Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out

income distribution

VIEWS: 54 PAGES: 13

									                                                                                    1


                 INCOME DISTRIBUTION AND INFLATION.

    AN EMPIRICAL ANALYSIS FOR WESTERN EUROPE (1995 – 2006).



                             Miltiades N Georgiou PhD



                                     ABSTRACT

In the present paper it will be pointed out with panel data that inflation worsens
income distribution, but not in a linear way. More specifically, at lower inflation
levels as inflation goes up, then income distribution worsens more rapidly than in the
case of higher inflation levels (hyperinflation). The sample covers annually most
western European countries for the period 1995 - 2006. The contribution of the
present paper is that at any level of inflation northern-western European countries
have a better income distribution than in the case of the whole sample average. Panel
data analysis is made feasible by means of Eviews software package.


Keywords: Income distribution, Inflation, Wage Rigidity, Econometric Model with
Panel data (single equation).
JEL classification: C23, D3, E31, J2.




Dr. M. N. Georgiou has an M.Sc. (Economics) from Stirling University, and a Ph.D.
(Economics) from the University of Thessaly in Greece. He left in July 2008
Emporiki Bank as a Department Head on Market Analysis. He has contributed papers
to: “Hellenic Bank Association”, “Economic Review of Commercial Bank of
Greece”, “Applied Research Review”, “Applied Financial Economics Letters”,
“Max Planck Institute of Economics” (4) and “Social Science Research Network
(SSRN)” (23). He is familiar with Eviews software package



Author confirms that this article has never been published by any other journal.
Further, this article expresses only author’s personal ideas and of nobody else.

Home address: Dr. M. N. Georgiou, 69 Eptanisou Street, Athens 11257, Greece.
E-mail: mng@insitu.gr




              Electronic copy available at: http://ssrn.com/abstract=1542696
                                                                                       2


PART 1. THEORY



        Regarding income inequality, according to Bhattacharya (2001, p.6), the

standard mainstream argument (often called the Keynes-Fisher-Hamilton hypothesis)

is that business that are net debtors gain through inflation. Inflation reduces the real

value of their nominal obligations, thus benefiting debtors and hurting creditors.

Further, since wages are set in nominal terms, inflation erodes their purchasing power.

The view that inflation increases income inequality is shared by Shiller (1996), Buliŕ

(2001), as well as (Kane and Morisett, 1993). Further, according to Buliŕ (2001)

inflation effect on income distribution is non-linear. In other words, the reduction of

inflation from a low level to an even lower yields a negligible gain to gini coefficient.

A good review of theories about inflation and income inequality is in Crowe (2006).

        In the present paper it will be shown that at lower inflation levels as inflation

goes up, then income distribution worsens more rapidly than in the higher inflation

levels (see graph 1.). This is in agreement with the study of Tobin (1972) (that wage

rigidity is higher at lower inflation levels).




               Electronic copy available at: http://ssrn.com/abstract=1542696
                                                                                         3


PART 2. THE MODEL



After the above, the following model (1) can express the positive impact of consumer

inflation on income inequality measured by gini coefficient.

giniit = c0 + c1 ln_inflit + errorit            (1)

It is ex-ante expected that c1 > 0. The subscript [i] stands for the country, while [t] for

the year. Data are annual and their source is Eurostat. Variable [ln_infl] stands for

the natural logarithm of the annual consumer inflation. The sample covers most

western European countries covering annually the period 1995 – 2006, making a

sample size of 154 observations, as shown in table 3.

             The method of GLS will be used to handle heteroskedasticity, since,

according to Yaffee (2003, p.10) for large samples the methods of “fixed effect” as

well as “random effect” are not efficient when there is heteroskedasticity (either

between time periods, or between cross sections). For equation (1) there are basically

two types of model, the “fixed” and “random” effects. The appropriate choice

depends on whether one treats constant terms αi’s as some fixed numbers or ‘random

drawings’ from a specific distribution. As the correlation structure of the error term is

ignored, a more efficient estimation method would be the Generalized Least Squares

(GLS) provided that there is no correlation between the x’s and the α’s. GLS requires

weighting the observations of y and x by Σ –(1/2):

                 1       1 − ϑ ΄ 
∑
    −1 / 2
             =     IT − 
                          T ii 
                                   
                 σ
                                 
                       σ2
where θ =
                   σ 2 + Tσ α
                            2




First one obtains an estimate θ by estimating the equation:

yit − yi. = β ΄ ( xit − xi. ) + (uit − ui. )                   (2)
                                                                                        4


Once the component variances have been estimated, one forms an estimator of the

composite residual covariance and GLS transforms the dependent and regressor data

(Baltagi, 2001; Davis, 2002). In this paper GLS cross section weights method will be

preferred, because it yields a robust model. The panel data analysis is made feasible

by means of Eviews software package.

        The detailed results are shown in table 1 and diagnostic tests, based on

(Halkos, 2003) in table 2. It can be seen that model (1) as estimated meets the three

required criteria of heteroskedasticity, specification and normality. Further, there is no

serial correlation. Hence, model (1) is robust. It can be seen that (at 95%) the

coefficient of [ln_infl] is positive and statistically significant.

        The same model with the same method can be applied to the northern-western

European countries (which are: Austria, Belgium, Denmark, Finland, France,

Germany, Ireland, Netherlands, Norway, Sweden and UK). This sample has 112

observations. The detailed results are shown in table 4 and diagnostic tests, based on

(Halkos, 2003) in table 5. It can be seen that model (1) as estimated for the northern-

western European countries meets the three required criteria of heteroskedasticity,

specification and normality. Further, there is no serial correlation. Hence, model (1) is

robust. It can be seen that (at 95%) the coefficient of [ln_infl] is positive and

statistically significant. It is worth mentioning that the coefficient of [ln_infl] in the

case of the northern-western European countries is lower than the coefficient of

[ln_infl] in the whole sample.
                                                                                      5


PART 3. CONCLUSIONS



       In the present paper it is shown that at lower inflation levels as inflation goes

up, then income distribution worsens more rapidly than in the case of higher inflation

levels (see graph 1.). This is in agreement with Tobin (1972) (that wage rigidity is

higher at lower inflation levels).

       It is also shown (with a simulation based on the estimated equation in graph 2)

that at any level of inflation northern-western European countries have a better

income distribution than in the case of the sample average (which refers to all western

European countries).
                                                                                       6


    REFERENCES



    1. Baltagi, B. H. 2001. Econometric Analysis of Panel Data. 2nd edn, John Wiley

and Sons, Chichester.

    2. Bhattacharya, J. 2001. Inflation, Real Activity and Income Distribution, Iowa

State University. [available also at:

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=260220&CFID=118951&CFTOK

EN=51039753].

    3. Buliŕ, A. 2001. Income Inequality: Does Inflation Matter? IMF Staff Paper,

Volume 48, No.1. [available also at:

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=875012&CFID=339221&CFTOK

EN=14066653].

    4. Crowe, C. 2006. Inflation, Inequality and Social Conflict, IMF WP/06/158.

[available also at:

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=920250&CFID=118951&CFTOK

EN=51039753].

    5. Davis, P. 2002. Estimating multi-way error components models with

unbalanced data structures. Journal of Econometrics, 106, 67–95.

    6. Halkos, G. E. 2003. Environmental Kuznets Curve for Sulphur: Evidence

Using GMM Estimation and Random Coefficient Panel Data Models. Environment

and Development Economics 8: 581-601.

    7. Kane, C. and Morisett, J. 1993. Who would vote for inflation in Brazil? World

Bank Policy Research Working Papers WPS1183.

    8. Shiller, R. 1996. Why Do People Dislike Inflation? NBER Working Paper,

No.5539. Cambridge, Massachusetts.
                                                                            7


   9. Tobin J., 1972. Inflation and Unemployment, American Economic Review,

62, 1-18.

   10. Yaffee, R.,A. 2003. A primer for panel data analysis. September, p10.

Available at: www.nyu.edu/its/pubs/connect/fall03/pdfs/yaffee_primer.pdf.
                                                                            8


APPENDIX

                                  1. All sample

Table 1. Results

Method                                     GLS cross section weights
c0                                                                     43,421
                                                                       (14,59)
c1                                                                       3,793
                                                                        (5,27)
          2
Adjusted R                                                               0,301
Durbin Watson                                                           1,887
Note: For n = 154 (at 95%) dU = 1,74961.
                                                                                                   9


Table 2. Diagnostic Tests1

TESTS                                            GLS Cross section             Critical values
                                                     weights                     (at 95%)
Heteroskedasticity                                    0,654                         3,914
Heteroskedasticity                                    0,076                         3,914
Heteroskedasticity                                    0,660                         3,841
Heteroskedasticity                                    2,290                         5,991
Heteroskedasticity                                    0,001                         7,815
RESET1                                                0,006                         3,841
RESET2                                                  0,022                        5,991
RESET3                                                  0,045                        7,815
Normality                                               1,858                        5,991
Notes:
Test 1: Regression of the squared residuals on X. That is, u 2 = x′t γ1 + v t,1
                                                             t



Test 2: Regression of absolute residuals on X. That is, | u t |= x′ γ 2 + v t,2 (a Glejser test)
                                                                  t


                                               ˆ
Test 3: Regression of the squared residuals on Y

                                               ˆ     ˆ
Test 4: Regression of the squared residuals on Y and Y 2

Test 5: Regression of the log of squared residuals on X (a Harvey test)

                                   ˆ
Test 6: Regression of residuals on Y 2

                                   ˆ
Test 7: Regression of residuals on Y 3

                                   ˆ
Test 8: Regression of residuals on Y 4

Test 9: Normality test (Jarque Bera)




1
    The diagnostic tests are based on Halkos (2003)
                                           10


Table 3. Data (all sample)

                   Country     Period
Austria                      1995 - 2006
Belgium                      1995 - 2005
Cyprus                       2003 - 2006
Denmark                      1995 - 2006
Finland                      1996 - 2006
France                       1995 - 2006
Germany                      1995 - 2005
Greece                       1995 - 2006
Ireland                      1995 - 2006
Italy                        1995 - 2001
Netherlands                  1995 - 2006
Norway                       2003 - 2006
Portugal                     1995 - 2001
Spain                        1995 - 2006
Sweden                       1999 - 2005
UK                           1995 - 2005
Source: Eurostat
                                                                                            11


                           2. Northern- western European countries

Table 4. Results

Method                                                GLS cross section weights
c0                                                                                    35,887
                                                                                      (12,03)
c1                                                                                      2,154
                                                                                       (3,50)
               2
Adjusted R                                                                              0,114
Durbin Watson                                                                          2,109
Note: For n = 112 (at 95%) dU = 1,70982.




Table 5. Diagnostic Tests2

TESTS                                            GLS Cross section        Critical values
                                                     weights                (at 95%)
Heteroskedasticity                                    2,470                    3,935
Heteroskedasticity                                    1,738                    3,935
Heteroskedasticity                                    2,459                    3,841
Heteroskedasticity                                    2,465                    5,991
Heteroskedasticity                                    0,334                    7,815
RESET1                                                0,042                    3,841
RESET2                                                  0,045                 5,991
RESET3                                                  0,047                 7,815
Normality                                               0,998                 5,991
Notes: The explanations are as in table 2.




2
    The diagnostic tests are based on Halkos (2003)
                                                                               12


                      Graph 1. All Sample (Historical Data)


                       Gini coefficient versus Inflation

       40


       35


       30
gini
       25


       20


       15
            0%   1%    2%    3%     4%      5%       6%   7%   8%   9%   10%

                                         inflation
                                                                            13




                              Graph 2. Simulation


          Gini coefficient in all sample and in northern-western
                                  countries

    40

    35

    30

gini 25

                                         gini all countries
    20
                                         gini northern-western countries
    15

    10
          0%     2%      4%       6%        8%       10%      12%     14%
                                    inflation

								
To top