What email address or phone number would you like to use to sign in to Docs.com?
If you already have an account that you use with Office or other Microsoft services, enter it here.
Or sign in with:
Signing in allows you to download and like content, and it provides the authors analytical data about your interactions with their content.
Embed code for: Econometrics-Determinants-of-Inequality
Select a size
This paper might have some grammatical errors though I believe the econometric theory is agreeable and further study is warranted on inequality and its determinants.
Determinants of Inequality
By: Joe Savarese
Income inequality in the U.S. has been a popular topic recently, especially the growing disparity that has been observed since the 1980’s. It poses questions of fairness in the social environment as well as potential challenges to the future economy. This paper discusses the determinants of income inequality and provides regression analysis using the gini-coefficient as the dependent variable measure. Several choice independent variables are tested to determine the statistical significance of their potential relationships to income inequality in the U.S.
The gini-coefficient is a measure of dispersion within a countries income distribution. The goal of this research is to develop a better understanding as to what causes the trends in income inequality. As of now there is much debate as to what may be driving the growing inequality in the U.S. and what the possible repercussions are. Economically, many believe that this trend may lead to negative consequences on growth and employment. Socially many question the fairness associated with cases of inequality as well as the potential detriment to justness within society.
Previous literature on income inequality is not yet well established in terms of U.S. studies only. Much of the work has been done on cross country analysis to understand the disparities between developing and developed nations. There has been some work on inequality in the U.S. focusing on changes in educational attainment being the driving cause. Stating that there is a skill-biased technical change in the advent of information technology substituting high education tasks for middle education work in turn raising upper middle class wages and lowering those below. Another paper focuses primarily on international trade factors playing the major role in trending inequality in the U.S. It suggests that its influence has been declining over the last decade shifting to technological exports as the predominant influence instead.
This paper attempts something a bit different in that it broadens the scope of analysis to see if there is evidence for several independent variables being determinants of income inequality in the U.S. over the last forty five years or so.
Three different specifications were ran and considered worthiest of note. The first was a simple linear model between the gini-coefficient as it is with several independent variables. The next specification looked to determine if these independent variables took time to truly effect income distribution because it’s possible that time is a necessary factor in the process of change. This was achieve through using the lags of the independent variables. The third was in response to the weakness of the prior to specifications. The dependent variable this time was modified so that the regression was run on the changes in the gini-coefficient rather than just as it is but without lagging the independent variables.
The chosen independent variables include government spending which takes into account transfer payments, public investments, and other ways in which income potentially becomes redistributed. This variable is split up into two separate variables because of the innate differences in which they will be spent. The two variables are state/local government spending and federal government spending. The expected sign of both of these variables is to be negative, that is when government spending increases inequality falls because income with a progressive tax code becomes redistributed.
Next variable included is inflation which holds an important place in monetary theory when it comes to a households earnings. Inflation influences the value of assets one is holding, whether it be in the actual cash money someone holds or longer term less liquid assets of a household’s portfolio. What is expected is that this would decrease inequality because of structural differences in rich and poor household portfolios. The rich have access to a greater variety of assets and have the ability to hold those investments for longer. Therefore inflation would actually lessen the value of their future earnings while the poorer households would experience less depreciation as their holdings would be in the more liquid and accessible forms.
Another independent variable that could play an essential role in the determination of inequality is union membership. The percent of the work force in unions is important because it’s approximately a measurement of a workers bargaining power, that is, how much say do workers, particularly at the lower end of the income spectrum, have in determining their wages. Also, it could be considered a measure of protection that they have against being exploited by corporatists. The expected sign of this variable is negative as well because the more power workers have, typically of the lower income brackets, the better they can achieve fair wages and secure employment.
Finally the model needed a variable that would be a proxy for trade openness and integration. Globalization has structurally changed the economy in many ways and it’s possible that it could have an effect on income inequality. It would be expected that increases in this would actually increase inequality in the sense that there are few with enough capital to truly benefit from this and as many employments have actually started to become outsourced it becomes harder for U.S. workers to earn a competitive living.
Data sources for the gini-coefficient were given by the Census Bureau, while all data for the other independent variables was taken from the Federal Reserve for Economic Data (FRED) except for the data on union membership. That data was taken from a Georgia State University project. All of these links can be found in the appendix. For government spending the data was given total expenditures and then divided by real GDP for percent format, inflation is given as a percent of money supply growth, and union membership as union workers as percent of labor force. Foreign direct investment was taken in the form of millions of U.S. dollars so it needed some modification to better fit the model. Using data from real GDP from FRED foreign direct investment as a percent of GDP was generated by first turning it into billions of dollars and dividing it by GDP. Data for the gini-coefficient could only be found dating back to 1967 so that is as far back as the time series runs up until the year 2012.
The results below in the three tables are of the three specified models described above respectively. All of the tests run on these regressions as will be explained are shown in the appendix. Initial regression results suggested the possibility of some autocorrelation so below is Table-1 showing the newey-west estimators.
Table-1 Newey-West Estimators of lag(2) (Correcting for autocorrelation):
Table-2 below shows the second specification where all independent variables are lagged, also using newey-west estimators to correct for autocorrelation.
Table-2 Newey-West Estimators of lag(2). Lagged independent variables:
Table-3 below is another attempt to obtain the proper specification. This was run with the original independent variables but with a modified dependent that is the changes in the gini-coefficient instead.
Table-3 OLS Estimators. Changes in the gini-coefficient as the dependent (L stands for lagged):
Taking a look at the results from Table-1, state and local government spending (SLS) was the one variable that was not significant contrary to above stated theory. Federal spending (FS) had an expected negative coefficient providing evidence to the relationship that increases in federal spending tends to decrease inequality. The coefficients for inflation (Inflation), union membership (UM), and foreign direct investment (FDI1) were consistent with expected results as well. That is, inflation and workers participating in worker unions may improve income equality while the growing economic dependence on trade integration, specifically in already developed economies like the U.S., may tend towards more inequality. These appear to be worthy results but are not in fact the most conclusive because of specification troubles. Ramsey’s reset test shows that perhaps there exist omitted variables or other specification errors. The second model from Table-2 that was run was the model of lagged independent variables which shows interesting but spurious results. By using the AIC test it was found that of the three specifications this one was the most correctly specified. All the variables were statistically significant, including state and local government spending this time, although its coefficient was of the opposite sign predicted, showing a positive correlation with income inequality. Finally, Table-3 was interesting in that all the independent variables became insignificant perhaps saying something about the model or theory in general. Whether the results are discouraging because of the model used or discourage the other two models as faulty is unknown. Despite this, it did not experience the same faults of the error terms that the first two models did.
Inequality, as it is a growing trend for U.S. households and a global epidemic across countries is an interesting topic that is hard to grasp. This field is in need of much further analysis to get at the true root of its causes as well as its implications for society. The results above attempted to get at some of the determinants of income inequality and provided evidence of relationships between income inequality and federal government spending, union membership, foreign direct investment, and inflation. What should be gleaned from this evidence is that it’s very important to take a look at the roles of federal spending policies and their role in helping or hurting households from the trending disparity. Treatment of workers, their ability to bargain, and avoid potentially scrupulous exploitations. The changing structure of the U.S. economy and the merits of the international trade strategies being employed. As well as monetary policy and the effects of manipulation U.S. currency on incomes and purchasing power. Future inequality involves many unforeseeable consequences for the future, requires much more intensive analysis, and should be addressed carefully in the coming years.
-David H. Autor, Lawrence F. Katz, Melissa F. Kearney, “Trends in U.S. Wage Inequality: Revising Revisionists”, The Review of Economics and Statistics, May 2008, Vol. 90, No. 2, Pages 300-323, President and Fellows of Harvard College and the Massachusetts Institute of Technology.
-J. David Richardson, 1995. "Income Inequality and Trade: How to Think, What to Conclude," Journal of Economic Perspectives, American Economic Association, vol. 9(3), pages 33-55, Summer.
Federal Government Spending:
State and Local Government Spending:
Foreign Direct Investment:
Below in A-Table-1 are the results for the first models test before replaced with newey estimators. Ramsey’s is close but still must be rejected. Suggests no heteroskedasticity but a possible problem with autocorrelation.
A-Table-1. OLS regression run and diagnostic tests.
Next is the OLS regression results and tests for the second specification. Ramsey’s suggests omitted variables despite AIC test showing it as the best of the three. There is no suggestion of heteroskedasticity but also has potential autocorrelation.
A-Table-2: OLS regression run for lagged independents and diagnostic tests:
Finally the last of the models OLS regression and tests. Ramsey’s test suggests no omitted variables, passes the heteroskedasticity test, as well as the serial correlation test.
A-Table-3: Change in Gini specification and diagnostic tests:
Below is a test for multicollinearity between the independent variables. There appears to be multicollinearity between FDI and UM as well as FDI and SLS.
AIC for the three specifications in their respective order. The order from best to worst judging by these functional form tests are the second specification, then the first, and lastly the third.
A-Table-5:sults are discouraging because of the model used or discourage the other two models as faulty is unknown. Despite this, it did not experience the same faults of the error terms that the first two models did.
Next is the OLS regression results and tests for the second spe