Estimated Impacts of the Proposed Public Charge Rule on Immigrants and Medicaid
Appendix A: Methods
The findings presented in this brief are based on Kaiser Family Foundation analysis of Wave #2 the 2014 Survey of Income and Program Participation (SIPP). SIPP enables us to directly measure individuals’ immigration status at the time they entered the U.S. and health coverage and includes measures of health status. This approach differs from that used by DHS (described in detail in Appendix C), which was based on a combination of multiple administrative data sets and applied a number of broad assumptions. While SIPP has the advantage of directly measuring citizenship and immigration status, 2014 is the most recent year of data available. Because 2014 was a year of substantial transition for Medicaid due to the implementation of the Affordable Care Act, we also ran our analysis using the 2016 American Community Survey (ACS) to see if the time lag in data was affecting our results. The ACS analysis examined citizens versus non-citizens and led to very similar results.
We classified people as not having LPR status when originally entering the U.S. based on a SIPP question that asks, “What was [respondent’s] immigration status when he/she first moved to the United States?” In addition to measuring people who might adjust to LPR status in the future, who would be subject to a public charge determination (unless they fall into an exempt category), this measure includes noncitizens who have adjusted to LPR status since arriving into the U.S. It also includes nonimmigrants and undocumented immigrants who do not have a current pathway to adjust to LPR status. Our testing of different citizenship measures led to overall similar patterns. The 2014 SIPP shows 20 million noncitizens, including 8.7 million of whom originally entered the country without LPR status. It also shows an additional 18.8 million citizens living in a household with a noncitizen (10.1 million of whom live in a household with a noncitizen who entered the country without LPR status). Due to underreporting of noncitizens and legal immigration status in the SIPP, these estimates may reflect an undercount of the total noncitizen population and especially the undocumented population. Given this potential undercount—and that the group of noncitizens without LPR status includes some individuals who have since adjusted to LPR status as well as nonimmigrants and undocumented immigrants who do not have a pathway to adjust to LPR status— our analysis of characteristics that DHS could consider negative in public charge determinations focuses on shares rather than absolute numbers of affected individuals.
For the estimates of the share of noncitizens without LPR status living within the U.S. who have characteristics that DHS could weigh negatively in a public charge determination under the proposed rule, we used SIPP to measure age, household size, poverty and work status, insurance status, enrollment in public programs, education, English proficiency, and health status and classified each factor as positive or negative based on the proposed rule’s description of how DHS would consider the characteristic. DHS’ implementation and operationalization of its assessment of factors may differ from SIPP’s measurement of characteristics. In the preamble to the rule, DHS provided some data analysis of characteristics of the noncitizen population compared to citizens and discussed how certain characteristics correlate with enrollment in public benefit programs. They relied on older SIPP data (Wave 1 of the 2014 SIPP, which reflects 2013) and, in most cases, did not break out the non-LPR population in tables presented. Thus, their estimates are not directly comparable with ours.
In our analysis of household income, we use 125% of the Census poverty threshold, or $23,819 for a family of three in 2014. Census poverty thresholds are measured slightly differently than HHS poverty guidelines but lead to similar poverty levels for incomes of similar household size. In the proposed rule, DHS proposes a specific definition of a household to be used in the calculation of household income and notes that, while similar in concept to rules used by some government programs, their proposed definition varies in some ways. Thus, the final income cutoff for a particular family to meet the 125% of poverty rule as implemented may differ from our measurement or that used by other programs.
SIPP includes monthly measures of health insurance coverage. We coded individuals with at least one month of Medicaid or CHIP coverage during the 2014 calendar year as a Medicaid/CHIP recipients. Our analysis of 2014 SIPP finds 67.8 million total Medicaid/CHIP enrollees. This figure is low compared to current administrative estimates of 76 million, largely reflecting a well-documented “undercount” of Medicaid enrollment in survey data. Our analysis also finds that 14.1 million Medicaid/CHIP enrollees lived in a household with a noncitizen, 4.7 million of whom are noncitizen Medicaid enrollees. These data on Medicaid enrollees reflect both an undercount of noncitizens in the survey data (as noted above) as well as an overestimate of the share of noncitizens participating in Medicaid as it includes some who may be reporting emergency Medicaid or other state or local health assistance programs as Medicaid coverage.
For estimates of potential changes in coverage due to public charge policies, we present several scenarios using different disenrollment rates for Medicaid and CHIP. These disenrollment rates drew on previous research that showed decreased enrollment in Medicaid and CHIP among immigrant families after welfare reform.1 For example, Kaushal and Kaestner found that after new eligibility restrictions were implemented for recent immigrants as part of welfare reform, there was 25% disenrollment among children of foreign-born parents from Medicaid even though the majority of these children were not affected by the eligibility changes and remained eligible.2 Using this 25% disenrollment rate as a mid-range target, we assume a range of disenrollment rates from a low of 15% to a high of 35%. However, it remains uncertain what share of individuals may disenroll from Medicaid and CHIP in response to the proposed rule. Although the welfare reform experience is instructive of chilling effects among immigrant families broadly, it was associated with changes to program eligibility for immigrants. In contrast, this rule would change the potential consequences of participating in programs on an individual’s immigration status.
|Characteristics that DHS Could Consider in Public Charge Determinations by Citizenship Status, 2014|
|Potential Positive or Negative Factor?||Heavily Weighted?||Non-LPR Noncitizen||Total Noncitizens||Citizens|
|17 or younger||Negative||7%||9%||24%|
|18 to 61||Positive||89%||83%||57%|
|62 or older||Negative||5%||8%||19%|
|Less than Three People in Household||Positive||22%||21%||38%|
|Three or More People in Household||Negative||78%||79%||62%|
|No Physical or Mental Health Disability||Positive||96%||95%||87%|
|Physical or Mental Health Disability||Negative||4%||5%||13%|
|Excellent, Very Good, or Good Health||Positive||91%||91%||87%|
|Fair or Poor health||Negative||9%||9%||13%|
|Physical or Mental Health Disability and No Private Coverage||Negative||Y||3%||4%||7%|
|Less than 125% Federal Poverty Level (FPL)||Negative||34%||29%||18%|
|125% to less than 250% FPL||Positive||33%||32%||22%|
|250% FPL or more||Positive||Y||33%||38%||59%|
|No Private Coverage||Negative||59%||55%||30%|
|TANF or General Assistance||Negative||Y||4%||3%||4%|
|Low-Income Medicare beneficiary||Negative||Y||1%||2%||4%|
|Receiving Any Public Benefit||Negative||Y||26%||29%||27%|
|Not Receiving Any Public Benefit||Positive||74%||71%||73%|
|Employed (and age 18+)||Positive||62%||59%||47%|
|Not employed (and age 18+)||Negative||31%||32%||29%|
|Not employed and not a full time student||Negative||Y||27%||29%||27%|
|Has high school diploma or higher (and age 18+)||Positive||53%||56%||68%|
|No high school diploma (and age 18+)||Negative||40%||35%||8%|
|Does Not Have Limited English Proficiency||Positive||73%||76%||99%|
|Limited English proficiency||Negative||27%||24%||1%|
|Any Negative Factor||94%||94%||89%|
|Any Heavily Weighted Negative Factor||42%||47%||45%|
|Notes: For each individual subject to a determination, DHS would take into account the totality of his her circumstances and would retain discretion on how to weigh specific circumstances and factors; no single factor would govern a determination. How DHS would implement and operationalize its assessment of factors under the rule may differ from how SIPP measures characteristics.
Source: Kaiser Family Foundation analysis of 2014 Survey of Income and Program Participation data.
Appendix C: Summary of DHS’s Medicaid Estimates
Using administrative and survey data, DHS estimated that about 142,000 individuals would disenroll from Medicaid per year and that this would lead to a $1.1 billion annual decrease in federal Medicaid expenditures. As discussed below, DHS included a number of broad assumptions in its analysis. DHS does not account for a chilling effect in its estimates of disenrollment noting uncertainty related to estimating prospective disenrollment and that the proposed rule changes enrollment incentives versus eligibility policy. Instead, DHS assumes that all individuals directly affected by the public charge rule (i.e., those applying to adjust status) drop coverage but no disenrollment effects among their family members or among other noncitizen families. However, DHS recognizes that, “when eligibility rules change for public benefits programs there is evidence of a chilling effect that discourages immigrants from using public benefits programs for which they are still eligible.” It also notes that previous studies examining the effect of welfare reform changes showed enrollment reductions ranging from 21% to 54% due to this chilling effect, it does not account for a chilling effect in its estimates of disenrollment.
Number of Medicaid Beneficiaries Impacted
Appendix C Table 1 shows how DHS estimates the number of Medicaid beneficiaries impacted by the proposed rule:
- DHS starts with an estimate of average annual Medicaid enrollment of 64,281,954. They report that they draw this figure from a 5-year average annual calculation based on the most recent 5 years of administrative data available. However, when calculated based on the cited data, we find average annual Medicaid enrollment of 72,215,654 from January 2014-July 2018, the most recent month available. Even if DHS is using an earlier period that includes 2013 data (which would result in an artificially low estimate, since 2013 is before the Affordable Care Act Medicaid expansion), the average annual enrollment number we calculate is 68,701,856.
- DHS then estimates the number of households that may be receiving Medicaid by multiplying its estimate of total Medicaid recipients by the average household size nationwide. This calculation assumes that household size is the same across households with and without Medicaid enrollees.
- DHS then estimates the number of households with a noncitizen who may be receiving Medicaid by multiplying its household estimate by the share of the total population that is noncitizen. This calculation assumes that households with a Medicaid enrollee have the same proportion of noncitizens as the general population.
- Finally, DHS multiples this estimated number of households with a noncitizen who may be receiving Medicaid by the average size of households that include noncitizens to estimate that 5,685,422 Medicaid enrollees live in a household with a noncitizen. This calculation assumes that households with a noncitizen receiving Medicaid are the same size as households with a noncitizen who is not receiving Medicaid. As described above, our analysis of SIPP revealed a much larger number of Medicaid enrollees reside in a household with a noncitizen.
|Appendix C Table 1: DHS Methods to Estimate Number of Medicaid Enrollees Affected by the Proposed Rule|
|Measure||Data Point Used||Calculation||Calculation Method|
|Medicaid Average Total Number of Recipients||64,281,954||Based on 5-year average from Monthly Medicaid and CHIP Application, Eligibility Determination, and Enrollment Reports and Data. Each annual total calculated by averaging the monthly enrollment population over each year.|
|Households that May be Receiving Medicaid||24,349,225||64,281,954/2.64||Divided the number of people receiving Medicaid by the Census estimated average household size of 2.64 for the total population.|
|Households with at least One Noncitizen who may be receiving Medicaid||1,697,141||24,349,225 x 6.97%||Multiplied the estimated number of households receiving Medicaid by the share of the total U.S. population that is a noncitizen (6.97%)|
|Medicaid Recipients Who are Members of Households Including Non-Citizens||5,685,422||1,697,141 x 3.35||Multiplied the estimated number of households with at least one noncitizen receiving Medicaid by the average household size for those who are foreign-born using the Census estimate (3.35)|
Number of Medicaid Disenrollees
Appendix C Table 2 shows how DHS estimates the number of individuals that would disenroll from Medicaid under the proposed rule:
- DHS estimates the share of individuals that would disenroll from public programs by dividing the five-year annual average of the total number of people who adjusted to LPR status by the total noncitizen population, finding that 2.5% of noncitizens apply to adjust status each year.
- DHS applies this 2.5% disenrollment rate to its previously calculated estimate of Medicaid recipients who are members of households including noncitizens to estimate an annual enrollment decline of 142,136. This calculation assumes that everyone applying for adjustment of status within a year would disenroll. It does not account for any chilling effects that would lead to disenrollment among a broader group of individuals.
|Appendix C Table 2: DHS Methods to Estimate Number of Medicaid Disenrollees|
|Measure||Data Point Used||Calculation||Calculation Method|
|Anticipated share of Disenrollees||2.5%||544,246/22,214,947||Divided the number of immigrants that adjusted to LPR status annually by the total non-citizen population|
|Number of Medicaid Disenrollees||142,136||5,685,422 x 2.5%||Multiplied previous estimate of Medicaid recipients with a noncitizen in the household by the anticipated share of disenrollees (2.5%)|
Reductions in Medicaid Expenditures
Appendix C Table 3 shows how DHS estimates reductions in Medicaid expenditures associated with Medicaid disenrollment under the proposed rule:
- Using administrative data, DHS estimates total annual Medicaid spending of $477 billion. They then divide this average annual spending amount by their earlier estimate of average total annual enrollment to estimate average annual spending of $7,426 per enrollee. The Office of the Actuary (OACT) for the Centers for Medicare and Medicaid Services projects that average per enrollee Medicaid spending was approximately $7,200 in 2013, rising to $7,648 in 2017. These figures are a weighted average across all eligibility groups in Medicaid. There is wide variation in Medicaid spending per enrollee across eligibility groups, as DHS notes. Noncitizen Medicaid enrollees are more likely to be enrolled in low-cost enrollment groups such as adults without disabilities than the overall Medicaid population; thus, their average per enrollee spending is likely lower than the overall average for the Medicaid population.
- To estimate the reduction in Medicaid expenditures, DHS multiples their previous estimate of the anticipated annual enrollment decline (142,136) by their estimate of average per enrollee spending ($7,427). The estimate that DHS uses for average per enrollee spending is similar to that reported by (OACT) as well as other administrative data for total (federal and state) spending. Further, the total Medicaid payment amount used by DHS appears to include both federal and state spending. However, DHS indicates that their initial calculation just represents declines in federal expenditures and later inflates their overall estimated expenditure decreases across all programs by 50% to reflect estimated additional reductions in state expenditures to account for state matching funds.
|Appendix C Table 3: DHS Methods to Estimate Reductions in Medicaid Expenditures|
|Measure||Data Point Used||Calculation||Calculation Method|
|Average Annual Medicaid Payments||$477,395,691,240||5-year average based on Expenditure Reports from MBES/CBES|
|Average Annual Medicaid Payment per Person||$7,426.59||$477,395,691,240/ 64,281,954||Divided average annual Medicaid payments by previous estimate of average annual total number of Medicaid recipients|
|Anticipated Reduction in Medicaid Expenditures||$1.1 billion||142,136 x $7,426.59||Multiplied previous estimate of anticipated number of disenrollees by the average annual benefit per person|