Description

Katie Leinenbach
1 Summary Statistics
Mean Standard Deviation Minimum Maximum
2003 13750714.2 32587628.4 -128490 777987403
2004 15381278 36702311.7 1 820253000
2005 17450505.6 37806316.6 1 939134000
2006 21207631 47723258.8 -2667140 1074625000
2007 23890388.4 51816232.7 1 1203374820
2008 26897102.5 57173473.3 1 1361805561
2009 28192299.5 48084630.4 1 583975318
2010 30109277.6 71908357.9 1 2793923000
2011 35832356.5 77177143.5 -54283503 2059698345
2012 38093291.7 87040637.2 -1213043 1882619896
2013 39983569.1 81152473.9 -274185 1817294160
2014 37395915 89020259.1 -25850677 1993553072
2015 34126799.8 87414848.8 -33577 2042616756
2017 41812669.2 102877932 -27988 2753803541
Table 1: Summary Statistics – Uncompensated Care
Mean Standard Deviation Minimum Maximum
2003 197297261 340308984 -1757898 4722758791
2004 218309037 380538410 154394 5525730727
2005 238903830 420571526 1 6398553843
2006 263933948 465770546 -104189 7784094716
2007 287793491 509740587 63650 8577046126
2008 313230097 557736008 4 9293788259
2009 343868995 615365959 119236 9846464732
2010 368766204 662403468 306861 10185415748
2011 396589746 715044896 -27582223 10572291195
2012 420822735 768898836 -11799711 11865320139
2013 447328716 836590634 94880 12751708196
2014 481186418 908414182 6624 13376352387
2015 520701412 971208353 9368 14143533186
2017 607157402 1169439971 124513 16863431079
Table 2: Summary Statistics – Total Revenue
2 Mean Uncompensated Care by Hospital Type

Figure 1: Average Uncompensated Care by Hospital Type
3 Simple Difference in Differences

4 Event Study
Table 3: Event Study
Full Sample 2014 Treatment
Dependent Variable Uncompensated Care Uncompensated Care
treat interaction-16 49,504,308.8** (17,823,510.1)
treat interaction-15 54,297,490.6** (17,962,951.1)
treat interaction-14 53,972,839.4** (17,426,471.5)
treat interaction-13 53,921,814.7** (16,737,876.3)
treat interaction-12 43,427,264.4* (17,221,608.7)
treat interaction-11 46,581,656.5** (14,991,530.2) 1,918.6 (1.48e+10)
treat interaction-10 46,529,904.3** (14,671,766.5) 904,998.9 (1.48e+10)
treat interaction-9 47,368,317.2** (14,400,568.5) 3,217,375.3 (1.48e+10)
treat interaction-8 46,468,103.3** (14,369,832.7) 3,369,353.0 (1.48e+10)
treat interaction-7 43,535,329.8** (13,490,259.0) 225,069.9 (1.48e+10)
treat interaction-6 42,928,082.2** (12,688,978.4) -1,437,469.5 (1.48e+10)
treat interaction-5 44,004,937.7*** (12,002,916.9) -4,814,362.2 (1.48e+10)
treat interaction-4 40,881,696.2** (11,920,126.4) -5,137,377.0 (1.48e+10)
treat interaction-3 38,651,480.9*** (10,329,192.8) -7,770,279.2 (1.48e+10)
treat interaction-2 37,974,495.1*** (10,054,529.9) -8,598,780.0 (1.48e+10)
treat interaction-1 39,007,173.5*** (9,948,084.9) -6,745,971.2 (1.48e+10)
treat interaction0 28,270,440.0** (9,413,134.9) -19,693,724.3 (1.48e+10)
treat interaction1 39,425,624.9* (18,346,062.4) -24,415,450.1 (1.48e+10)
treat interaction2 3,825,466.6 (12,949,897.9) -33,650,974.9 (1.48e+10)
treat interaction3 8,281,466.7 (6,643,740.9) -38,469,545.0 (1.47e+10)
treat interaction4 7,283,604.1 (4,962,929.7) -35,520,167.7 (1.47e+10)
treat interaction5 -49,600,636.6 (1.47e+10)
Fixed-Effects: —————————— ————————
State Yes Yes
Year Yes Yes

S.E.: Clustered by: State by: State
Observations 41,941 36,641
R2 0.03314 0.08756
Within R2 0.00373 0.00984
Table 4: Event Study Estimates
5 Sun and Abraham

6 Sun and Abraham Graph

7 Callaway and Sant’Anna

8 Rambachan and Roth
Unfortunately I had an error in my code that I could not debug. I am confident the issue is in createSensitivityResults, but I could not isolate which component caused the error.
9 Discussion
I find that uncompensated care decreases after states adopt the Medicaid expansion. Comparing the estimates of the 2014 treatment group, the two-way fixed effect finds a decrease in uncompensated care. Looking further at the data, the event study finds that uncompensated care began decreasing 6 years prior to the state adopting Medicaid expansion, although the most significant change in the decrease happens the year after treatment. The Sun and Abraham estimate finds a reduction in uncompensated care starting three years before the adoption. It estimates the largest decrease in uncompensated care the year of adoption.
The different estimators all calculate a decrease in uncompensated care after the states adopt the Medicaid expansion, illustrating that the results are not driven by the type of estimator. This gives us confidence in our findings.
10 Reflection
Surprisingly, I found the components of the report, i.e. graphs, tables, and the latex document, to be the hardest part of this assignment. I have always made my graphs in Excel, and this assignment helped me remember why. I will need to get more familiar with how to change the axes and other elements of graphs like the Callaway and Sant’Anna graph, as these are not easily produced in Excel. I will need to improve my Latex skills as well to improve my report for the next assignment.