Research Paper on Presence of Bias in Econometric Analyses

Introduction

Interpreting data is a very critical step in data analysis. The process involves data modeling and inferential analysis to identify relationship and causal relationships. Therefore, during analysis of data any errors inherent in the data or the model must be identified and solved to ensure accurate interpretation and generalization. For instance, in regression models, misspecifications can be found, and their diagnosis with ordinary residual analysis such as Ordinary Least Squares (OLS) estimator enables corrective actions to be taken thus realizing accurate predictions. Misspecifications and bias results in wrong coefficients thus wrong conclusions. Bias that require attentioninclude endogeneity bias and selection bias.The purpose of this paper is seeking and assessing evidence of these biases in the OLS estimates of married women's wage elasticity concerning hours of work. The paper will start by discussing the presence of bias in econometric analyses, and it causes, then summarize the main data features, then discuss empirical model design, then how to treating endogeneity-bias in elasticity estimates, main results of comparative analysis and finally present a conclusion.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Save 25%

Bias in Econometric Analysis

Although bias is said to exist in most of the educational branches, there are branches where its effects can be detrimental and cause confounding wrong conclusions. As such in every branch where the outcome of analysis will be generalized and be used to make a decision that affects people, it is prudent to ensure that the effects of the biases are minimized or made to be as little as possible. Some of the results affect the development of nations and between nations thus caution should be taken as wrong policy interpretation will destroy the existing institutions in the hope of making better ones and if the decisions are informed by results that are not robust, consistent over time irreparable damage can occur

Take for instance the branch of economics where most decisions that pertain to economics and drivers of economic development are discussed, evaluated and policies made; great caution must be exercised here as wrong conclusions can result to the formulation of wrong policies. Moreover, as the genre of economics grows new information that can change how the decisions are made come to surface and econometrics based on the textbooks which are severely biased and sometimes not current cannot be relied upon to make a judicious decision. It is important therefore to make the modeling and interpretation of data be devoid of the influence of time and data type collected.

Types of Bias

Bias in analysis can be caused by some things such as incompleteness of data, wrong modeling or from sample selection and endogeneity biases. Data collected in cross-sectional surveyis important in undertaking confirmative data analysis whose main aim is to forecast and test model for use in policy persuasion. Also, the data analysis is also used to test for correlation and causality which are not the same but important in understanding data that is to be analyzed. Moreover, the quality of these aspects, correlation, and causality, require that certain aspects such as randomness be present otherwise the aspect will be termed as biased. In different forms of research such as consumer demand studies and household labor supply, bias is likely to occur due to non-randomness of the observations (Blau& Kahn, 2007). Another important point to note is that sometimes samples are incomplete thereby affecting estimation for modeling using categorical data and incomplete data samplesresult in biased estimates. The bias increases if the missing the observations are not at random, and corrective measures to reduce bias such as data imputation have to be implemented.

When faced with incomplete samples several statistical issues arise that affect the reliability and credibility of the results obtained. The statistical issues are because incompleteness of the samples may be due to randomness or non- randomness. When a researcher uses incomplete samples the |, is biased because the missing parts are assumed to be zero of which if the data were collected they will not be zero. This results into loss of efficiency and precision which could arise if the researcher used the complete section of the sample while ignoring the extra available data information.Another important bias is sample selection bias. Sample selection bias arises as result of individuals or groups for data analysis are selected in ways that does not meet the assumption of randomness. Sample selection can be corrected using the Heckman model. Endogeneity bias results because some of the variables used as explanatory variables are determined within the model or are determined together with the dependent variable affecting the relationship between the error terms and the resulting estimates. In this case, an explanatory variable or variables are correlated with the error term. Endogeneity can also be caused by measurement error, simultaneous causality or omitted variables.

Selection Bias, Endogeneity Bias, and the Labor Supply Among Women Workers

One area where these form of bias are likely to occur is in the labor market. Labor supply in the form of hours and wage rate are both determined by the model making then endogenous thus using one to estimate another result in endogeneity bias. Also, the utilization of the number of hours worked excludes the characteristics of the women who never worked resulting in selection bias. In the US the wage elasticities among married women have been identified as shrinking challenging the established notion of the existence of a gap between male and female wage elasticities (Heim, 2007). Moreover, it has been found that laborforce participation and hours of work over the period 1980-2000 resulted into the dramatic reduction in wage elasticity to about 50-56 percent (Blau& Kahn, 2007). These developments can be attributed to the increase in wage opportunities for women as traditional gender roles become disbanded and new roles start forming up which do not limit the participation of the women in the labor market. It is believed that this wage elasticity gap exists at the micro level analysis due to the limited focus and concentration of the attributes that affect the microenvironment. Nevertheless, microeconomic are important as they form the basis for macroeconomic analysis. Additionally, microeconomicestimates continue to vary across studies (Keane & Rogerson, 2012; Eika, Mogstad& Zafar, 2014) with labor supply elasticities at the microeconomic being based on the hours worked which results into smaller elasticities than those presented in macroeconomic models for the labor market.

Summary of the Main Data Features

In this paper, the aim isseeking and assessing evidence of bias, caused presumably by selectivity and endogeneity, in the OLS estimates of married women's wage elasticity concerning hours of work.The dependent variable which is also the primary modeled variable is the wife's hours of work. The variable is measured in a continuous manner showing the hours worked by the wife of the household. Other major variables that are used to be able to explain the primary variable arewife's hour of work, wife's age, education, hourly wage/earning, work experience, number of children under 18, and husband's income (or family income). As stated earlier women traditionally involved in such as household tasks, market work or leisure which attracted no pay. But the changes in family dynamic and aspirations to better standards of living women role has been changing (Eika, Mogstad& Zafar, 2014). Male support though limited is also identified as a key driver of this change. Wife's age is measured as a continuous variable with age ranging between 25 and 60 years. Young spouses tired of just sitting and waiting for the husband to bring income home are engaged in entrepreneurial activities that generate income and given that they still have an opportunity to learn they have become very active increasing the hours spent by women in remunerative employment.

Education is measured categorically as the highest level received. Education has also become a universal need as women seek for equality not at home but in the workplace. Jobs that were by tradition men oriented are now open to anyone with the necessary skills required opening an avenue for women involvement in employment that earns then money. Although they may be limited in experience, they are willing to put more hours to learn and be experienced even with little pay. Sometimes the demand of household work may limit the number of hours available for engagement outside the hose especially when there are small children at home who require a lot of attention and care (Heim, 2007). Hourly wage rate is a continuous variable showing what the women receive per working hour. This variable is only observed for women with positive hours worked creating a selection bias. Average hourly wage rate is considered a better measure in this case as takes into account the fact that not all laborers are paid on an hourly basis. Moreover, hours worked and wages are determined together by the model causing endogeneity bias.

Another driver to women involvement in remunerative employment or income generating activity is the availability of extra family income for investment or the limitation in family income to meet the household needs driving the family to expand labor supply by involving the women at home. Family income is measured as a continuous variable showing the amount of money that the family earns from all sources combined. These factors individually and jointly effect on the participation of the women in the labor market, and even it influences the wage rate they are going to receive. As societal changes open up ways for the engagement of women in the supply of labor, especially in the sectors that were by tradition male-oriented, more factors than the ones stated here will come to the surface and will illuminate the discourse on self-selection and endogeneity bias.

Descriptive Statistics

Table 1 shows the descriptive statistics of variables used in this study. According to the results the average age of the women was 41 years with a standard deviation of 10 years. Moreover, each household had an average of one child under 18 years, and average wage rate was US$18.41. Among the respondents, the average family income was US$ 105, 903 with the wife having an experience of an average of 5 years.

Table 2 presents the percentages of women who were working and their highest degree received. According to the results, about 80 percent of the women were working. The results also show that about 50 percent of the women had received bachelor's degree with another 21 percent having received an Associate of arts degree. Masters recipient was about 23 percent while other degrees' recipients were below 2 percent with medical degrees having the least recipient.

Highest college degree Received Percentages

Associa...