The Effects of the State Children s Health Insurance Program (SCHIP) on Children s Health Insurance Coverage. Hua Wang

Transcription

1 The Effects of the State Children s Health Insurance Program (SCHIP) on Children s Health Insurance Coverage Hua Wang Department of Health Policy and Administration University of North Carolina at Chapel Hill December 2004 Abstract This paper analyzes the effects of availability of the State Children s Health Insurance Program (SCHIP) on children s SCHIP enrollment and Medicaid enrollment in addition to more commonly focused overall public and private coverage and the uninsured rate. Linear probability models with controls for state and quarter of year fixed effects were estimated using data from the National Health Interview Survey (NHIS) for the years 1995 through My results show that nationwide among children for whom SCHIP became available, 5 percent enrolled in SCHIP, less than 3 percent gained public coverage, 1.4 percent lost private insurance, and 1.1 percent became no longer uninsured. The spillover effect of SCHIP implementation on Medicaid enrollment among children under age 6 and with family income below poverty level was 5.5 percent. SCHIP s effects on health coverage were greater for lower-income children, older children, and states implementing only stand-along programs. These findings, many of which are similar to other recent national estimates, will contribute to the upcoming Congressional discussion about the reauthorization of SCHIP. 1. Introduction The uninsured rate of children increased from 14.1 percent in 1993 to 15.3 percent in Uninsured children have less access to health care than children with coverage (Patel, 2001). After the Clinton administration failed to implement the national health insurance 1

2 plan in the mid-1990s, the political pressure moved toward incrementally broadening health coverage for at least children. The result is the State Children's Health Insurance Program (SCHIP), a joint federal-state health insurance program created by Congress in With $40 billion federal matching funds apportioned for the next 10 fiscal years ( ), SCHIP was regarded as the single largest expansion of health coverage since the enactment of Medicaid and Medicare in It aims to expand health insurance to children whose family incomes are too high to be eligible for Medicaid but too low to afford private insurance. Approximately 4 million children, 40 percent of the thenestimated 10 million uninsured children would be covered by the new program (Wooldridge, 2003). Compared to the Medicaid program, states were granted considerable flexibility in SCHIP program design. All states have implemented their SCHIP programs between 1997 and In 2002, more than 5 million children under age 19 were ever enrolled in the program nationwide. As early as year 2005 Congress may begin to discuss SCHIP reauthorization. The take-up rates of most public programs are below expectations historically. For example, the enrollment rate of Medicaid is only around 30 percent among eligible children. Another concern policy makers have upon public health insurance programs is the crowd-out of private coverage by public programs. The concern was raised from the extensive studies on crowd-out effects of the previous Medicaid expansions in the 1990s, though the study results vary largely. In order to prevent low enrollment and crowd-out of private coverage, SCHIP legislation explicitly mandated states to take special efforts in the outreach and enrollment process; design and implement anti-crowd-out procedures such as waiting-period or co-payment requirement; and facilitate children eligible for Medicaid to enroll Medicaid during SCHIP outreach. Since the implementation of SCHIP, states have seen rapid increase in SCHIP enrollment and large and even bigger increase than SCHIP enrollment in Medicaid take-up as well. However, mainly due to lack of data, no study has ever estimated the effects of SCHIP implementation on SCHIP or Medicaid enrollment at the national level. 2

3 Several recent studies have provided national estimates about the effects of SCHIP on overall public and private coverage and the uninsured rate among SCHIP eligible children. In order to correct for endogeneity bias associated with program eligibility, these studies used instrumental variables regressions, which are less efficient estimators. This paper estimated the effects of SCHIP implementation on SCHIP enrollment and Medicaid enrollment, which are not available in previous studies. The spillover effect of SCHIP on Medicaid participation was tested among children under age 6 and with family income below the Federal Poverty Level (FPL) because they are not eligible for SCHIP and only eligible for Medicaid program in any state by law. Instead of more commonly used program eligibility, this paper looked at the effects of program availability of SCHIP on health coverage outcomes. Program availability can be measured accurately and is believed to be exogenous at the child level so its effects can be estimated by more efficient estimators. This paper used linear probability models controlling for state and time fixed effects. The National Health Interview Survey (NHIS) files, the only national continuous survey that started to collect SCHIP enrollment information as early as 1999, provided complete data for this study. The findings confirmed the recent estimates (Lo Sasso and Buchmueller, 2004) that the SCHIP program had a small but significant effect on enhancing public coverage and reducing the uninsured rate during its first several years. The crowd-out effect is also found small. In addition, this paper finds that the SCHIP availability increased SCHIP enrollment by about 5 percent and as a spillover effect increased Medicaid participation by 5.5 percentage points among the youngest and poorest children. 2. SCHIP implementation Within four years since the passage of the SCHIP legislation in October 1997, all states (including District of Columbia) established and implemented their own SCHIP programs but at different time. Most states (33 in toatal) began enrollment during 1998, 8 states 3

4 started as early as 1997, another 8 started in 1999, and the last 2 states began enrolling children in 2000 (see Figure 1) (Thompson, 2002). Figure 1. SCHIP implementation: number of new programs in the quarter of the year Program Type 1997 (Law passed) Total Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Medicaid-SCHIP Separate-SCHIP Combination Total Source: Background for the Report to Congress (2003). The federal legislation offered states three options for program design. A state can use the federal fund to expand its current Medicaid program, create a separate stand-alone program, or implement a combination of both. By March 2002, 16 states were operating Medicaid expansion programs (M-SCHIP), 15 were operating separate SCHIP programs (S-SCHIP), and 20 were operating combination programs (C-SCHIP) (CMS-website). The main reason why some states chose the Medicaid expansion approach is because they can take advantage of their existing Medicaid operation system. On the other hand, states that opted for a separate program wanted to use the flexibility under the federal law to design a program that best suits the state s need (Thompson, 2002). Given the flexibility by law to modify their programs after initial implementation, state programs have been evolving ever since. Prior to SCHIP, 27 states had pre-existing state-run non-medicaid programs for lowincome children and many of them were replaced by SCHIP. For example, 9 states discontinued their pre-existing programs and transferred enrollees to SCHIP, and the other states maintained some or all of their pre-existing plans alongside SCHIP to serve children who are ineligible for Medicaid or SCHIP. Therefore, children under preexisting state programs may be moved to SCHIP, which is usually more generous than state pre-existing programs. 4

5 States invested heavily in outreach and enrollment for SCHIP programs. Outreach strategies usually consist of state-wide media campaigns and community-based efforts to reach hard-to-reach families. To improve program enrollment, states commonly use the following ways to streamline and simplify the application and enrollment process: joint application for SCHIP and Medicaid, mail-in application option, minimal documentation requirement, no asset test, no face-to-face interview, and 12-month continuous eligibility period. Several states even have begun to extend coverage to parents in order to spur enrollment of eligible children. Many states report that they adopted some of the effective SCHIP outreach strategies into their Medicaid outreach and enrollment procedures. States are generally satisfied with the speed of program enrollment, which has increased rapidly from year to year. The federal law identified basic eligibility criteria for SCHIP and left considerable discretion to states in how they could set up eligibility criteria. It authorized states to expand income eligibility up to 200 percent federal poverty level (FPL) or 50 percentage points above the Medicaid eligibility level in effect on March 31, 1997 (since some states had already provided extensive coverage under Medicaid) (Thompson, 2002). In SCHIP implementation, the income eligibility level and the expansion of income eligibility vary across states, age groups, and program types. By March 2001, 25 states had set their income eligibility levels for SCHIP at 200 percent FPL (authorized level by federal law); 16 states set their thresholds lower than that level; and 10 states set their eligibility standards higher than that level. One interesting feature of SCHIP income eligibility criteria is that states are permitted to establish income threshold above the levels authorized by the federal law through the use of income disregards. 43 states disregard certain type of income when assessing program eligibility. Most states provide 12-month continuous eligibility and annual eligibility re-determination. Overall, M-SCHIP programs tend to have lower income eligibility levels than separate-schip programs. In states implementing combination SCHIP programs, the Medicaid expansion programs usually cover lower income children and the separate programs cover higher income children. 5

6 Unlike Medicaid, which set different income eligibility thresholds for different age group of children, the income eligibility limit is the same for all ages of children in SCHIP. Because state Medicaid programs are much less generous to older children, the expansion of income eligibility is much greater for them in SCHIP. The average expansion of income eligibility is 61 percent FPL for children age 1-5 and it is 130 percent FPL for children age (Thompson, 2002). Another important eligibility criteria set up by the SCHIP legislation is that only uninsured children are eligible for SCHIP. Children covered by private insurance plans are not allowed to switch to SCHIP. Medicaid children or even uninsured children who are eligible for Medicaid are not eligible for SCHIP because the federal matching rate is higher for SCHIP children than for Medicaid enrollees. Moreover, the legislation mandates states to establish rules that prevent crowd-out of private coverage. Waiting period is mostly used as the anti-crow-out provision. SCHIP law also requires states to cooperate with Medicaid programs and implement screen and enroll procedure during SCHIP outreach and enrollment process, in which a SCHIP applicant should be first screened for Medicaid eligibility and then if found eligible, will be enrolled into Medicaid program. States reported that the screen and enroll procedure and the Medicaid enrollment reforms have helped Medicaid take-up increase. In summary, all states have implemented SCHIP programs to expand eligibility for public insurance to near-poor children though at different time and with different program design. SCHIP law mandates innovative strategies in states program implementation, including aggressive outreach and enrollment, anti-crowd-out private coverage, and facilitation of Medicaid enrollment. States administrative reports suggest effectiveness of the new program. 3. Econometric Analysis 3.1. Model specification 6

7 This paper estimated the effects of SCHIP implementation on children s health insurance, including SCHIP enrollment, Medicaid enrollment, overall public coverage, private coverage, and the uninsured rate. The hypotheses are the SCHIP program would increase children s participation in SCHIP and Medicaid (through spillover effect) and therefore the overall public coverage; it may reduce private coverage (through crowd-out). The change in the uninsured rate would depend on how much the crowd-out effect will cancel out the increase in the overall public coverage. SCHIP implementation is indicated by SCHIP availability, which is the independent variable of primary interest in this study. Program availability indicates whether SCHIP program is available to a child, or in other words, whether the state of a child s residence had implemented SCHIP program by the time of data collection. The timing of SCHIP program implementation varies by state and time so the program availability also varies by state and time, though it is a child-level variable in this study. The effect of SCHIP availability for children on health coverage in the regression is identified from the variation in the timing of SCHIP implementation across states. It indicates on average by how much the dependent variable would change for a child if the child s residence state goes from pre SCHIP implementation to post SCHIP implementation. Alternatively, program eligibility can be used to indicate program implementation at the individual level. In fact, eligibility is more commonly used in SCHIP the literature and previous Medicaid expansions literature. In those studies the coefficient on the eligibility variable shows the average change in the dependent variable if a child is made eligible for the new program. This paper prefers program availability to eligibility because of two reasons discussed below. First, program eligibility suffers endogeneity problem, which has to be corrected by inefficient estimators like the instrumental approach used in recent SCHIP studies. Second, calculating individual child s eligibility status is difficult and may inevitably involve measurement errors. The difficulties come from the fact that states are using various income disregards in their income eligibility criteria and they change their criteria 7

8 over time. Current data sources do not collect detailed sources of income, which makes it difficult to assign eligibility. In contrast, program availability can be accurately determined and is less likely to be endogenous. SCHIP program is available to children in a state once it was implemented in that state. Program evolvement does not affect program availability. Therefore, program availability can be accurately assigned based on the time of survey interview for each child, which is available in most surveys. One potential problem with availability is that the timing of SCHIP implementation is a state-level policy so it may be endogenous. However, 45 out 51 states and D.C. started their programs within one and half years (from 4 th quarter of 1997 to 1 st quarter of 1998) from the enactment of SCHIP law in October Within 4 years since the passage of federal law, all states implemented SCHIP programs. The fact that most states launched their programs within a short period of time implies that variation in the timing of program implementation is less likely due to policy endogeneity and more likely due to administrative delays. In order to eliminate any potential endogeneity bias related to SCHIP availability, this study used multiple years national data, control for state Medicaid thresholds, and put in the model full sets of state ( State ) and quarter of year (Quarter ) dummies. State s Medicaid eligibility thresholds were set up before SCHIP program and they may reflect characteristics of a state in adopting a public program. State dummies control for unobserved time-invariant state level characteristics. Quarter of year dummies control for quarterly national trend in the dependent variables. Other control variables include child level and family level socio-demographic characteristics. Particularly, children s general health status levels are also included to control for potential endogeneity of selfselection due to health needs. q The model specification is described as follows: Insurance cist = c0 + α c1availabilityst + α c2states + α c3quarterq + α c4 α X + ε ist cist 8

9 where Insurance cist represents the five types of coverage this paper will address: SCHIP enrollment, Medicaid enrollment, public coverage, private coverage, and uninsurance. The subscripts has the following meaning: c indexes type of coverage, i indexes individual child, s indexes state of a child s residence, and t indexes time. The subscriptions attached to other variables in the regression have the same meaning as those to the dependent variable. Availability st, which indicates availability of SCHIP program to a child is the explanatory variable of primary interest. States and Quarter are groups q of state dummy and quarter of year dummy variables, respectively. Selected child level and family level characteristics are included in X ist. ε cist is the error term. Because more than half of the states had pre-existing state-run health programs for children and many of them were replaced by SCHIP, this study hypothesized that in these states SCHIP enrollment would increase quicker and bigger than states with no pre- SCHIP programs. To test this, when looking at the SCHIP enrollment effect, an interaction between SCHIP availability and a dummy variable indicating the existence of pre-existing health programs was added to the regression. A positive coefficient on the interaction term would indicate that SCHIP implementation had a bigger effect on SCHIP enrollment in states with pre-existing programs. This paper also tested whether timevariant state-level characteristics would affect the results by adding the full set of state and year interaction terms. Although the dependent variables are all dichotomous, linear probability models were estimated for the ease of interpretation and computation. Robust standard errors are calculated to solve potential heteroskatasticity in the error terms of linear probability models. The results are also adjusted by survey weights in the data Sub-populations of interest In addition to the full sample, the regression is applied to 6 sub-samples in order to estimate the effects of SCHIP implementation among sub-populations of interest. The 6 sub-samples are the following: 3 groups of states implementing 3 types of SCHIP 9

10 programs (M-SCHIP, S-SCHIP, and C-SCHIP) respectively, children from relatively lower-income families (family income less than 300% FPL), older children (15 to 18 years of age), youngest children in poverty (children under 6 years of age and with family income less than 100% FPL). The three types of SCHIP programs are quite different from each other, for example, M- SCHIP programs must follow all the strict rules set up for traditional Medicaid whereas S-SCHIP programs can design their program features quite freely and do not have to obey any Medicaid legislation. Thus, different type of programs may affect children s coverage differently. In addition, states choosing the same type of programs may share common but unobserved characteristics that are related to children s coverage. Therefore, it is necessary to test whether there is a differential effect among different type of programs. However, if we simply replace the overall program availability variable with availability for each of the three types of SCHIP program, these separate availability variables may be endogenous due to state policy endogeneity. To avoid endogeneity bias, this study groups states that chose the same type of program and analyze them separately. The fourth sub-population is children from relatively lower-income families. It is chosen because SCHIP targets children with family income ranges roughly from 100% FPL to 200% FPL. Since family income level may move up or down, this study will look at children whose family income falls into a wider range (<300% FPL) than the target levels. SCHIP programs increase the income eligibility thresholds more largely for older children compared to the younger ones, which is due to the lower eligibility levels for older children under Medicaid and the universal income threshold across age groups under SCHIP. Therefore, this study will look at the effects of SCHIP among the fifth sub-population separately: older children who are years of age. The sixth sub-population consists of youngest children under age 6 who are in poverty. This population is chosen specifically to test the spillover effect of SCHIP implementation on Medicaid enrollment. This group of children is ideal for testing the 10

11 spillover effect because they are only eligible for traditional Medicaid programs in any states according to Medicaid and SCHIP laws. Thus, they cannot be enrolled into SCHIP program. Therefore, the changes in Medicaid coverage among this group should be attributed only to the spillover effect of SCHIP program implementation on Medicaid enrollment. 4. Data The data source for this study is the National Health Interview Survey (NHIS) for the years 1995 through NHIS is the nation s primary source of general health information for the U.S. resident civilian non-institutionalized population. It collects information on health indicators, access to and use of health care, and health-related behaviors annually based on a stratified multistage sample design. NHIS is conducted by the National Center for Health Statistics (NCHS), a component of the Center for Disease Control and Prevention (CDC), U.S. Public Health Services, Department of Health and Human Services. NHIS has continuously collected data since The NHIS data are appropriate for this study because of their three features. First, the survey is of high quality overall and the insurance coverage variables are desirably reliable. The accurate measure of health coverage in NHIS results partly from the course of data editing, where one critical procedure is to check the names of respondents health plans and reassign respondents into appropriate enrollment category if their actual plan names indicate clearly a wrongly-reported coverage category. The back-edit process eased our worries that respondents may systematically misreport their public coverage as private coverage since many public programs such as Medicaid managed care programs and separate SCHIP programs have adopted many features of private plans, which may cause enrollee to believe they are under a private plan (Lo Sasso and Buchmueller, 2004). The second reason NHIS data are suitable for this study is because the data cover the span of SCHIP implementation nationwide. The first state started enrolling SCHIP children in June 1997 and the last one started in July Therefore, 1995 through 2002 NHIS data, which include information for every state and year, provide an opportunity to analyze SCHIP implementation as a natural experiment of children s 11

12 insurance. Last, NHIS is one of the first national surveys to collect SCHIP enrollment information as early as The availability of SCHIP coverage provides the opportunity to estimate SCHIP take-up at the national level. This study analyzed children under 19 years of age who participated in NHIS during The sample size is over 216,500 observations. Based on summary statistics adjusted for survey weights, less than one-fifth of the population is covered by some type of public insurance, mainly Medicaid program (17%) (see Table 1). SCHIP coverage, which is collected in the survey only after 1999, is quite limited with an average of 2.8 percent for the years 1999 through The majority of the sample-more than 67% have private coverage. The uninsured accounts for 14%. SCHIP program is available to less than 60 percent children across 8 years. Most of the children (56%) are from states implementing C-SCHIP; the rest is split evenly into states with M-SCHIP and states with S-SCHIP. Children s average age is about 9 and one-fifth of them are years old. White children account for more than 75%, blacks 15%, Asian 3%, and others less than 7%. In 12% of the families the highest educated person has only below high-school education; in 26% families a person who went beyond college is the one with highest education. 77% families have annual income above $20,000 and based on FPL, 58% families income are less than 300% FPL. More than half of the children reported as in excellent health; only 2% reported having fair or poor health. 5. Regression Results The effect of SCHIP implementation on health insurance coverage: overall and by program type The study results show positive effects of SCHIP on children s health coverage in its first five years across states. Among children to whom SCHIP became available, SCHIP implementation increased their SCHIP enrollment by more than 5 percent (see column 1 in Table 2). After including the interaction between availability and pre-schip program indicator, the effect is 4.4 percent among states with no pre-schip programs and 5.5 percent among those with pre-existing programs. SCHIP implementation increased 12

13 public coverage by 2.7 percent (p<.01). However, it decreased children s private coverage by 1.4 percent (p<.01). Overall, the uninsured rate of children changed slightly by 1.1 percent as a result of SCHIP implementation. The results do not change much if I control for the time-variant state characteristics by including the state and year interactions. After I break the full sample by SCHIP program type and estimate effects within states implementing the same type of program, we find similar patterns of SCHIP implementation on children s coverage as in the full sample. However, the magnitude of effect differs across groups of states. States that established separate insurance programs only had the largest increase on public coverage attributable to the new programs. Their programs increased children s public insurance by 5.0 percent (p<.01). Results also indicate that SCHIP program had the biggest crowed-out effect on private coverage states with S-CHIP programs: percent (p<.01) The effect of SCHIP implementation on health insurance coverage: lower-income children, older children, and youngest children in poverty Effects of SCHIP implementation among lower-income children and older children are stronger than that from the full sample. For children with family income less than 300 percent FPL SCHIP program has a bigger effect of 4.0 percentage points on public coverage than that for all children. The crowd-out effect of percentage points for this sub-sample leaves the reduction in overall uninsured rate by 1.5 percentage points. For older children where SCHIP has raised their income eligibility thresholds mostly, the public coverage grew by 2.9 percent, slightly larger than for all children. The result did not show crowed-out effect for this group of children. So the uninsured rate decreased by 2.1 percent (p<.05), which is the biggest reduction rate. Lastly, this study singled out children who are under 6 and in poverty to test spill-over of SCHIP on Medicaid since this sub-population should have been all covered by Medicaid and not eligible for SCHIP. The results supported our spill-over hypothesis. SCHIP implementation had no effect on SCHIP enrollment among this group of children but 13

14 increased their Medicaid enrollment by 5.5 percent (p<.01). However, the uninsured rate did not seem to be affected by SCHIP program, probably due to the big crowd-out effect for this sub-sample. 6. Discussion This study has produced a number of important policy-relevant and methodological findings. In the policy area, this study provided the first estimates of SCHIP enrollment rate, which can be seen as the effectiveness of mandated special efforts in SCHIP outreach and enrollment in all states. The expected spillover effect of SCHIP on Medicaid enrollment was found in this study. This study also confirmed the small but statistically significant effects of SCHIP on overall public coverage increase, the crowedout of private insurance, and the reduction of uninsured rate that were recently reported by other national estimates using different approaches. Moreover, this study found that SCHIP had greater effects on health coverage for older children, who were covered least in Medicaid programs and seemed to have a better chance to gain public coverage under SCHIP. In the methodological area, this study demonstrated that program availability can serve as a good indicator for program implementation in program evaluation. Using availability can avoid complicated eligibility assignment and because availability is likely to be exogenous at the individual level, efficient estimators can be used directly. This study also finds a simple way to test the spillover effect of SCHIP on Medicaid participation by focusing on the youngest and poorest sub-population that are not directly affected by SCHIP. References: AAP-website. State Insurance Program. Bollen, K. A., Guilkey, D. K., & Mroz, T. A. (1995). Binary Outcomes and Endogenous Explanatory Variables - Tests and Solutions with an Application to the Demand for Contraceptive Use in Tunisia. Demography, 32(1), Byck, G. R. (2000). A comparison of the socioeconomic and health status characteristics of uninsured, state Children's health insurance program-eligible children in the united states with those of other groups of insured children: implications for policy. Pediatrics, 106(1 Pt 1),

15 Census-website. Health Insurance Coverage: 1999(-2001). CMS-website. Fiscal Year 2002 Number of Children Ever Enrolled in SCHIP- Preliminary Data Summary. Cunningham, P. J., Hadley, J., & Reschovsky, J. (2002). The effects of SCHIP on children's health insurance coverage: early evidence from the community tracking study. Med Care Res Rev, 59(4), Currie, J., & Gruber, J. (1996). Health insurance eligibility, utilization of medical care, and child health. Quarterly Journal of Economics, 111, Currie, J., & Gruber, J. (1996a). Health insurance eligibility, utilization of medical care, and child health. Quarterly Journal of Economics, 111(2), Currie, J., & Gruber, J. (1996b). Saving babies: The efficacy and cost of recent changes in the Medicaid eligibility of pregnant women. Journal of Political Economy, 104(6), Cutler, D. M., & Jonathan, G. (1996). Does Public Insurance Crowd Out Private Insurance. The Quarterly Journal of Economics, 111, Dafny, L., & Gruber, J. (2000). Does Public Insurance Improve the Efficiency of Medical Care? Medicaid Expansions and Child Hospitalizations. NBER Working Paper No. w7555. Damiano, P. C., Willard, J. C., Momany, E. T., & Chowdhury, J. (2003). The impact of the Iowa S-SCHIP program on access, health status, and the family environment. Ambul Pediatr, 3(5), Dubay, L. H., I. Kenney, G.M. Haley, J.M. Leibovitz, H. (2002). Highlights from The Urban Institute s SCHIP Evaluation (Presentation). Urban Institute. Dubay, L. H., J. Kenney, G. (2002). Childen's eligibility for Medicaid and SCHIP: A view from The Urban Institute. Grossman, M. (1972). On the Concept of Health Capital and the Demand for Health. The Journal of Political Economy, 80(2), Gruber, J. (2000). Medicaid. NBER working paper series # Gruber, J. (2000). Medicaid. Harris, D. M. (2003). Contemporary Issues in Healthcare Law and Ethics. Joyce, T., & Racine, A. (2003). Have the vaccine for children and the State Children's Health Insuarnce Programs improved immunization coverage and delivery? Working paper. Kaestner, R., Joyce, T., & Racine, A. (1999). Does Publicly Provided Health Insurance Improve the Health of Low-Income Children in the United States? : National Bureau of Economic Research Working Paper: Lave, J. R., Keane, C. R., Lin, C. J., & Ricci, E. M. (2002). The impact of dental benefits on the utilization of dental services by low-income children in western Pennsylvania. Pediatr Dent, 24(3), Lave, J. R., Keane, C. R., Lin, C. J., Ricci, E. M., Amersbach, G., & LaVallee, C. P. (1998). Impact of a children's health insurance program on newly enrolled children. Jama, 279(22),

16 LoSasso, A. T., & Buchmueller, T. C. (2004). The effect of the State Children's Health Insurance Program on health insurance coverage. Journal of Health Economics, 23 (2004), Mann, C., Rowland, D., & Garfield, R. (2003). Historical overview of children's health care coverage. Future Child, 13(1), Mofidi, M., Slifkin, R., Freeman, V., & Silberman, P. (2002). The impact of a state children's health insurance program on access to dental care. J Am Dent Assoc, 133(6), ; quiz NCHS. (2000). Data File Documentation, National Health Interview survey, 1997 (machine readable data file and domumentation). Newhouse, J. P., Manning, W. G., Duan, N., Morris, C. N., Keeler, E. B., Leibowitz, A., Marquis, M. S., Rogers, W. H., Davies, A. R., Lohr, K. N., & et al. (1987). The findings of the Rand Health Insurance experiment--a response to Welch et al. Med Care, 25(2), Oral health in America: a report of the Surgeon General. (2000). J Calif Dent Assoc, 28(9), OSG-Website. National Call to Action to Promote Oral Health. Patel, K. (2001). Down and out in America: children and health care. J Health Soc Policy, 13(4), Pauly, M. V. (1986). Taxation, Health-Insurance, and Market Failure in the Medical Economy. Journal of Economic Literature, 24(2), Rosenbach, M. E., M. Czajka, J. Irvin, C. Coupe, W. Qrinn, B. (2001). Implementation of the state children's health insurance program: momentum is increasing after a modest start. Mathematica Policy Research. Slifkin, R. T., Freeman, V. A., & Silberman, P. (2002). Effect of the North Carolina State Children's Health Insurance Program on Beneficiary Access to Care. Arch Pediatr Adolesc Med, 156(12), Szilagyi, P. G., Holl, J. L., Rodewald, L. E., Shone, L. P., Zwanziger, J., Mukamel, D. B., Trafton, S., Dick, A. W., & Raubertas, R. F. (2000). Evaluation of children's health insurance: from New York State's CHild Health Plus to SCHIP. Pediatrics, 105(3 Suppl E), Thompson, T. G. (2002). The state children's health insurance program: A summary evaluation of states' early experience with SCHIP. Secretary of Health and Human Services. Ullman, F., & Hill, I. (2001). Eligibility under State Children's Health Insurance Programs. Am J Public Health, 91(9), Wooldridge, J. (2000). Introductory Econometrics: A Modern Approach. Wooldridge, J. (2003). Interim Evaluation Reprot: Congressionally Mandated Evaluation of the State Children's health Insurance Program. Mathematica Policy Research, Inc. and The Urban Institute. 16

17 Table1. Summary of variables in NHIS Variables Mean Health insurance SCHIP 0.02 Medicaid 0.20 Public 0.23 Private 0.61 Uninsured 0.16 SCHIP program SCHIP availability 0.60 M-SCHIP 0.19 S-SCHIP 0.19 C-SCHIP 0.61 Lower limit (Medicaid thresholds) (% FPL) Child characteristics Age Age age_ Male 0.51 White 0.71 Black 0.17 Asian 0.03 Other race 0.09 US born 0.94 General health Excellent 0.51 Very good 0.28 Good 0.18 Fair or poor 0.02 Family characteristics Family size 4.53 (range: 1-25) Income >$20k 0.73 Highest education No high school 0.16 High school 0.58 College 0.26 Sample size 216,517 17

18 Table2. The effect of SCHIP availability on health insurance: overall and by program type Full Sample M-SCHIP S-SCHIP C-SCHIP SCHIP coverage CHIP availability.051**.221** **.937**.016**.000 (.005) (.029) (.002) (.002) (.052) (.027) (.003) (.002) R Observations 192,186 36,934 38, ,442 Public Insurance CHIP availability.027**.030** **.050**.069**.015**.019* (.004) (.006) (.009) (.012) (.009) (.015) (.006) (.008) R Private Insurance CHIP availability -.014** -.025** * -.043** -.072** -.019* (.005) (.007) (.011) (.016) (.011) (.017) (.008) (.010) R Uninsured rate CHIP availability -.011** * (.004) (.006) (.009) (.012) (.009) (.013) (.007) (.008) R Observations 213, , ,445 State year interactions N Y N Y N Y N Y Notes: All regressions control for age, gender, race, born in U.S. or not, family size, highest education level of parent, family income over $20,000 or not, general health status, state Medicaid income thresholds, state, quarter of year dummies. The sample size is the same for regressions of public insurance, private insurance, and uninsured rate. Robust standard errors in parentheses * significant at 5%; ** significant at 1% 18

19 Table3. The effect of SCHIP availability on health insurance: lower-income groups and older children < 300% FPL > 14 years of age < 6 years & < 100% FPL SCHIP coverage CHIP availability.077**.317**.037**.136* (.008) (.040) (.010) (.065) (.000) (.000) R Observations 122,357 37,900 14,565 Medicaid coverage CHIP availability.055**.059 (.023) (.031) R Public insurance CHIP availability.040**.038**.029**.036** (.007) (.009) (.008) (.012) (.022) (.030) R Private insurance CHIP availability -.025** -.030** * (.007) (.010) (.011) (.016) (.017) (.024) R Uninsured rate CHIP availability -.015* * (.006) (.009) (.010) (.015) (.017) (.025) R Observations 135,710 41,978 16,354 State year interactions N Y N Y N Y Notes: All regressions control for age, gender, race, born in U.S. or not, family size, highest education level of parent, family income over $20,000 or not, general health status, state Medicaid income thresholds, state, quarter of year dummies, and survey weights. The sample size is the same for regressions of Medicaid, public, and private coverage and the uninsured rate. Robust standard errors in parentheses * significant at 5%; ** significant at 1% 19