Determinants of insurance enrollment

Ownership of health insurance policy is a mechanism for protecting an individual family ’ s financial security. It is also a means for a risk-aversion strategy for the cost of medical care, loss of productivity time during the illness, and in a more serious case death. This study examines the factors that influence the ownership of insurance policies at an individual level using a binary logistic regression model. The data used in this study was taken from the Health and Retirement Study (HRS), the fifth wave in 2002 with 3206 respondents of the survey, whereby thirty-nine percent of the respondents happened to have an insurance policy scheme. The outcome of this model indicated that retirement status, household income, years of schooling, and marital status variables were all found to have a statistically (at 95% confidence level) associated with ownership of insurance policy. Contrastingly, the other covariates, namely: age at the time of the survey, race, and gender of the respondents had insignificantly relationship with the ownership of insurance policy.


INTRODUCTION
The motivation for this article resides to understand what drives the insurance policy ownership among these aged groups of individuals. This study will be relevant to policymakers as it will deepen their understanding of the factors that influence the decision of individuals to get insurance. The main objective of this article is to evaluate the determinants of insurance coverage in the 52-86-yearold population. Moreover, it also estimates the predictors of insurance policy ownership among the sampled respondents. This study will be relevant to policy-makers as it will deepen their understanding of the factors that influence the decision to get insurance 1 . This knowledge is expected to help policy-makers decide on the best strategies to adopt to increase insurance enrollment 2 . Therefore, it is imperative to recognize the factors which influence the purchase of insurance policy at an individual level of the above age group. This article examines the factors that affect health policy ownership using the fifth wave of the Health and Retirement Study. The paper is organized as follows. Section two provides a brief account of the pieces of literature and empirical findings on the ownership of insurance. Section three presents the empirical method of the study. Section four summarizes the dependent and explanatory variables and how they were measured operationally. The next section provides the result of descriptive statistics and binary logistic regression. This section also offers the discussion part of the result. Finally, the article finishes with some concluding remarks and policy implications.
Theoretical framework As far as the household level of income is concerned, the larger the level of income, the more life insurance consumer can afford to purchase. Thus, with the household income of an individual, I expected to 'have a positive relationship with the life insurance consumption 3 . The level of education of a given individual positively affects his/her demand for life insurance ownership 4 . As argued by Truett and Truett (1990) an individual's higher level of education is associated with a strong desire to protect dependents and safeguard their standard of living 5 . Hence, I hypothesized a positive association between insurance ownership and the level of individual education, since higher education is attributable to the corresponding purchasing power and access to information on a health insurance policy in general 6 .
Among the demographic measure, 'empirical evidence of the factors influencing health insurance uptake indicates 3 Kjosevski, 2012, p. 240 4 Ibid., p, 241. 5 Truett & Truett, 1990 quoted in Ibid. 6 Badu, et al., 2018; that age plays an important role' 7 in insurance policy enrolment. This is because health insurance coverage increases with age 8 . It 'was attributed to an increment in additional healthcare needs and increased older people's financial security. It is hypothesized that an increase in age would influence ownership of insurance policy as an individual risk-avoiding strategy. In connection to this, the other predisposing factor that makes it more likely that individuals will seek insurance policy ownership is gender 9 . In addition to age, gender is expected to be associated with the probability of owning an insurance policy. Within the category of a social factor, marital status also predisposes an individual to acquire an insurance policy 10 . The other variable associated with the attainment of insurance policy is race. Particularly, in some countries, racial disparities are reflected in insurance status, whereby racial/ethnic minorities are more likely to experience being devoid of an insurance policy 11 . As compared to other races, the majority race, (82%) 12 of the white race is expected to be positively associated with the probability of owing insurance policy.

RESEARCH METHODOLOGY
The data set of this article was taken from the Health and Retirement Study (HRS), the fifth wave in 2002. It was collected by the National Institute of Aging at the University of Michigan Institute for Social Research. The concern of this study is to interpret the outcome variable as a likelihood of owning an insurance policy or not. For this paper, a total of 3206 respondents with complete data on the key dependent variable were included in the data analysis. Descriptive statistics and a binary logistic regression model were used to overview the sample's characteristics and identify factors associated with insurance policy ownership 13 . The logistic regression analysis was used to determine the relationship between a binary response of the outcome variable and continuous or categorical independent variables of this paper 14 . Thus, as compared to Linear Probability Model (LPM), the logit model was employed in the empirical estimation of this paper because 'the logit model can overcome the problem associated with LPM' 15 which is 'plagued with heteroscedasticity, non-normality of the distribution term, low R 2 and non-fulfillment of the 0 < (Y1/X1) < 1 restriction of binary model' 16 . Moreover, the other justification employing the logit model is to realize 'the advantage of being more robust such that the independent variables do not have to be normally distributed or have equal variance in each group, does not assume a linear relationship between the independent and dependent variable, does not assume homogeneity of variance and does not assume normality of error term' 17 .
Suppose the outcome variable of this study y can be assumed only dichotomous response values, which for convenience and without any loss of generality will be the value of 1 if ownership of insurance policy occurs and 0 if it does not. In this case, let the probability of y = 1 be p and the probability of y = 0 be (1-p) 18 . Then the expected value of y is the probability that owning an insurance policy will occur. In this regard, when the logit model is applied, it can be expressed as follows 19 : Whereby the maximum likelihood estimable equation of binary logistic regression model is given as Logit( ) = 0 + 1 1 + 2 2 + ⋯ + (1), where 0, 1, ……, p are regression parameter; X1, X2….X p are explanatory variables, and is the probability of success 20 . Accordingly, for this paper, Insurance Policy ownership (y) = 0 + 1 retirement status + 2 age + 3household income + 4 years of schooling + 5martial status + 6 race + 7 gender 21 .

Measurement of variables
In this econometric estimation, health insurance enrollment is defined as the dependent variable of interest of this article. This dependent variable is equal to whether or not a person has health insurance (0 or 1). In other words, it assumes a value of 1 if the person is currently enrolled in the insurance scheme otherwise 0 22 . In the estimation of a binary logistic regression model, demographic and socio-economic characteristics of the respondents were included as an explanatory variable. The predictors' variables examined in the study were selected based on the factors cited in the literature as influencing insurance policy ownership. The demographic measure comprises age, race, and gender. Socio-economic factor includes retirement status, income, and years of schooling. Accordingly, the seven explanatory variables of this model that are likely to influence the ownership of insurance policy are defined as retirement status, age at the time of the survey, household income, years of schooling, marital status, race, and gender of the respondent. Table 1 below summarizes the dependent and explanatory variables and how they were measured operationally. Health Insurance 0 = "no" 1 = "yes" Key Independent Variable Operational Measurement Retired 0 = "no" 1 = "yes" Age Continuous positive whole numbers in years HH Income Amount earned measured in thousands Educational years Continuous positive whole numbers in years Married 0 = "no" 1 = "yes" White 0 = "no" 1 = "yes" Gender of the respondent 0 = "male" 1 = "female" Journal of Research in Social Science and Humanities, 2(1), June 2022, -37

RESULTS AND DISCUSSION
Descriptive statistics Table 2 presents the percentage of distribution of dependent and independent variables. Overall, thirty-nine percent of the respondents happened to have an insurance policy scheme, and sixty-one percent of the sub-sample do not have a health policy. In the overall sample, the retirement status mean was 62 %, with a Standard Deviation (STD) of 0.484. Thirty-nine percent of the sample respondents do not have retirement status. The average age was 67 years (STD = 3.676) and the household income was 45, 263.9. Seventy-three percent of the respondents were married individuals. Among the total sample, the majority of the respondents are white (82%) with an STD = 0.384.

Logistic regression and discussion
Binary regression is used to assess the strength of a relationship between one dichotomous dependent and independent variable. This regression method helps in predicting the value of a dependent variable from one or more independent variables. This model also helps in predicting how much variance is being accounted for in a single response by a set of independent variables. Among the binary regression, the logit model was used to analyze data to predict an outcome of insurance ownership among 3206 respondents of the survey. As table 4 of the next page illustrates, the final model was able to explain between 78.0 % and 10.6 % of the variance. The model was found to fit the data adequately (Hosmer and Lemeshow's X 2 = 42.627. P < .001). Overall, the model was able to correctly predict 61.8 % of all cases. Seven predictors were included in the model, using the Enter method. Table 3 below depicts that, three of these predictors successfully predicted insurance ownership since they have positive coefficient as the table shows clearly the relationship between the predictors and the outcome. The beta (B) is the predicated change in Log Odds -for 1 unit change in the independent variable, there is Exp (B) change in the probability of the outcome 23 . 23 Mayers, 2013.
When we come to coefficient interpretation, retired individuals in comparison to those who are non-retired are less likely to have an insurance policy because they have a negative coefficient of significance. Those who have high household income is more likely to have insurance. Table 3 above shows the most important outcome of the logistic analysis of insurance. The B column of this table indicates the regression coefficient for each independent variable of the model 24 . The Wald statistic and the respective significance indicate which of the independent variable predict insurance policy ownership 25 . Whereas the Exp (B) column indicates the odds ratios for each independent variable 26 . Specifically values greater than 1 indicate a greater likelihood of insurance policy; values less than 1 signify reducing likelihood 27 . Retirement status, age, and marital status were associated with decreased likelihoods of insurance ownership. Income, years of schooling, and gender were associated with an increased likelihood of insurance ownership. These three predictors were more than once are more likely to be received insurance policy (OR 1.003; 100.3 % CI: 1.001 -1. 004, 1.147;114.7% CI: 1117114.7% CI: -1178111.1% CI: .945-1307 respectively). 24 Mayers, 2013, p. 458 25 Ibid. 26 Ibid. 27 Ibid.  In the above binary logistic regression model, the assumption for linearity and multicollinearity were satisfied. Particularly for the multicollinearity test see table  5 of Collinearity Statistics below which shows the "Tolerance" data to be not too close to 0 (preferably nor below.2) and the VIF figure not exceed 10 29 . The Table  suggests that we are fine on both accounts.

CONCLUSIONS AND POLICY IMPLICATION
The retirement status, household income, years of schooling, and marital status variables were all found to have a statistically (at 95% confidence level) associated with ownership of insurance policy. Contrastingly, the other covariates, namely: age at the time of the survey, race, and gender of the respondents had insignificantly relationship with the ownership of insurance policy. Thus, there is two policy implication of this article. The policy options for promoting insurance ownership among the 52-86-year-old population could be aided by the two major findings of this article 30 . Higher household incomes are 28 See in page 6 29 Mayers, 2013, p. 457 30 Markowitz, Gold 7 Rice, 1991 significant predictors of insurance policy ownership. And hence, economic development program geared at (a) improving income of the retired married individuals of the society 31 ; (b) policies should be aimed at ensuring that the majority of the society to attain a secondary education level will increase the probability of owning an insurance policy 32 . Furthermore, the insurance industry may use the findings to inform actions to enhance insurance schemes for individuals.