21
THE SOCIAL VALUATION OF INCOME: A SURVEY APPROACH IN
TURKEY
Yoko Nagase*, David Evans** and Erhun Kula***
*Oxford Brookes University ()
**Oxford Brookes University ()
***Bahcesehir University, Istanbul ()
ABSTRACT
A model relating equity welfare weights to income is developed in order to estimate the elasticity of marginal social valuation of income and test for iso-elasticity over a wide range of income using survey data obtained from a sample of leading Turkish politicians, who are instrumental in actual policy making. While formal statistical testing suggests that a constant unitary value for this welfare parameter cannot be rejected, the calculation of the elasticity measure, based on the point estimates of the regression parameters, reveals considerable variation in value over the full income range. A large-scale survey, conducted along the same lines, may help to resolve the iso-elasticity issue.
JEL CODES: D60, D61, H24.
KEYWORDS: Social cost-benefit analysis, welfare weights, elasticity of marginal social valuation of income, Turkish politicians.
I INTRODUCTION
The question of how to weight benefits and costs impacting on individuals or clearly identifiable socio-economic groups such as households, regions and generations that are differentiated by income is a crucial issue in social cost-benefit analysis (CBA). The importance of welfare weights in both intra and inter-temporal contexts is made clear in the official CBA guidance issued by governments and supra-national bodies; see, for example, HM Treasury (2003) for the UK and the European Commission (2008) for the EU. In an inter-temporal context, both the choice of discount rate and the extent to which it may decline over time are vital issues in relation to the weighting of project impacts on future generations, especially in relation to climate change mitigation policy; see Gollier (2002), Gollier et al (2008), Gollier and Weitzman (2010), Nordhaus (2007), Stern (2006), Weitzman (1998, 2001 and 2007). In the context of regional policy, decisions on the amount of funding allocated by governments, or supra-national bodies, for investment projects in relatively poor regions are likely to be sensitive to the equity weighting placed on social cost and benefit impacts; see Evans et al (2005), Evans and Kula (2011), Kula (2002 and 2007) and Sezer (2006 and 2007).
Welfare weights reflect the marginal social valuations placed on cost and benefit impacts experienced by individuals, or groupings of individuals, in different income classes. Providing income is expressed on an appropriate equivalence scale (e.g. the ‘OECD-modified scale’ as proposed by Hagenaars et al. (1994)) and suitable adjustments are made for any cost of living differences, then income differences should properly reflect differences in living standards (see Cowell and Gardiner, 1999). It is assumed in this paper that the income comparisons made do properly reflect differences in living standards so that the only remaining concern is what weighting to assign to the benefits and costs impacting on those in the different income classes. The extent to which equity welfare weights decline with income can be formally considered by reference to the welfare parameter commonly known as the elasticity of marginal social utility of income (e). From a policy maker’s perspective, (e) can be regarded as the elasticity of marginal social valuation of income, or income inequality aversion parameter. [In some contexts (e) is also referred to as a relative risk aversion parameter; see, for example, Barsky et al (1995) and Gollier (2002)]. In this paper we are concerned with the policy maker’s perspective who are active politicians.
The main theoretical approaches to the estimation of e, together with their relative strengths and weaknesses, are discussed in Stern (1977), Cowell and Gardiner (1999), Evans (2005 and 2008) and Creedy and Guest (2008).They include behavioural evidence, based on both lifetime consumption behaviour and consumer demand for preference independent goods, and the revealed social values of policy-makers as indicated, for example, via the extent of progression in countries’ personal income tax structures. The empirical evidence on (e) is quite wide-ranging and rather dependent on the theoretical approach adopted with a mid-range value that is close to unity; see Cowell and Gardiner (1999), Evans (2005 and 2008), Kula (1984, 1985 and 2004) and Pearce and Ulph (1999). Most of the reported evidence on (e) is based on direct survey approaches such as Barsky et al (1995) who focused on pensioners and Amiel et al (1999) who targeted students. There is scope for further work based on surveys and this study aims to elicit the views of politicians regarding the appropriate social valuation of income. This paper takes up the case in Turkey which has a fast growing economy and welfare weights are an important political issue. Much of the empirical work on (e) has assumed that an iso-elastic social utility, or social valuation, function is relevant but this is quite restrictive where a wide range of income is under consideration. This study is designed to leave the question of iso-elastic social preferences open to empirical investigation.
The paper is divided into four main sections excluding the introduction. Section II focuses on the theoretical properties of the social valuation function and, based on a suitably flexible functional form, an equation linking welfare weights to income is derived. In section III we consider the questionnaire designed to deliver the welfare weight data based on the direct responses of the sample politicians. Aside from political party, other variables incorporated into the questionnaire that may influence a respondent’s social preferences include gender, age, education, family circumstances and current standard of living based on income. In section IV the regression results based on the Turkish sample survey data are reported, both for politicians in different political parties and for all politicians. In each case, formal tests for iso-elastic social preferences are conducted and values of e are estimated at different levels of income. Finally, in section V, the main conclusions based on the empirical findings are reported and suggestions are made for extending the work.
II THEORETICAL APPROACH AND MODELLING
We start with a consideration of the main theoretical properties that a social valuation function is expected to exhibit, selecting a suitably flexible functional form for the purpose of empirical estimation and testing. We then derive an expression for the elasticity of marginal social valuation of income (e) showing that its algebraic form allows for the possibility of a constant, declining, or rising value of (e) as income changes. Finally, in this section of the paper, we consider the welfare principle upon which survey respondents are requested to reveal the social valuation weights (Wi) placed on small income changes (ΔYi) experienced at different income levels (Yi).The revealed Wi based on the survey results can then be used in the regression analysis to estimate (e) and test for iso-elasticity of social preferences with respect to income.
Social valuation function
From the perspective of the social planner or policy-maker, the social valuation of income (V) takes the following general functional form:
V = f (Y) ………………………………… (1)
Where,
Y = household income (on a suitable equivalence scale),
Expected theoretical properties of the function:
A) V increases monotonically with income (dV/dY > 0)
B) d2V / dY2 0 (dV/dY falls as Y rises)
Conditions A and B guarantee concavity of the social valuation function.
A suitably flexible specific functional form for V is as follows:
V = aY + bYC …………………………………………………….. (2)
Conditions A and B requires the following restrictions to be placed on parameters b and c:
c < 1; b > 0 if c > 0; b < 0 if c < 0
So,
dV/dY = MV = a + bcYc-1 ……………………………………….. (3)
Required: if ‘a’ is negative then, a + bcYc-1 0 for all values of Y.
Now define a low income household as having disposable income = Y1 which is above the income required to meet basic living standards. This is important since, theoretically at least, it should guarantee diminishing marginal social valuation of income over the full range of income above Y1. All other income levels considered, Yi, are higher than Y1 (Yi > Y1 ), so MVYi < MVY1. If MVY1 is normalised at unity, then MVYi represent a series of marginal social valuation weights (welfare weights, Wi) that decline in value as Yi increases. So we have,
Wi = MVYi / MVY1 = (a + bcYic-1) / (a + bcY1c-1) ……………… (4)
Setting (a + bcY1c-1) = 1 we have,
Wi = a + bcYic-1 = a + bcY1c-1(1+ gi)c-1 ………………………. (5)
Where,
bcY1c-1 = 1-a; gi = (Yi – Y1)/Y1
Elasticity of marginal social valuation of income (e)
Working from equation 3 (or, alternatively, equation 5) the derived expression for e is as follows:
e = [dMV/dY](Y/MV) = [(c-1)bc] / [aY1-c + bc] ………………….. (6)
Equation 6 allows for the possibility of e rising, falling, or remaining unchanged, as income changes:
(i) If a = 0, then e remains at the constant value c-1 as income changes;
(ii) If a > 0, then e falls as income rises;
(iii) If a < 0, then e rises with income.
Since equation 3 allows for the possibility of constant, falling or rising e, as income increases, the specific social valuation function given by equation 2, along with the associated parameter restrictions, is sufficiently flexible for the purpose of empirical estimation and testing via regression analysis.
There is a special case that is not covered by the generality of the social valuation function as expressed in equation 2 and that is where we have an iso-elastic relationship in which e takes the value -1. For this special case only, the following equation applies:
V = Log Y …………………………. (7)
From equation 7 we have:
dV/dY = MV = Y -1 ……………………………. (8)
And therefore,
e = (dMV/dY)[Y/MV] = -1 ...... (9)
Principle on which equity welfare weights are constructed
For any given marginal welfare loss or gain (ΔW*) resulting from a small change in income (ΔYI) experienced by each of N individuals on N different incomes (Yi) we have:
ΔW* = ΔY1MVY1 = ΔY2MVY2 = ΔY3MVY3 ..= ΔYNMVYN ……. (10)
Where Y1 = Lowest income and YN = Highest income
From equation 10 we have,
MVYi / MVY1 = ΔY1 / ΔYi …………………… (11)
For i = 1 to N.
Setting MVY1 to unity, equation 11 yields a series of equity welfare weights (Wi) that can be applied to monetary benefits and costs impacting on individuals with different incomes. So, we have,
Wi = ΔY1 / ΔYi …………………… (12)
Where 0<Wi <1
The income changes required to deliver equal welfare gains or losses can be obtained using a direct survey-based approach to reveal the social valuation principles of the targeted respondents. A questionnaire designed to elicit such information is outlined in the following section. Using the survey data on Wi we can run non-linear regression equations, based on equation 5, in order to determine the empirical nature of the underlying social valuation function and whether or not the marginal social valuation of income is iso-elastic.
III SURVEY SAMPLE DATA AND REGRESSION RESULTS
YOKO I FEEL THAT THIS SECTION SHOULD BE MERGED WITH
THE FOLLOWING SECTION BUT CUT SUBSTANTIALLY
ESPECIALLY OLD SECTION III. (would you be avble to do that)
Broad aim of the survey
The main aim of this survey is to determine how respondents feel that the elected representatives of society (e.g. governments of countries or supra-national authorities such as the European Commission) should weight changes in income affecting poor, middle-income and rich persons. For instance, respondents’ views have potentially important implications for welfare reforms and social policy relating to a wide range of issues including taxation, social security, education and health-care.
Personal characteristics and circumstances
Gender:
Country of residence:
Political party:
Age in years (please enter X in the appropriate box below):
20-29 / 30-39 / 40-49 / 50-59 / 60-69 / 70 and overEducation: Did you go to University?
If yes, Degree subject(s) studied?
Family circumstances (please enter X in the appropriate box below):
Single,No Dependants / Single,
With Dependant(s) / Married / Partner,
No Dependants / Married / Partner
With Dependant(s)
Annual disposable income: (Please follow instructions given in Note B to Table 1 and make appropriate entry in Table, as directed).
Instructions on completing the questionnaire
The main information required relates to the completion of Table 1 below. A wide range of after-tax incomes is shown in the table, starting at a low level of income (Y1), which is a little in excess of the amount required to cover basic living costs, and ending with the income of a very rich person (YN).
Now consider that we have a society consisting of N individuals with each person having a different income ranging from Y1 to YN. Furthermore, assume that these individuals are able-bodied adults of similar age (without dependants) facing the same prices for consumer goods and services. In fact, the only defining difference between them relates to the disposable incomes shown in Table 1.
Suppose that the individual with the lowest income Y1 is required to make a small contribution of TL100 towards a government initiative designed to raise funds to help finance charitable works overseas. This reduction in the individual’s disposable income involves a given loss of welfare to the country as a whole (the loss of income of TL100 is entered in the column headed Y1=10).
Now for each of the individuals on higher disposable incomes, ranging from Y2 to Y15 in Table 1, please indicate the deduction of income that you feel is required in each case, to just match the loss of welfare to society resulting from the TL100 income deduction incurred by the lowest income individual on income Y1.
Table 1 Social valuation of income and revealed social preferences
After-tax annual income (TL000)
Yi / Y1=10 / Y2=18 / Y3=24 / Y4=36 / Y5=60 / Y6=72 / Y7=96 / Y8=120Income
deduction
(in TL) / 100
Your
Income
After-tax annual income (TL000)……continued
Yi / Y9=150 / Y10=180 / Y11=210 / Y12=240 / Y13=300 / Y14=360 / Y15=480Income
deduction
(in TL)
Your income
Definition: After-tax income = Disposable income = Gross income less income tax and national insurance contributions.