + "<br />
$ + %<br />
+ ",,,<br />
$ + " %<br />
+ ",,<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,,<br />
$ + " %<br />
+ ,,<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,,<br />
$ + " %<br />
+ ",,<br />
$ + %<br />
+ "<br />
$ + %<br />
3 7<br />
) 8 6<br />
+ ,,<br />
$ + " %<br />
+ ,<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,<br />
$ + " %<br />
+ ,,,<br />
$ + %<br />
+<br />
$ + %<br />
+ ""<br />
$ + %<br />
+ ,,<br />
$ + " %<br />
+<br />
$ + %<br />
+ "<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + " %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + "%<br />
: ! / 7 // 9 ; ; ; ; 9 ; ; ; ;<br />
< . 7 ; 9 ; ; ; ; 9 ; ; ;<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 8<br />
+ 3 " * 3 4<br />
54"4" G3 4 4 ) * # C3 ) D<br />
)GK & )? )K&<br />
3 4 ?3?J ?3D' 3<br />
25
4 'C3 4 + *<br />
3 - 3 / )GK 3<br />
3 . ! // ! # 4 5 ! ! . ! ) # !<br />
$ / + 0 % +<br />
$ + " %<br />
3 7<br />
) 8 6<br />
+<br />
$ + %<br />
- -<br />
+ ,<br />
$ + %<br />
+ ",<br />
$ + %<br />
+ "<br />
$ + %<br />
- ! . $ / 0 % + ",<br />
$ + %<br />
1 ! . / ! $ / 0 % +<br />
$ + %<br />
2 / ! ' $ / 0 % + +<br />
$ + "% $ + %<br />
3 + ,,,<br />
$ + %<br />
). / +<br />
$ + %<br />
9 # / ' 5 4 0 +<br />
$ + "%<br />
+ ,,,<br />
$ + " %<br />
+ ,,<br />
$ + %<br />
+ ,<br />
$ + %<br />
+ ,,,<br />
$ + " %<br />
+ ,,<br />
$ + %<br />
+ ,<br />
$ + %<br />
+ ,,,<br />
$ + " %<br />
+ ,,<br />
$ + %<br />
+ ,<br />
$ + %<br />
3 7<br />
) 8 6<br />
+ ,,<br />
$ + %<br />
+<br />
$ + %<br />
+ ",,,<br />
$ + %<br />
+<br />
$ + %<br />
+ "<br />
$ + %<br />
+ ,<br />
$ + "%<br />
+ ,,,<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+ ,<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
: ! / 7 // 9 ; ; ; ; 9 ; ; ; ;<br />
< . 7 ; 9 ; ; ; ; 9 ; ; ;<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 8<br />
+ 3 " * 3 4<br />
54"4" G3 4 4 ) 3 4<br />
?3?J ?3D' 3<br />
+ ",<br />
$ + %<br />
+ "<br />
$ + %<br />
+ "<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
26
Table 4C. Interaction effects. Percentage po<strong>in</strong>ts.<br />
! "<br />
$ / + 0 % + ,<br />
$ + %<br />
3 . ! // ! # 4 5 ! ! . &5 7 !<br />
- 6 ! 7 !<br />
3 7<br />
3 7<br />
6 - = 1 6 - = 1<br />
+ ,<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
- ! . $ / 0 % +<br />
$ + %<br />
1 ! . / ! $ / 0 % +<br />
$ + %<br />
2 / ! ' $ / 0 % +<br />
$ + %<br />
# + + +<br />
$ + % $ + % $ + %<br />
+ ,<br />
$ + "%<br />
+ ,,<br />
$ + %<br />
+ ,<br />
$ + %<br />
+<br />
$ + %<br />
+ ,,<br />
$ + " %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+ ,<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
: ! / 7 // ; ; ; ; ; ; ; ;<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 8<br />
+ @ A 3 " * 3 4<br />
54"4" G3 4 4 ) 3 ) ' )GK<br />
D J )K& 3 4<br />
?3>' ?3JJ 3<br />
27
4 ' 3 4 + *<br />
3 = 5 3 - 3<br />
3 . ! // ! # 4 5 ! ! . &5 7 !<br />
- 6 ! 7 !<br />
3 7 3 7<br />
$ "<br />
& ! 7 7 7 +<br />
$ + %<br />
+<br />
$ + %<br />
- ! . + " ,,<br />
$ + %<br />
1 ! . / ! +<br />
$ + %<br />
2 / ! ' + +<br />
$ + "% $ + %<br />
! "<br />
$ / + 0 % + "<br />
$ + %<br />
+<br />
$ + "%<br />
- ! . $ / 0 % + ,,<br />
$ + %<br />
1 ! . / ! $ / 0 % +<br />
$ + %<br />
2 / ! ' $ / 0 % + + "<br />
$ + % $ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + "%<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+ "<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
: ! / 7 // ; ; ; ; ; ; ; ;<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 8<br />
+ @ A *<br />
+ 3 " * 3 4 54"4" G3<br />
4 4 ) 3 ) ' )GK D J<br />
)K& 3 4<br />
?3?J ?3D' 3<br />
28
4 D3 + * *<br />
N- 3<br />
) # >3 ! + ,,<br />
$ + " %<br />
3 7<br />
) 8 6<br />
+ "<br />
$ + %<br />
- 6 ! 7 !<br />
+<br />
$ + %<br />
+<br />
$ + %<br />
3 7<br />
) 8 6<br />
+ ,,,<br />
$ + " %<br />
+ ,<br />
$ + %<br />
+ ,,,<br />
$ + %<br />
( ? ) # ,( 9 ; 9 ; 9 ; 9 ;<br />
? ) # , 9 9 ; ; 9 9 ; ;<br />
@ ! 7 / ! 7 // ! !<br />
# 4 / ! 5 # ' ><br />
/ 7 ! !<br />
" " "<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 4 "
4 &3 # * 3 Q >5753 - 3<br />
3 . ! // ! # 4 5 ! ! . ! ) # !<br />
3 7<br />
) 8 6<br />
- -<br />
3 7<br />
) 8 6<br />
7 + +<br />
+ +<br />
$ + % $ + %<br />
$ + % $ + %<br />
' + " +<br />
+ "<br />
$ + % $ + %<br />
$ + %<br />
: ! / 7 // ; ; ; ; ; ; ; ;<br />
( . ! 9 ; 9 ; 9 ; 9 ;<br />
! / . . ! +" + + +<br />
9 /<br />
+ 9 1 2 1 2 1 2 3 4 Q # C<br />
# " ! 54"4" G3 4 * 1<br />
2 4 'C 3 4 4 ) *<br />
+ * 3 3<br />
30
Appendix<br />
4 ")3 "<br />
A #<br />
9 +<br />
9<br />
(! 7<br />
9 +<br />
) ! ! 7<br />
9 +<br />
0<br />
* !<br />
9 +<br />
: ! ' &5 7<br />
! 7 ! . !<br />
! 7<br />
9 +<br />
: ! ' ) #<br />
! 7 ! . !<br />
! 7<br />
9 +<br />
&5 7<br />
! 7 ! .<br />
!<br />
$ % $ %<br />
$ %<br />
$ %<br />
$"%<br />
$ % $ ? "%>$ % $ ? %>$ % $&5<br />
B<br />
7 %>$) # %<br />
2<br />
χ<br />
6 / ! " + + + +<br />
* $ ! 7 ! % +" + " + +<br />
8 ! ! " + " + + + ,,,<br />
2 * + + + + ,,,<br />
) ! ! "" + + + +<br />
9 " " "" " + + + + ,,,<br />
* ! 7 ' + + + " + ,,<br />
* $ ! . . % + + " + + ,,,<br />
6 ! ! 5 4 + + + + ,,,<br />
B ! 5 4 + + + " "+ ,,<br />
3 7 " + + + + ,,,<br />
& ! " + + + + ,,,<br />
* C"" C " + + +" + ,,,<br />
9 ! 1>??(23<br />
+ 9 4 @C H A 1D2 U 1&23 4 2<br />
χ<br />
&3&K 1 2 K3J' 1 23 1 2 0 3<br />
6 # 4<br />
) #<br />
! 7 ! .<br />
!<br />
31
4 ">3 N 1?