"Fair" Inequality ? An International Comparison of Attitudes to Pay Differentials

Lars Osberg Dalhousie University

Tim SmeedingSyracuse University

July 2, 2005 Society for the Advancement of Socio-EconomicsBudapest

Inequality & Public Policy

Greater inequality in market incomes implies more people would benefit from redistributionDemocracy + self-interest of Median voter implies

• higher taxes & more redistribution when market incomes are more unequal ??

USA is important counter exampleMore inequality & poverty than in other affluent

OECD nations • Governments do less about it

Why is US policy different ?

Policy outcomes depend on public preferences + institutional structure

H0: American “preferences” different Greater emphasis on = opportunity ??More acceptance of mobility/change ??

• economics literature emphasizes POUM (Prospect Of Upward Mobility)

• Alesina, di Tella and MacCulloch (2001), Alesina and la Ferrara (2001), Alesina and Angeletos (2003), Benabou and Ok (1998) and Piketty (1995)

But are US attitudes different?

“Preferences” for redistribution mingle: Perceptions of actual outcomes Perceptions of possibility & costs of change Values about desirable outcomes Values about legitimate process & agents of change

HA: USA not particularly different in values Large & subtle sociological & survey literature - ignored in

economics literature • Examples: Kelly and Evans (1993), Kluegel et al (1995) Svallfors

(1997) Suhrcke (2001).

USA not an outlier in international survey data except perhaps in average attitudes to government as agent of

change ?

Are attitudes different ? Plan:

So why not just ask people ? Average answers not so different

• Distribution of attitudes may be key for political economy What do the answers mean ?

Many issues mingled in “inequality” or “redistribution” “Should earn” / “Do earn” data in ISSP

Preferences for aggregate inequality Leveling up or down ? Max/mean & Mean/min

Distribution of preferences for leveling Determinants ??

• USA is different in % religious fundamentalism Conclusion

& the bottom line will be ……??

Hard to find big cross-national differences in attitudes – on average General preference for “less inequality” even given

underestimation of extent of actual inequality Polarization in US attitudes is different from

elsewhere – especially Europe Similar preferences across nations for “leveling down”

at top of distribution + massive under-estimation of actual top end inequality

Anglo countries have less concern for social minima

Are Income Differences “Too Large”?

Most people, in all countries agree or agree strongly

Tiny minority everywhere “disagree” USA is not particularly “different” USA also not an outlier when asked to

agree/disagree: “Inequality continues to exist because it benefits the rich and powerful” “Large income differences are necessary for a country’s prosperity” Importance of “Knowing the right people”, “Well-educated parents” ,

Education/Ambition/Ability/Hard Work

Country YearStrongly Agree Agree

Neither Agree Nor Disagree Disagree

Strongly Disagree Total

Australia 1999 17.8 53.1 17.1 11.6 0.4 100

Austria 1999 40.4 45.8 9.1 4.7 0 100

Canada 1999 28.1 42.5 15.7 11.2 2.6 100

France 1999 60.3 27.2 7.4 4.5 0.7 100

Germany 1999 20.5 55.2 14.3 9.1 0.9 100

Italy 1992 53.2 36.3 6.3 4 0.2 100

Netherlands 1987 19 47.4 13 17.6 3 100

Norway 1999 22.4 50.1 13.8 12 1.8 100

Spain 1999 35.9 53.4 7.4 3.1 0.2 100

Sweden 1999 29.2 41.9 18.1 8.4 2.4 100

Switzerland 1987 19.1 48.5 20.7 10.4 1.4 100

UK 1999 31.7 50.6 11.6 5.4 0.6 100

US 1999 25 41.2 21.5 9.2 3.2 100

Source: ISSP 1999, 1992, 1987

Attitudes to Inequality: Are Income Differences Too Large?

What do such questions mean ?

To fix ideas, suppose that we lived in a just society, then: “should get” income = “do get” income  (1)                Yi * = Yi

A

 (2)           YiA > Y*min Y*min 0

(3) YiA < Y*max Y*max

If society is not just, then some may get “too much” while others get “too little”

Yi * = b0 + b1 YiA

If an individual believes society is fair, then b0 = 0 b1 = 1 Y*max > Yi

A > Y*min

A

“Do Earn”

YA

“Should Earn”

Y*

Y* = YA

YJ

Y*min

Y2Y1

e

ad

cY*max

Figure IEquity in Earnings

3 dimensions of Pay Norms

(1) the ethical floor to minimum earnings (Y*min );

(2) the ethical ceiling to maximum earnings (Y*max);

(3) the desired degree of levelling, relative to the current income distribution, among “acceptable” incomes (b1 ).

• Note: Jasso ratio = ln (Yi A / Yi *)

• implicitly sets b0= 0

Inequality – an issue of “comparing two

frequency distributions f(y)” ?? Atkinson JET 1970

“Inequality” has two meanings Differences between individuals/groups in rewards

• Relative income ratio is only necessary information Dispersion of rewards in a population

• Estimate of population densities needed• Gini/Theil/Atkinson – all need to know f(yi)

Subjective perceptions of income frequencies based on biased, self-selected small samples Population densities estimated with great error

• Most respondents place selves in “middle” regardless of actual level of income

Are ethical attitudes driven by $ differentials or population frequencies?• E.g. Black/white, male/female, skilled/unskilled differentials • – how relevant is the number of each to equity norms ?

“Do Earn” & “Should Earn” Inequality

ISSP 1999, 1992, 1987 what salaries do people actually make ? what salaries should they make ?

• skilled factory worker, doctor in general practice, chairman of a large national company, lawyer, shop assistant, owner/manager of a large factory, judge in the country’s highest court, unskilled worker & federal cabinet minister.

• + Bus driver, secretary, brick layer, bank clerk in 1987

“should earn” is conditional on “do earn” For political economy, subjective reality is key “Should earn” implicitly controls for estimation errors,

process, family “need”, hours work, etc.

Preferences for Aggregate Inequality

For each person - calculate subjective inequality index (CV,Gini, Theil)Actual Inequality (CVA, GiniA)“Ethical Inequality” (CVE, GiniE)Ratio – “tension” ?

Is “Ethical inequality” = 0 ? (nowhere) What is tension between actual & fair

inequality ? (constant)

Actual & Ethical Inequality

All nation averages - GiniA~.46; GiniE ~.34Trend & Level of actual earnings inequality

not reflected in subjective estimates of actual inequality

All countries accept some earnings inequality

USA not particularly different from others “Should Earn” inequality is less than “Do

Earn” inequality in all countries (Ratio ~ .75)

Two Distributional Problems

A summary statistic (Gini, Theil, CV) cannot reveal where the concern with inequality lies max, min or leveling in between ?

The Distribution of attitudes & estimates may matter crucially for political economy

What is the socially acceptable range of incomes? How much inequality is

perceived ?

ISSP queries span the range very top (chairman of a large national company) very bottom (unskilled worker)

Do nations differ in acceptable range ? Between middle & top ?

• Not very much - & desired range is approx 2:1 Between middle & most disadvantaged ?

• Significantly – Anglo countries especially Actual extent of salary differences is massively

underestimated

CEO Compensation & Pay of Production Workers in Manufacturing

CEO Compensation

Production Worker in Manufacturing Ratio

Rank by Ratio

UK (1) 762,305 22,654 34 4Australia (3) 740,565 19,582 38 2Japan(1) 524,454 29,974 17 8France(3) 619,047 16,699 37 3Sweden(3) 504,468 21,192 24 5Germany(1) 491,561 26,465 19 7US(1) 1,404,309 29,391 48 1Canada(2) 481,651 23,436 21 6Note: 2001 (US $)

Americans' Subjective Perception of Objective Reality: Do Earn CEO/Do Earn Skilled Worker

00.020.040.060.080.1

0.120.140.16

1 2.5 4 5.5 7 8.5 10 11.5 13 14.5 16

1987

1992

1999

Canadians' Subjective Perception of Objective Reality: Do Earn CEO/Do Earn Skilled Worker

0

0.05

0.1

0.15

0.2

0.25

0.3

1 2.5 4 5.5 7 8.5 10 11.5 13 14.5 16

1992

1999

West Germans' Subjective Perception of Objective Reality: Do Earn CEO/Do Earn Skilled Worker

00.020.040.060.080.1

0.120.140.16

1 6 11 16

1987

1992

1999

Preferences for Levelling Across Country-Year Pairs

Should Earn Ratio of Chairman to Skilled Workers

Country-Year Median Mean Median Mean Median Mean

USA 1999 2.86 4.38 14 16 0.67 0.52 5.43

Canada 1999 4.00 6.41 8 5 1.25 2.27 7.02Australia 1999 4.00 5.41 8 11 1.00 1.38 3.67New Zealand 1999 4.75 6.31 6 6 0.79 0.57 5.58UK 1999 5.33 7.20 3 3 1.00 1.30 6.18North Ireland 1999 4.38 6.20 7 7 0.31 0.37 7.33Spain 1999 2.50 2.77 16 27 0.00 0.02 1.32Portugal 1999 5.00 6.80 4 4 0.45 1.93 8.55France 1999 6.25 8.71 1 1 2.14 3.21 8.98Netherlands 1987 3.24 4.27 11 17 0.40 0.31 3.67Switzerland 1987 2.14 3.21 19 23 0.22 0.11 5.50Germany (W) 1999 4.80 6.01 5 8 0.93 0.45 4.61Austria 1999 4.00 5.47 8 10 0.67 0.81 4.54Norway 1999 1.72 1.84 20 30 -0.05 -0.12 0.93Sweden 1999 2.14 3.05 19 24 0.50 0.77 3.01Israel 1999 4.00 4.79 8 13 0.25 0.57 3.32Japan 1999 6.00 8.30 2 2 0.29 ######## 7.04

Ratio RankMale Ratio - Female

RatioStd Deviation of Ratio: Males &

Females

Distributions of Should-Earn Ratios Across Countries: 1999Table I: Means, Medians and Rankings: All Individuals

Country

Mean Med Mean Med Mean Med

United States 9.680 6.667 3.013 5 2 2.131 1.978 0.153 6 7 4.037 3.236 0.801 5 4

Canada 9.792 6.250 3.542 4 2 2.144 1.931 0.213 5 6 3.961 3.064 0.897 3 3

Australia 6.110 5.000 1.110 12 9 1.935 1.791 0.144 12 12 2.975 2.686 0.289 12 12

New Zealand 7.982 5.555 2.427 8 6 2.024 1.875 0.149 9 10 3.515 2.857 0.658 8 10

United Kingdom 10.932 6.667 4.265 3 2 2.305 2.100 0.205 2 2 4.038 3.255 0.783 4 2

North Ireland 7.994 5.442 2.552 6 5 2.018 1.858 0.160 10 11 3.507 2.943 0.564 6 6

Spain 3.139 2.800 0.339 15 11 1.607 1.556 0.051 15 15 1.773 1.875 -0.102 15 13

Portugal 7.804 5.333 2.471 10 7 2.007 1.902 0.105 11 8 3.499 2.813 0.686 9 11

France 11.092 7.143 3.949 2 1 2.319 2.132 0.187 1 1 4.210 3.296 0.914 2 1

Germany 7.553 6.000 1.553 11 4 2.112 2.000 0.112 7 5 3.306 2.880 0.426 10 7

Austria 7.779 5.333 2.445 7 7 2.030 1.883 0.147 8 9 3.461 2.866 0.595 7 9

Norw ay 3.206 2.609 0.597 14 12 1.610 1.564 0.046 14 14 1.906 1.667 0.239 14 15

Sw eden 4.018 2.941 1.077 13 10 1.718 1.591 0.127 13 13 2.079 1.791 0.288 13 14

Isreal 7.750 6.000 1.750 9 4 2.212 2.083 0.129 4 3 3.277 2.879 0.398 11 8

Japan 12.347 6.515 5.832 1 3 2.258 2.027 0.231 3 4 4.488 3.053 1.435 1 5

Difference: Column 9 - Column 10

Country Rank by Mean &

Med MeanMin Ratio

Mean MeanMin

Ratio

Median MeanMin

Ratio

Mean MaxMin Ratio

Median MaxMin Ratio

Mean MaxMean

Ratio

Median MaxMean

Ratio

Difference: Column 2 - Column 3

Country Rank by Mean &

Median MaxMin Ratio

Difference: Column 5 - Column 6

Country Rank by Mean &

Med MaxMean Ratio

In all countries, survey respondents estimate there to be far smaller income differentials than actual data indicates

Across countries, little variation in average respondents’ acceptable “Top End “ inequality Max/Mean “Should Earn” ratio is approximately 2 USA in middle of pack

US, UK Canada show larger acceptable bottom end inequality Mean/Min “Should Earn” approx 3.2

Distribution of Preferences for leveling: b1 in Yi * = b0 + b1 Yi

A

Figure 5.1United States Social Inequalities ISSP Years 1987-1999: Should Earn / Do Earn

Slope Coefficient (Beta) Over Time, Both Sexes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

U.S. 1987

U.S. 1992

U.S. 1999

Table 5.2United States Social Inequalities ISSP Years 1987-1999:

MaxMean Ratio Over Time, Both Sexes

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

U.S. 1987

U.S. 1992

U.S. 1999

Figure 5.3United States Social Inequalities ISSP Years 1987-1999:

MeanMin Ratio Over Time, Both Sexes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

U.S. 1987

U.S. 1992

U.S. 1999

Figur e 5 . 8

U n it ed S t at es , U n it ed K in gdom, C an ada & Nor way 19 9 9 : D ist r ibut ion of Do- E ar n / S hould

E ar n S lope C oeffi cien t ( B et as) : B ot h S exes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

B et a

U.S.

U.K.

Canada

Norway

Figur e 5 . 9

U n it ed S t at es , U n it ed K in gdom, C an ada & Nor way 1 9 9 9 : D is t r ibut ion of M ax S hould E ar n

/ M ean S hould E ar n ( M axM ean ) R at io: B ot h S exes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

B et a

US MaxMean

UK MaxMean

Cdn MaxMean

Nor MaxMean

Figur e 5 . 9

U n it ed S t at es , U n it ed K in gdom, C an ada & Nor way 19 9 9 : D ist r ibut ion of M ax S hould E ar n

/ M ean S hould E ar n ( M axM ean ) R at io: B ot h S exes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

B et a

US MaxMean

UK MaxMean

Cdn MaxMean

Nor MaxMean

Figur e 5 . 10

U n it ed S t at es , U n it ed K in gdom, C an ada & Nor way 19 9 9 : D ist r ibut ion of M ean S hould E ar n

/ M in S hould E ar n ( M ean M in ) R at io: B ot h S exes

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

B et a

US MeanMin

UK MeanMin

Cdn MeanMin

Nor MeanMin

Implications & Explanations ??

Current actual trends are for widening inequality – particularly at top end But is extent & trend of inequality perceived ? USA – leveling preferences are bimodal

• Levelers & status quo defenders split Hardening of American attitudes against large

differentials at top Erosion of consensus on minimum standards Polarization of attitudes - not a recipe for political

stability “Bi-causal” perceptions are highly susceptible to

“framing” and salience – possibly unstable ?