1 An Assessment of the Effect of the Private Health Insurance Incentives Scheme on Dental Visits Ynon Gablinger, Elizabeth Savage, Jane Hall CHERE, University of Technology Sydney January 2006 Abstract Several initiatives that were introduced by the Commonwealth government in the late 990s with the intention to prop up private health insurance membership rates are examined here with regard to their impact on dental visits. It is hypothesized that improved access to health services conferred upon by the private health insurance led to an increase in dental visits. Using a non-linear decomposition technique on data from 995 and 200 we find evidence to support this hypothesis.. Introduction Ever since its inception in the 970s, participation rates in private health insurance (PHI) have been generally falling. In an attempt to reverse this trend and ease the mounting pressure on the public health system, the commonwealth government initiated a couple of incentives schemes in the late 990s. The first the Private Health Insurance Incentives Scheme (PHIIS), in 997 pledged a subsidy to low income earners who took up private health insurance. The second the Private Health Insurance Rebate (PHIR), introduced on January 999 extended the first to apply to all income levels. More specifically, the incentives schemes provided a 30 per cent rebate for all PHI expenses. The individual, on his side, could enrol in one of three forms of private health insurance: hospital cover only, ancillary cover only, or both hospital and ancillary cover. Dental health services are largely covered under the ancillary insurance. Thus, to the effect that the rebate scheme prompted individuals to take up private ancillary cover, it presented a new subsidy for dental insurance, albeit an indirect one; according to Spencer (200, p.39) an estimated $36-$345 millions per year of government expenditure has been spent on dental rebates, nearly twice as much as the estimated expenditure on public dental care. Once insured, an individual not only enjoys the low cost of dental treatments, but also the benefit of short or indeed no waiting time, which is very substantial for people who rely on the public system (quote). This furthers his or her likelihood of visiting the dentist. The Scheme also penalized high income earners who did not take up private health insurance.
2 2 In subsequent years, membership in private health insurance funds rose only marginally: ancillary insurance rates rose from 3.7% in July 997, at the introduction of PHIIS, to 32.9% in March 2000 (The Private Health Insurance Administration Council, Statistical trends in membership and benefits, What eventually brought the desired increase in membership was the introduction of the Lifetime Health Cover in Ancillary insurance rates rose to 39.2% at once, and have been steadily increasing since. 3 However the effect of the Lifetime Health Cover should not affect dental visits because it is very late in the sample period and most new insurers will not have used health services yet, and because the new entrants were motivated to enter by a deadline, so it is unlikely that they will use it to go to the dentist, at least in the short run. We focus on dental services for a good reason; there is an urgent need to improve dental access, especially among lower income people; the most common complaint lower income people express with regard to health is about teeth. Poor people cannot afford treatment of oral diseases which is one of the most expensive diseases to treat and quickly develop bad teeth. This, in turn, may lead to serious health and social problems: The loss of teeth makes eating fresh fruits and vegetables difficult, and a diet heavy in soft, processed foods exacerbates more serious health problems, like diabetes. The pain of tooth decay leads many people to use alcohol as a salve. And those struggling to get ahead in the job market quickly find that the unsightliness of bad teeth, and the selfconsciousness that results, can become a major barrier. (The Moral-Hazard Myth, Malcolm Gladwell, The New Yorker, 29/08/05). In Australia, in particular, oral health among adults lags behind that of most industrialized countries. Reportedly, tooth decay is the most common health condition. 4 A person who does not have PHI either has to pay the full cost of the treatment, or, if he or she holds a concession card, the treatment is free. But without a widespread public dental hospital system, very long waiting times and a shortage of dentists, about 90 percent of dental services are provided by the private sector. Our data shows that between 995 and 200 there has been an increase in the share of population that has visited a dentist in the last year. In order to assess whether all or some of this increase is a result of the incentives schemes, we run two logit regressions, one for each survey year, with a dummy for whether or not the individual has visited the dentist in the last year as the dependent variable. We then use a non-linear decomposition technique to isolate the part of the increase that cannot be explained by observable variables. This part is assumed to represent the impact of the incentives schemes on dental visits. We find that as much as 70 percent of the increase may be attributed to the incentives schemes. 2 Lifetime Health Cover was inducted on July 2000 in order to encourage people to take out hospital insurance earlier in life, and to maintain their cover. Under this initiative, people who take out hospital cover early in life will be charged lower premiums throughout their life, relative to people who take out cover later. 3 The rates for hospital insurance closely follow the same pattern: from 3.9% in June 997 through 32.2% in March 2000, to 43.0% in July Edentulism (the loss of all teeth) and periodontal (gum) disease come in third and fifth place respectively.
3 3 2. Data and Methodology We use the National Health Surveys of 995 and 200. The surveys provide information with regard to health status, health-related actions (dental consultations), health-related habits (smoking, exercise), health insurance, and socio-economic and demographic characteristics. Many of the questions are relevant only to adults of 5 or 8 years of age (health insurance, education, labour force status). Under the assumption that the dietary and oral preventive care practices of the child are largely influenced by his parents, in most such cases we extended the status of the adult in the income unit to its dependents. Thus, for example, the income level of the head of family was applied to its dependents. Similarly, education, labour force status, place of birth, level of exercise, and health insurance were treated the same way. In the case of health insurance it is essential since most private health insurance packages automatically extend to dependents. The two surveys differ in their sample selection method. While in 995 information on each and every member of the household was collected, in 200, information was collected only on one adult (8 years or older), one child aged 7-7, and all children aged 6 or lower within each household. As a result, the 995 sample is twice as large as the 200 sample. However, in 995 a random sub-sample was excluded from various questions (e.g., education level, dental visits) and the removal of those observations rendered the two samples practically equal in size. Information regarding dental visits was derived from the question when was the last time you consulted a dentist or dental professional. While the questionnaire offered a menu of possible answers, we grouped them into two categories: less than 2 months and 2 months or more. 5 Persons who answered don t know (less that half a percent in each sample) were excluded, as well as infants aged 5 years or younger. Figure presents the distribution of answers for each year. Percentage of Sample 56% 54% 52% 50% 48% 46% 44% 42% 40% Figure : The Distribution of Dental Visits, 995 and % 54.6% 50.5% Year year or less more than year 49.5% 5 Ideally we would have information about the number of visits in, say, the last year, but unfortunately, the question that inquires about the number of visits restricts the timeframe to the previous two weeks!
4 4 We estimate a binary logit model where the dependent variable assumes unity for an individual who has visited a dentist in the last 2 months and zero otherwise. The explanatory variables include age, sex, income unit type, labour force status, education level (highest qualification), geographical area of residency, whether native born, level of exercise (in the past two weeks), and smoking habits. To control for income we use the income decile of the income unit. Two key dummy variables control for the type of private health insurance that the individual owns, one for individuals with ancillary insurance only, and one for individuals with both hospital and ancillary insurance. We expect both coefficients to be significantly positive, with a somewhat added strength to the coefficient of the first dummy since people who took up ancillary insurance alone may have done so with a clear intention of using the services of a specialist in particular, a dentist. Last, we include interaction terms between the ancillary dummies and the unit income decile in order to capture the different rates at which the legislation impacted different income groups for ancillary covered persons. 3. Decomposition We are interested in isolating the proportion of the observed difference in dental visits that is due to the incentives scheme introduced in the intermediate years. The standard approach for decomposition in linear models is the so called Blinder-Oaxaca technique, developed independently by Blinder (973) and Oaxaca (973). Most commonly applied in the context of racial or sexual differences in wages, the technique involves estimation of separate regressions for two groups, in our case two years, and then use a simple construction to decompose the difference in average outcomes between the two groups to differences in average group endowments (difference in variables), and differences in treatment/behaviour (difference in coefficients). Formally, for a linear model y = β x + ε, we use the exact relation y = βˆ x for each group, and with the aid of an interim component of which we subtract and then add back, we obtain the Blinder- Oaxaca decomposition as given in equation. In this equation, the first component on the right represents the share of the difference that is due to variables and the second represents that of the coefficients. y ( x x ) ˆ β + ( ˆ β β ) x 0 ˆ ˆ ˆ y = β x β x y y = () The Blinder-Oaxaca decomposition, however, is inadequate in the context of non-linear models. In this context, Gomulka and Stern (990) developed a parallel approach for a Probit model that accounts for differences over time. The approach involves decomposing the difference between averages of predicted probabilities rather than averages of outcomes. Fairlie (2003) expands on their method and demonstrates its applicability to Logit models as well. Formally, for a non-linear model y = F ( β x) + ε, we use the asymptotic relation y = ( N ) F( βˆ x) for each group, and again, with the aid of an interim component we obtain the decomposition as given in (2).
5 5 y 0 0 ( ˆ N ) ( ˆ ) N ( ˆ N β x F β x F β x ) F ( ˆ β x ) 0 N 0 F i i = + i y 0 0 i= N i= N i= N i= N 0 i (2) where again, the first component on the right stands for the share of the difference that is due to variables and the second stands for the share that is due to coefficients. 0 Notice that both () and in (2) we arbitrarily chose to use x and β as the arguments of the interim component. An equally valid alternative would be to use 0 and β instead. The decomposition would then look like y 0 ( ˆ N ) ( ˆ ) N ( ˆ N β x F β x F β x ) F( ˆ β x ) N 0 F i i = + i y i= N i= N i= N i= Here, the first component on the right stands for the share of the difference that is due to coefficients and the second stands for the share that is due to variables. The existence of two alternatives none of which is more correct than the other is known in the literature as the index problem. In this paper we report both alternatives. 4. Results 4.. Trends and Statistics Table shows, per surveyed year, the percentage of persons who reported to have consulted a dentist in the previous 2 months. The bottom line shows that between 995 and 200 the percentage of persons who reported to have consulted a dentist in the previous 2 months has increased by 5 percent. 6 We would like to assess this rise in dental usage in light of the policy legislation that was enacted in the intermediate years. Since the legislation renders oral health benefits only to individuals who have private health insurance with ancillary cover, a distinction is made between individuals who have a PHI with ancillary cover and individuals who do not (i.e. those who have PHI without ancillary cover or do not have a PHI at all). The comparison reveals a striking difference between the groups: the increase in recent dental visits for ancillary-covered persons, 7.94 percent, is four-fold that for ancillary-uncovered persons,.98 percent.. Table Percentage of Persons who Consulted a Dentist in the Last Year Ancillary Cover Difference (SE) No 40.49% 42.47%.98% (0.66%) Yes 53.94% 6.89% 7.94% (0.83%) Total 45.4% 50.46% 5.05% (0.52%) N 0 i. x 6 The increase is highly significant with a standard error of 0.5%.
6 6 The large increase in dental visits among ancillary-covered individuals comes, at least partly, from an increase in the share of ancillary-covered individuals between 995 and 200; a comparison of the surveys reveals that the share of ancillary-covered individuals has increased by 4.6 percent (from 36.5 percent in 995 to 4. percent in 200); it is reasonable to assume that a relatively large proportion of the newly joined individuals have used health services in general and dental services in particular more frequently or more recently. At any rate, since the newly enacted policies provide dental incentives only to ancillary covered individuals, this high percentage increase among ancillary covered individuals is a first indication that they did in fact positively impact dental visits; that the 7.94 percent increase is a response to the newly enacted incentives. To better assess whether this observed increase may be attributed to the introduction of the incentives scheme, we turn to regression analysis Regression Analysis We regress individual and income unit characteristics on a binary dependent variable that assumes unity for individuals who had a dental consultation in the last 2 months and zero otherwise. Parameter estimates are presented in table x and estimates of marginal effects in table x2. The sign of the coefficients is generally as expected, with the rate of dental visits increasing with income, education, and level of regular exercise; and decreasing with age. Women visit the dentist more often than men and so do foreign born individuals (and their income-unit members). By far, the variables that contribute most to the probability of a recent dental visit are the two ancillary insurance dummies. Individuals with ancillary insurance are at least 6 percent more likely to visit a dentist than those without. What we are primarily interested in, however, is not the factors that determine the rate of dental visits per se, but the extent to which the incentives schemes contributed to its increase. For that we need to closely examine the increase in the rate of dental visits and try to isolate the portion that is due to the incentives schemes. To this end, we employ Fairly s Logit decomposition technique (Fairly 200) to decompose the increase into a component accounted for by a change in the variables (endowments) between the years, and a component accounted for by a change in the coefficients. The latter represents the unexplained portion of the increase. Table 3a: Decomposition Results unrestricted Increase in rate of dental visits.0504 (se=0.005) First Index Due to Variables.0047 (9.4%) Due to Coefficients.0457* (90.6%) Second Index Due to Variables.039* (27.6%) Due to Coefficients.0365* (72.4%) Percentage of gap in parentheses * Significant at % level
7 7 As is evident from table 3a, changes in the model s coefficients account for the overwhelming majority of the increase in the probability of dental consultation 90 percent according to the first index and 72 percent according to the second. What this means is that we cannot explain much of the increase in the rate of dental visits by means of an increase in education, or ancillary cover or any other variable in our model. Most of the increase comes from an unexplained source. The question is can this large unexplained portion of the increase be attributed to the incentives schemes? Even though a change in policy could alter each coefficient in our model, some coefficients seem more prone to respond than others. A natural candidate would be the coefficient over the ancillary dummy. It would have been instrumental to sub-decompose the portion that is due to coefficients into sub-portions that represent the effect of specific coefficients. Unfortunately this type of exercise would be unproductive in that the outcome is specification-dependent (Jones 983). 7 Instead, we make a different kind of experiment. We run a regression in which we allow only the coefficients over ancillary and ancillary-income interaction variables to differ between the two years, while restricting all other coefficients to be equal. Formally, the regression takes the form Y Y X Z = β + γ 95 + γ 0 + ε, 0 0 X 0 Z 95 0 where X is a matrix of variables whose coefficients are restricted, 8 and Z represents a matrix of variables whose coefficients are unrestricted in our case, ancillary, and the interaction between ancillary and income. Such decomposition ensures that the coefficient portion is comes from a change in the ancillary and ancillary-income coefficients and should therefore reflect a conservative measure for the impact of the incentives scheme on dental visits. Table 3b: Decomposition Results - restricted Increase in rate of dental visits.0504 First Index Due to Variables.028 (55.7%) Due to Coefficients.0223 (44.3%) Second Index Due to Variables.035 (62.4%) Due to Coefficients.090 (37.6%) 7 Unlike the case of coefficients, we can decompose the component that is due to variables into subcomponents, each representing a specific variable or a group of variables. If not for the Lifetime Health Cover, It would have been of interest to examine the share of the increase that is due to a change in the frequency of ancillary coverage. Our data shows that the rate of ancillary coverage has increased by 4.5% (se=0.5%) between the years. However, the rate of ancillary coverage was influenced by the Lifetime Health Cover much more than it was by the rebate scheme. The Lifetime Health Cover should not influence our other results as it became effective only in July 200. However, by then, enrolment had to be in place in order to avoid the excess fee. Moreover, the rebate scheme was aimed at the issue of affordability of health services while the Lifetime Health Cover aimed at increasing membership rates. 8 Including the constant term.
8 8 * Percentage of gap in parentheses Table 3b demonstrates that the coefficient portion diminishes to about 40 percent; still a sizable share of the increase. This 40 percent may be regarded as a lower bound to the effect of the incentives schemes on the rise in the rate of dental visits between 995 and Illustration For a policy maker the relevant question with regard to dental health is: how much more likely is a non-insured person in 995 to be insured and visit a dentist by 200? We can demonstrate the increase in this likelihood for different individuals. An individual without ancillary insurance in 995 has a predicted probability of 38.2 percent for having visited a dentist in the previous year. In 200, had that individual purchased ancillary insurance, perhaps due to the incentives, that probability would have increased to 58.5 percent. It would be of interest to be more specific in the characterization of the individual in order to see differences between the rich and the poor; the young and the old; females and males. Table 4 demonstrates such a comparison. Table 4 compares the increase in the probability of a dental visit between 995 and 200 for the above types. The comparison is between a poor and a rich type and between ancillary-covered and ancillary-uncovered individuals. Table 4: Predicted Probabilities of a Dental Visit for Predefined Types 995 without Ancillary 200 with Ancillary Difference Sample Switching Probability Young, Low Income (.049)* Young, High Income (.033)* Old, Low Income (.033) negative Old, High Income (.037) 0.3 Females, Low Income (.049)* negative Females, High Income (.043)* Males, Low Income (.030) Males, High Income (.042) 0.09 *Significant at 5%
9 9 As expected, for both types the increase in dental usage for ancillary covered individuals is much higher than the one for uncovered individuals. The positive evidence is that for poor individuals the increase in the probability of dental usage is substantially higher than for wealthy individuals. A type I, second decile, ancillary covered individual has increased his probability by.20/.358=.587 while a similar type from the ninth decile has increased his probability only by.43/.452=.36. Similarly, for type II the numbers are.857 and.539 respectively. This is particularly important in light of increasing inequality in the access to dental care. There is not much difference between the behaviour of type I and type II. 5. Concluding Comments Improved utilization but not improved income. Since the poor are in greater need of oral health care than the rich, we need to go further to have a negatively sloped gradient over income deciles to align it with needs. Only caused greater salaries for dentists. That s the income redistribution.
10 0 References Blinder, Alan, S. (973). Wage Discrimination: Reduced Form and Structural variables, Journal of Human Resources 8, Fairlie, Robert, W. (2003). An Extension of the Blinder-Oaxaca Decomposition Technique to Logit and Probit Models, Discussion Paper No. 873, Economic Growth Center, Yale University. Gomulka, Joanna, and Stern, Nicholas (990). The Employment of Married Women in the United Kingdom, Economica 57(226), Jones, F. L. (983). On Decomposing the Wage Gap: A Critical Comment on Blinder s Method, Journal of human Resources 8, Oaxaca, Ronald, L. (973). Male-Female Wage Differentials in Urban Labor Markets, International Economic Review 4, Smith, Julie. (200). How Fair is Health Spending? The Distribution of Tax Subsidies for Health in Australia, The Australia Institute, Discussion Paper Number 43. Spencer, John, A. (200). What Options Do We Have for Organising, Providing and Funding Better Public Dental Care?, Australian Health Policy Institute at the University of Sydney in collaboration with The Medical Foundation University of Sydney, Australian Health Policy Institute, Commissioned Paper Series 200/02. Spencer, John, A. (2004). Narrowing the Inequality Gap in Oral Health and Dental care in Australia, Australia Health Policy Institute, commissioned Paper Series Sered, Susan S., Fernandopulle, Rushika (2005). Uninsured in America: life and death in the land of opportunity, University of California Press. Wenzlow, Audra, T., Mullahy, John, Wolfe, Barbara, L. (2004). Understanding Racial Disparities in Health: The Income-Wealth Paradox, Institute for Research on Poverty, Discussion Paper no
11 Table x Logit Estimates of a Dental Consultation in the Last Year Variable Name Constant (0.) 995 Year Dummy (0.39) Ages (0.0) Ages (0.072) Ages (0.065) Ages (0.063) Ages (0.07) Ages (0.085) Ages (0.088) Ages 75 over -2. (0.) Female (0.032) Income Unit: Single + Dep (0.074) Income Unit: Couple (0.05) Income Unit: Couple + Dep (0.048) Unemployed 0.09 (0.095) Not in Labour Force 0.06 (0.059) Foreign Born 0.05 (0.035) Qualification: Basic/Skilled (0.04) Qualification: Undergraduate 0.86 (0.054) Qualification: Postgraduate 0.39 (0.05) Location: Inner Regional (0.049) Location: Remote Areas (0.047) Exercise Level: Low 0.25 (0.04) Exercise Level: Moderate (0.043) Exercise Level: High (0.065) (0.) (0.076) (0.092) (0.07) (0.066) (0.073) -.56 (0.085) -.75 (0.093) (0.) (0.033) 0.77 (0.069) (0.046) (0.05) (0.4) 0.52 (0.047) (0.038) 0.87 (0.039) (0.057) (0.05) (0.04) (0.048) (0.039) (0.044) 0.38 (0.07)
12 2 Table x, continued Variable Name Smoker (0.042) Ancillary and Hospital (0.4) Ancillary only (0.239) Income Unit Decile (0.084) Income Unit Decile (0.083) Income Unit Decile (0.086) Income Unit Decile (0.088) Income Unit Decile (0.09) Income Unit Decile (0.092) Income Unit Decile (0.094) Income Unit Decile (0.095) Income Unit Decile (0.05) (Decile 2)*Ancillary*Hospital (0.93) (Decile 3)*Ancillary*Hospital (0.93) (Decile 4)*Ancillary*Hospital (0.7) (Decile 5)*Ancillary*Hospital (0.62) (Decile 6)*Ancillary*Hospital (0.59) (Decile 7)*Ancillary*Hospital (0.56) (Decile 8)*Ancillary*Hospital (0.53) (Decile 9)*Ancillary*Hospital (0.5) (Decile 0)*Ancillary*Hospital (0.53) (Decile 2)*Ancillary (0.428) (Decile 3)*Ancillary (0.383) (Decile 4)*Ancillary (0.35) (Decile 5)*Ancillary (0.34) (Decile 6)*Ancillary (0.32) (Decile 7)*Ancillary (0.299) (0.043) (0.33) 0.88 (0.275) (0.08) (0.08) 0.82 (0.087) 0.68 (0.09) (0.093) (0.096) 0.48 (0.0) 0.3 (0.02) 0.4 (0.4) 0.22 (0.9) (0.82) (0.75) (0.72) (0.69) (0.69) (0.68) (0.67) (0.7) (0.409) (0.422) (0.389) (0.382) (0.42) (0.36)
13 3 Table x, continued Variable Name (Decile 8)*Ancillary (0.36) (0.38) (Decile 9)*Ancillary (0.308) (0.407) (Decile 0)*Ancillary (0.346) (0.6) Sample Size 8,292 8,548 * SE in parentheses
14 4 Table x2 Marginal Effects df/dx (SE) Variable Name Ages ** (0.023) Ages ** (0.04) Ages ** (0.03) Ages ** (0.03) Ages ** (0.04) Ages ** (0.07) Ages ** (0.07) Ages 75 over ** (0.08) Female 0.058** (0.007) Income Unit: Single + Dep ** (0.06) Income Unit: Couple ** (0.02) Income Unit: Couple + Dep (0.0) Unemployed 0.02 (0.022) Not in Labour Force 0.025* (0.04) Foreign Born 0.02 (0.008) Qualification: Basic/Skilled 0.022** (0.009) Qualification: Undergraduate 0.043** (0.02) Qualification: Postgraduate 0.087** (0.0) Location: Inner Regional ** (0.02) Location: Remote Areas (0.0) Exercise Level: Low 0.049** (0.009) Exercise Level: Moderate 0.062** (0.00) Exercise Level: High 0.080** (0.04) -0.67** (0.08) ** (0.02) ** (0.06) ** (0.05) ** (0.07) ** (0.09) ** (0.02) ** (0.02) 0.046** (0.008) 0.04** (0.06) (0.0) 0.022* (0.02) (0.027) 0.035** (0.0) 0.02** (0.009) 0.043** (0.009) 0.062** (0.03) 0.093** (0.0) -0.03** (0.00) -0.05** (0.02) 0.047** (0.009) 0.065** (0.00) 0.07** (0.05)
15 5 Table x2, continued df/dx (SE) Variable Name Smoker ** (0.0) Ancillary and Hospital 0.67** (0.02) Ancillary only 0.92** (0.037) Income Unit Decile (0.020) Income Unit Decile (0.09) Income Unit Decile ** (0.09) Income Unit Decile ** (0.02) Income Unit Decile (0.02) Income Unit Decile ** (0.020) Income Unit Decile ** (0.020) Income Unit Decile ** (0.02) Income Unit Decile ** (0.022) (Decile 2)*Ancillary*Hospital -0.6** (0.048) (Decile 3)*Ancillary*Hospital -0.2** (0.048) (Decile 4)*Ancillary*Hospital -0.0** (0.043) (Decile 5)*Ancillary*Hospital -0.08** (0.040) (Decile 6)*Ancillary*Hospital ** (0.040) (Decile 7)*Ancillary*Hospital -0.64** (0.039) (Decile 8)*Ancillary*Hospital -0.32** (0.038) (Decile 9)*Ancillary*Hospital -0.3** (0.037) (Decile 0)*Ancillary*Hospital ** (0.038) (Decile 2)*Ancillary -0.86* (0.05) (Decile 3)*Ancillary (0.095) (Decile 4)*Ancillary (0.087) (Decile 5)*Ancillary -0.76** (0.078) ** (0.0) 0.63** (0.025) 0.66** (0.047) ** (0.09) (0.09) 0.042** (0.020) 0.039* (0.02) 0.054** (0.02) 0.077** (0.02) 0.092** (0.022) 0.070** (0.023) 0.09** (0.025) (0.043) (0.045) (0.043) (0.042) -0.07* (0.042) * (0.042) ** (0.042) -0.0 (0.040) (0.042) 0.76** (0.066) (0.09) (0.092) (0.095)
16 6 (Decile 6)*Ancillary (0.074) (Decile 7)*Ancillary -0.40* (0.075) (Decile 8)*Ancillary -0.2** (0.076) (Decile 9)*Ancillary ** (0.073) (Decile 0)*Ancillary (0.086) Note: ** indicates p<.05, * indicates p< (0.096) (0.09) (0.092) (0.099) (0.2)
17 7 Predicted Probabilities of Dental Visits in the Last 2 Months Per Income Decile: Type I without Ancillary Income Decile Predicted Probabilities of Dental Visits in the Last 2 Months Per Income Decile: Type I with Ancillary Income Decile
18 8 Predicted Probabilities of Dental Visits in the Last 2 Months Per Income Decile: Type II with Ancillary Income Decile Predicted Probabilities of Dental Visits in the Last 2 Months Per Income Decile: Type II with Ancillary Income Decile