The effect of cadaveric kidney donations on living kidney donations: an instrumental variables approach.
As of March 19, 2012, there are 113,465 patients awaiting an organ transplant. Only 28,535 transplants were performed in the previous year. (1) The critical shortage of organs is the result of an artificial price ceiling instituted by the National Organ Transplant Act (NOTA) of 1984, which legislates financial compensation for organs to be unlawful. For this reason, the current supply of organs is generated by altruistic cadaveric and living donors (LDs).
These donors willfully supply organs without any expectation of compensation, but the decision to donate is not the same for both cadaveric and LDs. Unlike cadaveric donors (CDs) where the functional value of the organ to the donor is small, LDs must consider the potential long-term healthcare costs associated with donating an organ. These costs include an increase in mortality risk, morbidity risk, loss in quality of life, and lost wages generated by time away from work during recovery. Potential LDs and their recipients must weigh these costs versus alternative sources of organs (cadaveric organs and other LDs) and consider their own future demand for organs when deciding to donate.
Even so, the number of LDs has increased dramatically from 1,769 in 1988 to 5,955 in 2008, an increase of 237% (Figure 1). Over the same period, the percentage of blood-related (or biological) donors has decreased from 93% to 60% of all LDs. Several factors can account for these changes: introduction of tax incentives for organ donation, compatibility websites, relaxing donor match requirements, and the implied responsiveness of potential LDs to increases in the number of persons on the waiting list for deceased donors (an increase from 40,000 to 52,000 from 2000 to 2009). (2)
The topic of organ donation has received considerable attention in both the academic literature and the press, but there is limited empirical evidence identifying potential causes for the increase in LD rates, the trade-offs between living and cadaveric kidney donations, and the change in LD composition from primarily blood-related donors to an increase in non-blood-related donors. (3) Two concurrent studies address the substitutability between living and cadaveric donations. Howard (2011) utilized geographic variation in waiting times for organ transplants to estimate the substitution effect. The author finds that a decrease of five kidneys from cadaveric donations is correlated with one additional living kidney donation. Sweeney (2010) exploits a discontinuity in panel reactive antibody (PRA) (4) blood levels within a regression discontinuity framework and finds a 10% increase in the likelihood of receiving a cadaveric donation decreases the likelihood of receiving a living kidney donation by 2-3%.
We find similar results as these studies, but we utilize a completely different identification method. In this article, we propose using variation in traffic safety laws between states as an instrumental variable to identify the substitution patterns between living and cadaveric kidney donations. A benefit of our approach is that we can analyze how obesity influences the supply of donations, the demand for transplantation, and the use of living versus cadaveric donations. We also analyze substitutability across classes of LDs, which helps to establish the plausibility of our results and the mechanisms underlying the substitution effects.
With respect to motor vehicle safety laws, we study how changes in seat belt, helmet, and speed limit laws indirectly affect the number of living donations by shifting the supply of cadaveric organ donations. Previous studies have considered how changes in these safety laws would affect motor vehicle fatalities and cadaveric donations. According to the Organ Procurement and Transplantation Network (OPTN), roughly 16% of cadaveric donations are the result of a motor vehicle accident (MVA). Therefore, policy changes intended to affect the number of motor vehicle fatalities indirectly affects cadaveric organ donations. Ashenfelter and Greenstone (2004) estimate that increasing the speed limit from 55 to 65 mph on rural highways increased fatality rates by 35%. Dee (2009) finds a 27% reduction in fatalities among motorcyclists when states require helmet use. Dickert-Conlin, Elder, and Moore (2011) build on this result by linking helmet laws to organ donations. The authors find that a national repeal of helmet laws would lead to a 1% increase in the supply of cadaveric organ donations. These studies provide support for a possible link between motor vehicle safety laws and cadaveric donations, but they do not consider how these changes in cadaveric donations may affect the supply of living donations.
The supply of living donations may also be influenced by changes in demand. We consider both state-level obesity rates and prevalence rates for end-stage renal diseases (ESRD) as potential demand shifters. The use of obesity rates to estimate shifts in the demand for organs is motivated by recent medical research connecting obesity and renal disease. The primary risk factor for type II diabetes, a leading cause of renal failure, is obesity. Hsu et al. (2006) find that obese patients are six to seven times more likely to develop renal failure than individuals of normal weight. In Sweden, Ejerblad et al. (2006) estimate that being overweight at age 20 tripled the odds of chronic kidney failure. These studies highlight the potential link between obesity/ESRD and the demand for kidney transplants.
A secondary link between obesity and the supply of living donations is generated by the allocation system used to determine who receives organs from cadaveric donations. Individuals who are morbidly obese are not considered "good candidates" for organ transplants. These patients receive less priority to access the supply of cadaveric donations. In these cases, LDs can circumvent the allocation system and donate directly to these individuals.
A potential third relationship of obesity with the supply of kidney donations is via the supply of cadaveric donations. Selck, Deb, and Grossman (2008) find that a higher body mass index (BMI) is also associated with a modest increase in organ yield among CDs, but this relationship is quadratic in nature. Consequently, large values of BMI or obesity can lead to fewer cadaveric donations as there would be fewer viable organ transplant candidates.
The substitution effect between living and cadaveric kidney donations is estimated using two-stage least squares where variations in motor vehicle safety laws serve as instruments for the supply of cadaveric kidney donations. (5) In order for these variables to be valid instruments, they must be independent of unobserved factors affecting both cadaveric and living donations. For example, the passage of traffic safety laws must be independent of both unobserved levels of altruism and the objectives of organ transplant physicians within a state.
Our primary findings are (1) a decrease in the supply of cadaveric donations leads to an increase in the use of LD transplantation. On average, living donations decrease by one kidney for a corresponding increase in cadaveric donations of 2-5 units; (2)a positive association between obesity and living donations does exist; (3) blood-related donors are non-responsive to changes in the supply of cadaveric donations, but non-blood-related donors are; (4) within the category of non-blood-related donors, anonymous donors are less responsive to shifts in cadaveric donations than donors who are known to the organ recipient.
In the following section, we provide a brief overview concerning changes in motor vehicle safety laws in the United States. Section III describes the data gathered from the OPTN as well as changes in state demographics used to conduct the analysis. Section IV presents the empirical model and results. Finally, Section V concludes and suggests paths for future research.
II. A BRIEF HISTORY OF MOTOR VEHICLE SAFETY LAWS
A. Speed Limits
Prior to 1974, speed limits were determined by states and local governments, but because of the energy crisis that began in the 1970s, politicians were motivated to set nationwide speed limits. The Emergency Highway Energy Conservation Act (EHECA) of 1974 set speed limits to a maximum of 55 mph, which was lower than any existing speed limit within the 50 states. In 1987, Congress modified the EHECA by allowing states to set speed limits not to exceed 65 mph on rural highways. Of the 50 states, 41 states did increase speed limits on rural highways to 65 mph. (6) Table 1 reports current maximum speed limits by state. (7)
In 1995, Congress eliminated the national speed limit restriction. States are now free to choose the maximum speed on both rural and urban highways. By 1997, most states had adopted speed limits of 70 mph or greater on rural highways, but only half set limits above the previous national speed limit of 55 mph on urban highways. As of June 2009, 33 states have speeding limits in excess of 65 mph on rural highways with the highest speed limit of 80 mph in Texas and Utah. Only 14 states have maintained the national speed limit of 55 mph on urban highways and the lowest speed limit of 50 mph is found in Hawaii.
According to the National Highway Traffic and Safety Administration, speeding is associated with roughly one-third of all fatal crashes. In 2007, 13,040 people died in speed-related crashes, yet the evidence of speed limits having a significant effect on fatalities is mixed. Ashenfelter and Greenstone (2004) report that fatalities per mile decreased by 15% immediately following the passage of the 1974 EHECA. The National Research Council publication Highway Statistics (1994) reports that the nationwide speed restriction prevented 3,000-5,000 traffic fatalities annually. Fatalities were thought to increase when the restriction was lifted on rural highways in 1987. However, Lave (1994) finds the increase in speed limits may have saved 3,113 lives between 1987 and 1988. Furthermore, Moore (1999) finds the 1995 repeal of the nationwide speed limit on urban highways led to 66,000 fewer road injuries in 1997 than in 1995.
B. Helmet Laws
In 1967, the federal government encouraged states to adopt universal helmet laws by making the issue a stipulation to qualify for federal safety programs and highway construction funds. The universal helmet law requires all riders, regardless of age, to wear a helmet. In the early 1970s, most states had adopted a helmet law with the exception of Michigan, which repealed the law in 1968. In 1976, Congress removed the mandate requiring states to have helmet laws in order to receive federal highway funds. By 1980, most states had repealed the helmet law or stipulated partial coverage laws, which would only apply to young riders (18 years old and younger). Currently, all but three states (Illinois, Iowa, and New Hampshire) have helmet laws. Since 1997, Arkansas, Florida, Kentucky, Pennsylvania, and Texas have moved from universal to partial helmet laws. In 2004, Louisiana moved from a partial to a universal helmet law. Table 1 summarizes the history of helmet laws by state. The changes in helmet law adoption and type between states have created a natural experiment used by researchers to study the effects of these laws on helmet use, fatalities, and demand (Dee 2009; Liu et al. 2008; Sass and Zimmerman 2000).
C. Seat Belt Laws
In 1984, New York was the first state to require front seat occupants to wear a seat belt. As seen in Table 1, mandatory belt laws are currently present in every state with the exception of New Hampshire. (8) In 30 states, the belt law is a primary law implying that police officers may stop a vehicle solely because the occupants are not wearing seat belts. In the remaining states, where belt laws are considered secondary laws, a police officer must have an alternative reason to stop a vehicle prior to giving a seat belt citation. Since 1993, 23 states have moved from a secondary belt law to a primary belt law. The average fine is $32 with five states issuing the lowest fine of $10 and Texas issuing the highest fine of $200. Cohen and Einav (2003) use data from 1983 to 1997 for all 50 states and the District of Columbia to estimate the effectiveness of seat belt laws to reduce fatalities. The authors find that a 10% increase in seat belt usage decreases automobile fatalities by about 1.35% or 500 lives annually.
The 1984 NOTA established the OPTN, a network of 57 Organ Procurement Organizations (OPO) located in 37 states and the District of Columbia. These OPO's are given monopoly rights to receive and transplant organs within a designated service area. The OPTN has maintained organ donation data collected from each OPO since 1988. (9) Both individual level and OPO level data are available, but only the OPO level data have geographic information on organ transplants. We use the OPO level data
to link state traffic laws with OPO locations. This article uses transplanted kidney donation counts from 1988 to 2008 with respect to cadaveric donations and living donations. (10) For each year, we aggregate kidney donation counts to the state level for 46 states (data are missing for Alaska, Montana, Idaho, and Wyoming) resulting in a total of 966 state-year observations. (11) LDs are disaggregated into blood-related and non-blood-related donors.
As illustrated in Figure 1, the number of LDs has increased dramatically over the last 20 years. Perhaps more interestingly, the percent of LDs who are not blood relatives to the organ recipient has also increased roughly from 7% in 1988 to 40% in 2008. One reason for the increase is a change in the surgical procedure used to harvest the kidney. Prior to the introduction of laparoscopic surgery, a 4-7 in. incision was needed to retrieve the kidney, which would significantly increase the pain and recovery time associated with the procedure. According to Schweitzer et al. (2000), the use of laparoscopic surgery has decreased hospital recovery time from 4 1/2 days to 3 days and has allowed donors to return to work 27 days faster. Clearly, this medical advancement decreases the non-pecuniary cost of donating an organ for a LD. These cost reductions are also a source of savings to organ recipients (i.e., a decrease in lost wages).
A second reason to observe an increase in non-related donors is the relaxing of match requirements. Previously, organ transplants would only occur between blood-related partners (typically siblings), but today transplants can occur between non-related partners of the same blood type. This policy change greatly expands the set of qualified donors for a given recipient.
While NOTA stipulates that it is unlawful to provide financial compensation to the donor, the ordinance does allow for payment to cover expenses directly incurred by the organ donor for the purposes of the donation (i.e., travel costs and lost wages). (12) These non-medical costs are not covered by insurance. Instead, these costs are paid out-of-pocket. By reducing time away from work, compensation associated with lost wages is also decreased. This incentive can potentially establish a link between non-blood-related donors and disposable income. The U.S. Bureau of Economic Analysis provides data on State Disposable Income Per Capita. The income data are inflation-adjusted to the base year of 2000. We use these data to test if the increase in non-blood-related donors is affected by higher levels of disposable income.
For each state, we collect information on helmet, seat belt, and speed limits laws from the Insurance Institute of Highway Safety. State population, health insurance coverage, race/ethnicity, gender, and age distribution data are collected from the U.S. Census. Marital status and obesity data are collected from the Behavioral Risk Factor Surveillance System. (13,14)
In an effort to control the effects of religion on organ donations, we collect data on church membership for Catholicism, Judaism, other religions, and atheists from the Association of Religion Data Archives: Churches and Church membership in the U.S. Decennial survey (1980, 1990, 2000). (15) Although no organized religion in the western hemisphere officially discourages organ donations, some religious groups may hold private beliefs that either encourage or discourage participation in organ transplants. (16)
We gather race and gender information from the Compressed Mortality files 1988-2006 for non-medical injury deaths to control the effects of demographic differences between the living population and the recently deceased. Non-medical injury deaths, such as suicide, are a source of cadaveric donations, as organs are typically not damaged as a result of medication or disease.
Descriptive statistics for the primary variables of interest (organ donations, traffic safety laws, and obesity rates) are found in Table 2. Over the sample period, the number of CDs outpaced the number of LDs by a ratio of 5-4 and the total number of cadaveric donations is nearly twice the amount of living donations, but the number of LDs did exceed the number of CDs from 2000 to 2005. Among LDs, 73% are from blood-related donors, 57% are female, 45% are between the ages of 35-49, and 36% are between the ages of 18-34. Among CDs, 60% are male, 10% are 11-17 years old, 30% are 18-34 years old, 26% are 35-50 years old, and 28% are greater than 50 years old.
Obesity and ESRD prevalence rates are collected from the Centers for Disease Control and Prevention. Obesity rates in the United States have risen steadily since the 1990s. Over the same time period, living donations increased by 197% and cadaveric donations increased by 69%. By definition, an obese individual has a BMI of 30 or greater. The average percentage of obese individuals by state in the sample is 18%. We find that 81% of the states in our sample report obesity rates between 10% and 24%.
Next, we analyze data on traffic safety laws. Large variations in helmet and seat belt laws between states are observed. This is expected due to many adoptions, repeals, and modifications to these laws during the period of our sample. Approximately 47% of the sample has enacted a full helmet law and 42% has enacted partial helmet laws. Seat belt laws are slightly less balanced with 60% of the sample enacting a secondary belt law and 30% of the sample enacting a primary seat belt law. However, speed limits display little variation as 56% of the states report rural speed limits equal to 65 mph and 57% report urban speed limits of 55 mph or less. The most popular rural speed limits are 65 and 70 mph. The most popular urban speed limits are 55 and 65 mph. The average speed limit is 59.6 mph on urban highways and 66.6 mph on rural highways. As observed in Table 1, maximum speed limits have regional tendencies such as being lower in the Northeastern region and higher in the Western region. The majority of speed limit changes occurred shortly after the 1987 modification to the EHECA and the 1995 repeal. (17)
In addition to the variables of interest, a secondary set of control variables are collected. The descriptive statistics for these variables are found in Table 3. These control variables can be grouped into the following categories: race/ethnicity, gender, age, marital status, religion, health insurance, and income. The demographic variables (gender, age, race/ethnicity, and marital status) capture unobserved effects that are specific to these sub-populations. The race and age variables are further disaggregated into two sub-samples, the living population and the recently deceased population. This division allows differences in race and age composition between the living and the deceased to affect the supply of cadaveric kidney donations separately.
The health insurance variables may affect organ donations in two ways. First, if a greater percentage of the population has health insurance, then an individual's health is less likely to deteriorate such that an organ transplant is necessary. Second, in the event an organ transplant is necessary, individuals with alternative forms of health insurance can cover the costs of transplants not paid by Medicare. (18)
IV. MODELAND RESULTS
Consider the following conceptual model of altruistic LDs. A LD's utility function takes the form of
(1) U(H, a, c, p) = max [H +a p [U(CD).sub.donee], H - c + a [U(LD).sub.donee]
where H represents the donor's Health; c, the cost to health for donating an organ; p, the probability the organ recipient receives an organ from a CD; and a, the level of altruism the donor has for the organ recipient. In this model, a potential organ donor is assumed to fully internalize the recipient's welfare into their own utility function. According to The National Kidney Foundation's Fact Sheet, patients receiving a kidney from a LD have a l-year survival rate of 97.9% versus a 94.4% survival rate when a kidney from a CD is used. (19) For this reason, we assume an organ recipient's utility is higher when receiving an organ from a LD, U(LD) > U(CD). In relative utility, the potential donor chooses not to donate if c > a[U(LD).sub.donee] - p [U(CD).sub.donee] or the cost of donation is greater than the relative expected benefit of supplying an organ from a LD. A potential donor faces several costs including pain associated with the procedure, lost wages due to time away from work for the procedure, and potentially lowers lifetime income due to a lower lifetime level of health. Advancements in medicine can decrease these costs over time leading to an increase in the supply of LDs, ceteris paribus.
Altruism plays an important role in this model. For our purposes, altruism is captured by allowing the donor's utility to be dependent upon the organ recipient's utility. The magnitude of a in the donor's utility becomes larger the more familiar a donor is with the organ recipient. Altruism creates a positive supply of organs at a price of zero. (20) The probability of receiving a cadaveric organ, p, is dependent on the size of the waiting list and the level of altruism in society associated with donating organs at death. As the waiting list increases or the expected level of future donations decreases, p decreases, thereby encouraging more living donations. If the level of altruism among CDs rises in society, p increases, causing LD rates to marginally decrease. However, if altruism rises among LDs, then the number of LDs also rises. In an effort to increase altruism among potential LDs, websites such as livingdonors.org and matchingdonors.com provide information pertaining to patients in need of an organ. These websites can increase LD rates by personalizing organ matches, thereby increasing altruism on the part of the donor (a increases).
In the conceptual model, the organ donor is the end decision maker, but in reality several agents may affect the decision. For example, the transplant physician serves as the agent for the organ recipient, who is the principal. A principal-agent problem arises as the physician must not only consider the health of the patient but also the health of other patients on the waiting list. Therefore, the physician may determine that the patient is not of good health and thus receives less priority to the organ (Mandelbrot 2007). The principal may also refuse a living kidney donation because she does not wish to impose a cost to a donor with whom she is familiar (Pradel et al. 2003).
The duality of altruism in the organs market plays an important role when estimating the substitution effect between living and cadaveric donations. Higher levels of altruism in a state will increase the number of cadaveric and living donations regardless of the substitution effects. Unfortunately, the level of altruism is an unobserved variable and would cause least squares estimates of the substitution to be biased. Second, this relationship may suffer from reverse causality in that living donations are affected by waiting times. A patient's waiting time is her individual price for the kidney. We must instrument for donations in the same fashion as we instrument for price when estimating demand equations. An instrumental variables approach is used to remove the endogeneity bias caused by unobserved changes in altruism and expected waiting times.
The substitution effect of cadaveric donations for living donations is estimated using 2SLS where traffic safety laws serve as instruments for the supply of cadaveric organs. Traffic safety laws function as exogenous shifters to the supply of cadaveric donations. These instruments are independent of altruism among organ donors.
The primary equation of interest is the LD equation
(2) [LD.sub.st] = a[CD.sub.st] + [theta][Obesity.sub.st] + [X.sub.st] [beta] + [[delta].sub.s] + [[epsilon].sub.t] + [v.sub.st]
where s indexes state, t indexes year, [LD.sub.st] measures living kidney donations per 100,000 individuals in the state, [CD.sub.st] are cadaveric donations per 100,000 individuals in the state, [X.sub.st] is a matrix of socio-demographic state variables (including race/ethnicity, health insurance coverage, age distribution, gender, marital status, disposable income per capita, population, and religion), and Obesity measures the percentage of the state population that is obese (BMI [greater or equal t] 30).
The error term captures all unobserved changes within a state and across time associated with living donations. These unobserved measures may include variation in the causes of hospital deaths, changes in medical technology, changes in physician/hospital practice in regards to organ transplants, and differences in altruism. We adopt a fixed effect specification where [[delta].sub.s] is a state-specific fixed effect, [[epsilon].sub.t,] a year-specific fixed effect, and [[upsilon].sub.st], an idiosyncratic state-year shock with mean zero and finite variance. (21) Given the monopolistic and regulated nature of the OPO's, this specification controls for unobserved changes in nationwide organ procurement policy and procedure via the year dummy variables.
A variable missing from our specification is the size of the organ waiting list. Naturally, potential LDs would consider the current shortage of kidneys in their designated organ market before deciding to donate, but this variable is econometrically endogenous. In a typical goods market, both price and quantity are determined by the intersection of the demand and supply curve. These variables are determined within the model; thus, both are econometrically endogenous. In the organs market, price is restricted to zero. The endogenous variables determined by the model become the quantity demanded and the quantity supplied when price is equal to zero. The size of the waiting list is the difference between these two endogenous variables; therefore, it too is endogenous. For this reason, we forgo using the size of the organ waiting list as an explanatory variable in the living organ donations equation.
Table 4 contains ordinary least squares estimates of the parameters [theta], [beta], and [alpha] under different model specifications. The first column in Table 4 provides estimates including variables found in all years and states of the sample. The second column includes marital status, and the third column includes a quadratic term for the obesity variable. Across the three specifications, we find a positive point estimate for the effect of cadaveric donations on living donations, but these estimates are not found to be statistically different from zero.
Similarly, the obesity coefficients are all positive, but indistinguishable from zero in the first two columns. In column 3, we follow Selck, Deb, and Grossman (2008) and consider nonlinear effects of obesity by including a quadratic obesity term. This specification does yield a statistically significant effect of obesity on living kidney donations at the 1% level. Living donations are found to increase as the obesity percentage increases, but at a decreasing rate. The association is maximized when the obesity percentage is equal to 20%. A concern of using obesity in this analysis stems from the potential that obesity is correlated with unobserved state-specific health shocks. We attempt to minimize this concern by including prevalence rates for ESRD as a secondary measure of health. This variable is disaggregated into patients with diabetes and those without. Both variables are found to be significant at the 1% level where an increase among non-diabetic (diabetic) prevalence of ESRD is found to increase (decrease) living kidney donations per 100,000 individuals.
Other notable observations are that the divorce percentage and Medicaid percentage are found to decrease LD rates. Medicaid is statistically significant at the 1% level in all specifications, and divorce is significant at the 10% level in the last specification. A 1% increase in divorce (Medicaid enrollment) decreases living donation rates by -0.03 (-0.05). The age variables are all statistically significant at the 1% level. The omitted category is the 75 and older age group. Therefore, as the percentage of the population under 74 years of age increases the number of living donations decreases. Finally, we do not observe a link between living organ donations and disposable income. All point estimates of income per capita are found to be positive, but none are statistically significant at conventional levels.
The previous regressions treat the supply of cadaveric donations as exogenous, but the positive correlations between living and cadaveric donations observed in these regressions may be the result of both variables being positively correlated with unobserved altruism. (22) As a result, these coefficients may be biased. A secondary potential source of bias is an omitted variable bias due to not controlling for waiting times (Howard 2011). When deciding whether to pursue living donation, transplant candidates look at waiting times. Waiting times in period t are a function of the number of cadaveric donations recovered in period t, as well as the number recovered in t - 1 and t - 2, etc. The organs themselves are not durable, but the individuals on the waiting list are. For the same reason, the number of cadaveric donations recovered in period t will affect living donation rates in future years. The fixed effects model assumes, incorrectly, that there is instantaneous adjustment, and that an increase in cadaveric donations in 1 year affects living donation rates in the same year but not in subsequent years. Therefore, a valid instrumental variable must cause a more permanent shift in the supply of cadaveric organs to identify changes in living donation rates not caused by intertemporal transitory shocks to the waiting list or unobserved levels of altruism.
We utilize an instrumental variables approach to remove the bias. We propose the following specification for the supply of cadaveric donations per 100,000 individuals
(3) [CD.sub.st] = [delta]([law.sub.st]) + [lambda] [W.sub.st] + [[delta].sub.s] + [[epsilon].sub.t] + [v.sub.st]
where [CD.sub.st] represents cadaveric kidney donations per 100,000 individuals in the state, [law.sub.st], a matrix of dummy variables capturing changes in traffic safety laws (full and partial helmet laws, primary and secondary seat belt laws, rural highway speed limits, and urban highway speed limits), and [W.sub.st], a matrix containing the age distribution of recently deceased individuals in a state who died as a result of a non-medical injury. These individuals are the most likely candidates to be cadaveric organ donors as their organs, unlike those of individuals dying from diseases such as cancer, are not typically damaged.
To gauge the strength of these instruments, we first estimate Equation (3) using four different specifications. Each specification progressively adds more instrumental variables in the following order: helmet laws, speed limits, seat belt laws, and injured death demographics. For each specification, the incremental F-stat is calculated for testing the validity of the instruments. These estimates are found in the lower portion of Table 7. The proposed instruments are found to be strongly correlated with the supply of cadaveric donations as indicated by the F-stat ranging from 4.94 to 17.49. (23)
We report the point estimates for the parameters of Equation (3) in Table 5. Both helmet laws are found to reduce the supply of cadaveric donations by a range of (-0.73, -1.34) donations per 100,000 in models OLS 4-OLS 7. A universal helmet law has a larger effect than a helmet law that targets young drivers. These results are consistent with the findings of Dee (2009) and Dickert-Conlin, Elder, and Moore (2011). Conversely, the supply of cadaveric donations increases as the speed limit on rural highways increases. A 1% increase in the speed limit increases the number of cadaveric donations, on average, by 0.9% donations per 100,000 and is statistically significant at the 10% level. Urban speed limits and seat belt laws are not found to have an effect on cadaveric donations conditional on helmet laws and rural speed limits.
We further test the validity of our instruments by providing a falsification test. The use of traffic safety laws as instruments should only have an effect on cadaveric donations generated from MVAs. The OPTN does collect data on the mechanism of death for each cadaveric donation beginning in 1995. Therefore, we collect data on the number of MVA donations and Non-MVA donations by state from 1995 to 2008. This subset is substantially smaller than the full sample (a reduction of 7 years) and limits the amount of variation in the traffic safety laws. In turn, we can only consider changes in speed limits and helmet laws as there is little within-state variation of seat belt laws. (24) We do consider the effect of a primary seat belt law where the omitted category is no seat belt law or a secondary seat belt law. We again estimate Equation (3) separating cadaveric donations into two groups: MVA donations per 100,000 and Non-MVA donations per 100,000. The estimates are found in Table 6.
The traffic safety variables are only statistically significant when considering MVA cadaveric donations. None of the traffic safety variables are statistically significant when predicting Non-MVA cadaveric donations. Both helmet laws are significant at least at the 5% level when all controls are used and are consistently negative across all the model specifications. Rural speed limits are found to have a statistically significant effect on MVA donations at the 5% level. A 10% increase in rural speed limits is found to decrease the number of cadaveric donations from MVA by -0.05 donations per 100,000. However, the opposite relationship is found with respect to urban speed limits. A 10% increase in urban speed limits increases the donation rate by 0.03 donations per 100,000 and this relationship is also significant at the 5% level. From these results, we can infer that the instruments are affecting the supply of cadaveric donations through MVA donations and not some other unobserved mechanism.
There are several studies that find MVA fatality rates to be higher on rural highways than on urban highways (Clark and Cushing 1999, 2002; Muelleman and Mueller 1996; Zlatoper 1989). As previously stated, both Lave (1994) and Moore (1999) find that increasing speed limits could potentially save lives. Therefore, it is not surprising that we too find mixed results. An increase in the urban speed limit may encourage individuals to use these highways instead of rural alternatives, thereby decreasing the number of fatal accidents via a substitution effect. However, there are several reasons for there to be heterogeneous effects between urban and rural highways. Clark and Cushing (1999, 2002) find that rural highway travelers are further away from trauma centers increasing the likelihood of death in the case of an accident. Muelleman and Mueller (1996) find that rural residents travel over more miles than urban travelers, thus increasing the potential for an accident. Finally, the construction of urban highways often prevents vehicles from leaving the road when an accident occurs, but rural highways do not have such safety measures and more harm is done to the vehicles once they leave the road.
When we evaluate the speed limit variables separately, as in specifications 6 and 7 of Table 6, urban speed limits are still found to be significant and robust over the different specifications, but the rural speed limit variable is no longer significant. These differences in point estimates from the more general model are caused by the time period considered by the MVA subset. These data are taken after the 1995 national repeal. Urban speed limits increase over this time period by an average of 1% annually versus 0.5% annually for rural highways. In the pre-1995 sample, only rural speed limits experienced changes. When the whole sample is used, as shown in Table 5, increases in rural speed limits lead to more donations, but urban speed limits do not have a significant effect when both variables are included in the specification.
Given the strength of the proposed instruments, we proceed by estimating the LD equation using 2SLS. The 2SLS estimates are found in Table 7. Each specification adds additional instrumental variables to the first stage regressions as is done in the previous section. The under-identification test proposed by Kleibergen and Paap (2006), which tests the rank condition of the instruments, finds that the first four specifications pass this test at the 1% level and the last specification passes the test at the 10% level. The first two specifications pass the weak instrument test proposed by Stock and Yogo (2005) at the 10% level and the third specification passes at the 20% level. However, only he second specification fails to reject the null hypothesis of over-identification at conventional levels using the Hansen J statistic. The first and third specifications reject the null hypothesis at the 5% level, but not the 1% level.
In all five specifications, we find a negative relationship between living and cadaveric donations. Living donations are found to decrease between -0.2 and -0.6 donations per 100,000 for one additional cadaveric donation per 100,000. This relationship is statistically significant in the first three specifications at the 1% and significant at least at the 10% level in the remaining models. (25) These estimates are in full agreement with those found in Howard (2011).
From these estimates, we infer the elasticity of substitution to be between -0.36 and -0.91 when evaluated at the mean of living and cadaveric donations per 100,000. (26,27) This result is important when considering measures to increase the supply of cadaveric donations. The elasticity estimates indicate that the level of LDs would decrease when an expansionary policy (such as an opt-out policy for donations) is adopted, but the overall effect is an increase in the supply of kidney donations.
Obesity is not found to have a direct effect on the supply of living donations in most specifications. Furthermore, the exclusion of the obesity variable does not affect the substitution effect estimate significantly. However, when a quadratic term is included in the obesity specification there is a positive and statistically significant effect on living donations at the 5% level. Living donations increase at a decreasing rate as the prevalence of obesity increases and the effect is maximized when obesity prevalence is between 19.5% and 24%.
The prevalence of ESRD does have a significant effect on both living and cadaveric donations. We disaggregate the ESRD variable into two groups, diabetic and non-diabetic patients. Although both patient groups are unhealthy, the diabetic group is composed of individuals with a poorer health state. In all specifications, non-diabetic ESRD prevalence is found to have a positive effect on both types of kidney donations at the 1% level. On average, an increase in non-diabetic ESRD prevalence of 100 patients per million increases both living and cadaveric donations by 0.1-0.6 donations. However, an increase of 100 diabetic ESRD patients per million decreases both types of donations by 0.3-0.6 donations and is significant at the 5% level. These results are consistent with Howard (2002, 2011) and Sweeney (2010) where relative health is an important determinant to receive an organ.
Several other variables are found to have statistically significant effects on the level of living kidney donations at the 10% level. Living donations are found to increase as the percentage of African-Americans rises, yet it is not clear from this result if African-Americans are more likely to donate an organ or are more likely to receive an organ when this percentage rises. The level of living kidney donations also increases as the percentage of older citizens (75 years or older) increases. Potentially, the likelihood of needing an organ transplant is highest among this group as health depreciates with age. Neither marital status nor income per capita is found to have a statistically significant effect on living donations.
With respect to religion, a 1% increase in the population of Catholics or Jews relative to atheists decreases living donations by -0.05 and -0.3 donations per 100,000 individuals, respectively. (28) Yet, an increase in the percent-age of other religious groups does not have a significant effect on living donations. Next, we examine health insurance coverage and find that any type of health insurance decreases living donations, but the effect is only statistically significant with respect to Military and Medicaid coverage. Access to healthcare provides individuals with alternative forms of treatment to delay the need for an organ transplant. However, individuals without health insurance may only seek medical services once their health has been severely affected. On average, a 10% increase in these two forms of health insurance coverage decreases the level of kidney donations by -0.5 donations per 100,000 individuals.
This section further investigates the relationship between living and cadaveric donations and tests whether the substitution patterns identified in the previous sections hold for all subgroups of LDs or only a few. To this end, we disaggregate the LD variable into blood-related donors and non-blood-related donors. We further disaggregate non-blood-related donors into anonymous donors and spouse or unrelated known donors. The previous instrumental variable equations are re-estimated for each of these sub-samples.
The first column in Table 8 provides point estimates of the substitution effect with respect to blood-related donations per 100,000 individuals. In all specifications, the point estimates are found to be indistinguishable from zero at conventional levels. Blood-related donors appear to be insensitive to changes in the supply of cadaveric donations. A few reasons may cause this result. First, search and transaction costs are smallest for blood-related donors. These donors are typically easier to locate and more likely to be good matches for the organ recipient. Next, these donors value the life of the organ recipient more than other donor types. Therefore, they are less likely to risk losing their relative by waiting for a cadaveric organ donation.
However, non-blood-related donors are found to be sensitive to changes in the supply of cadaveric donations. The substitution elasticity ranges from -1.30 to -1.58 (point estimates between -0.19 and -0.23) and the point estimates are statistically significant at the 1% level. When we disaggregate non-blood-related donors into anonymous donors and non-blood relatives (i.e., spouses and friends) we find that non-blood-related donors who are known to the organ recipient have an elasticity (point) estimate equal to -1.59 (-0.21). This estimate is statistically significant at the 1% level. However, anonymous donors are not found to respond to shifts in the supply of cadaveric donations.
There are three decision agents at work when deciding who can/will donate the organ. The first agent is the organ donor who does internalize the cost of donation and donates when the private benefits exceed this cost. The share of non-blood-related donors who are not spouses increased from 14% in 2000 to 26.6% in 2009. (29) There are two potential explanations. First, transplant candidates must incur some non-monetary cost of exhorting potential LDs. These costs are probably higher for non-blood-related and unrelated donors. Pradel (2003) finds that transplant candidates would rather receive organs from blood relatives and non-blood relatives than friends. A decrease in the supply of deceased donors makes transplant candidates more willing to incur the price of exhorting non-blood-related, unrelated donors. (There are anecdotes that circulate in the transplant community of transplant candidates, facing long times for deceased donor kidneys, joining churches for the express purpose of soliciting an LD.) Second, the elasticity of the supply of non-blood-related, unrelated donors to waiting time may be greater than for related, blood-related donors. Potential related, blood-related donors may donate to signal their love. The strength of the signal does not depend on the length of the waiting list.
The second agent is the transplant center handling the living donation. This institution has individual guidelines by which they determine who can donate an organ. Mandelbrot et al. (2007) surveys different transplant centers to determine what factors influence their decisions. These factors include: age of the donor, the BMI of the donor, and health condition of the donor (hypertension, diabetes, family cardiovascular history, and family history of renal diseases). However, the survey reveals that the individual transplant centers may relax some of the health restrictions among older donors and among donors who have a pre-existing emotional relationship with the recipient. Still, the transplant centers prefer blood relatives as opposed to other known relatives. The third agent is the organ recipient. Pradel et al. (2003) surveys a group of (potential) organ donors and recipients to measure their willingness to donate/accept donation depending on the type of nephrectomy. Potential donors did not differ in their willingness to donate depending on the type of nephrectomy. Yet, among organ recipients 89% agreed that the use of laparoscopic nephrectomy, which would minimize the harm to the organ donor, affected their decision to accept a kidney from an LD.
Pradel (2003) also finds that organ recipients are less receptive to cadaveric donations for fear that the organ is damaged or diseased. The overall preference of anonymous LDs to other donor types stems from the organ recipient forgoing living donations from known sources for fear of harm to the donor (Kranenburg et al. 2007; Kranenburg et al. 2009; Young et al. 2008). This would explain the relative difference between known non-blood-related donors and anonymous donors.
Using variation in traffic safety laws as instruments for the supply of cadaveric donations, we find an inverse relationship between the supply of cadaveric organs and the use of LDs for kidney transplantation. One additional cadaveric kidney donor per 100,000 individuals decreases LDs by -0.2 to -0.5 donations per 100,000 per year, corresponding to an elasticity of substitution of -0.36 to -0.91, on average. Among LDs, blood-related and unrelated anonymous donors are found to be insensitive to changes in the supply of cadaveric donations, but related non-blood-related donors (i.e., spouses and friends) are found to be responsive to shifts in the supply of cadaveric donations with an elasticity of substitution equal to -1.59.
The implication of these results is that there is not a one-to-one correspondence between the number of kidneys recovered from CDs and the number of kidney transplants. As more kidneys from CDs become available, the number of LD transplants will decrease. These types of effects are important to consider in evaluations of policies that affect the number of deceased donors and simulation models of organ allocation. Allocation policies that prioritize groups of transplant candidates unlikely to have LDs (e.g., African-Americans) will increase the total number of transplants. Patients in groups that receive lower priority will have to make greater use of potential LDs. We are not necessarily suggesting that the allocation policy be designed with this purpose in mind, only that these types of effects should be taken into account.
2SLS: Two-Staged Least Squares
BMI: Body Mass Index
CD: Cadaveric Donor
EHECA: The Emergency Highway Energy Conservation Act
ESRD: End-Stage Renal Disease
LD: Living Donor
MVA: Motor Vehicle Accidents
NOTA: National Organ Transplant Act
OPO: Organ Procurement Organizations
OPTN: Organ Procurement Transplantation Network
PRA: Panel Reactive Antibody
Ashenfelter, O., and M. Greenstone. "Using Mandated Speed Limits to Measure the Value of a Statistical Life." Journal of Political Economy, 112(2,Part2), 2004, S226-S267.
Becker, G., and J. Elias. "Introducing Incentives in the Market for Live and Cadaveric Organ Donations." Journal of' Economic Perspectives, 21 (3), 2007, 3-24.
Centers for Disease Control and Prevention. "Obesity Trends among US Adults: 1985-2008." Accessed June, 2009. http://www.cdc.gov/obesity/downloads/obesity_trends_2008.pdf.
Clark, D. E., and B. M. Cushing. "Predicting Regional Variations in Mortality from Motor Vehicle Crashes." Academic Emergency Medicine, 6, 1999, 125-30.
--. "Predicted Effect of Automatic Crash Notification on Traffic Mortality." Accident Analysis and Prevention, 34, 2002, 103-9.
Cohen, A., and L. Einav. "'The Effects of Mandatory Seat Belt Laws on Driving Behavior and Traffic Fatalities." Review of Economics and Statistics, 85, 2003, 828-43.
Cragg, J. G., and S.G. Donald. "Testing Identifiability and Specification in Instrumental Variables Models." Econometric Theory, 9, 1993, 222-40.
Dee, T. "Motorcycle Hehnets and Traffic Safety." Journal of Health Economics, 28, 2009, 398-412.
Dickert-Conlin, S., T. Elder, and B. Moore. "Donorcycles: Do Motorcycle Helmet Laws Contribute to the Shortage of Organ Donors?" Journal of Law and Economics, 54(4), 2011,907-35.
Ejerblad, E., C.M. Fored, P. Lindblad, J. Fryzek, J. K. McLaughlin, and O. Nyren. "Obesity and Risk for Chronic Renal Failure." Journal of the American Society of Nephrology, 17, 2006, 1695-702.
Epstein, R. "The Human and Economic Dimensions of Altruism: The Case of Organ Transplantation." The Journal of Legal Studies, 37(2), 2008, 459-501.
Howard, D. H. "Why Do Transplant Surgeons Turn Down Organs? A Model of the Accept/Reject Decision." Journal of Health Economics, 21, 2002, 957-69.
--. "Producing Organs." Journal of Economic Perspectives, 21(3), 2007, 25-36.
--. "Waiting Time as a Price for Deceased Donor Kidneys." Contemporary, Economic Policy, 29(3), 2011, 295-303.
Hsu, C., C. E. McCulloch, C. Iribarren, J. Darbinian, and A. S. Go. "Body Mass Index and Risk for End-Stage Renal Disease." Annals of Internal Medicine, 144(1), 2006, 21-28.
Kleibergen, F., and R. Paap. "Generalized Reduced Rank Tests Using the Singular Value Decomposition." Journal of Econometrics, 127, 2006, 97-126.
Kranenburg, L. W., W. C. Zuidema, W. Weimar, M. T. Hilhorst, J. N. M. Ijzermans, J. Passchier, and J. J. V. Busschbach. "Psychological Barriers for Living Kidney Donation: How to Inform the Potential Donors?" Transplantation, 84(8), 2007, 965-71.
Kranenburg, L.W., M. Richards, W.C. Zuidema, W. Weimar, M. T. Hilhorst, J. N. M. Ijzermans. J. Passchier, and J. J. V. Busschbach. "Avoiding the Issue: Patients' (Non)communication with Potential Living Kidney Donors." Patient Education and Counseling, 74(1), 2009, 39-44.
Lave, C., and P. Elias. "Did the 65 MPH Speed Limit Save Lives'?" Accident Analysis and Prevention, 26(1), 1994, 49-62.
Liu, B. C., R. Ivers, R. Norton, S. Boufous, S. Blows, and S. K. Lo. "Helmets for Preventing Injury in Motorcycle Riders." Cochrane Database of Systematic Reviews, 1, 2008, CD004333.
Mandelbrot, D.A., M. Pavlakis, G. M. Danovitch, S. R. Johnson, S. J. Karp, K. Khwaja, D. W. Hanto, and J. R. Rodrigue. "The Medical Evaluation of Living Kidney Donors: A Survey of US Transplant Centers." American Journal of Transplantation, 7, 2007, 2333-43.
Moore, S. "Speed Doesn't Kill: The Repeal of the 55-MPH Speed Limit." Policy Analysis, 346, 1999, 1-23.
Muelleman, R. L., and K. Mueller. "Fatal Motor Vehicle Crashes: Variations of Crash Characteristics within Rural Regions of Different Population Densities.'" Journal of Trauma and Acute Care Surgery, 41, 1996, 315-20.
Pradel, F. G. "Exploring Donors' and Recipients' Attitudes about Living Donor Kidney Transplantation." Progress in Transplantation, 13(3), 2003, 203-10.
Pradel, F. G., M. R. Limcangco, C. D. Mullins, and S. T. Bartlett. "Patients' Attitudes about Living Donor Transplantation and Living Donor Nephrectomy." American Journal of Kidney Diseases, 41(4), 2003, 849-58.
Roth, A. "Repugnance as a Constraint on Markets." Journal of Economic Perspectives, 21(3), 2007, 37-58.
Sass, T. R., and P. R. Zimmerman. "Motorcycle Helmet Laws and Motorcyclist Fatalities." Journal of Regulatory Economics, 18, 2000, 195-215.
Schweitzer, E. J., J. Wilson, S. Jacobs, C. H. Machan, B. Philosophe, A. Farney, J. Colonna, MD, B. E. Jarrell, and S. T. Bartlett. "Increased Rates of Donation with Laparoscopic Donor Nephrectomy." Annals of Surgery, 232(3), 2000, 392-400.
Selck, F. W., P. Debb, and E. B. Grossman. "Deceased Organ Donor Characteristics and Clinical Interventions Associated with Organ Yield." American Journal of Transplantation, 8, 2008, 965-74.
Stock, J., J. Wright, and M. Yogo. "A Survey of Weak Instruments and Weak Identification in Generalized Method of Moments." Journal of the American Statistical Association, 20(4), 2002, 518-29.
Stock, J. H., and M. Yogo. "Testing for Weak Instruments in Linear IV Regression," in Identification and Inference for Econometric Models: Essays in Honor of Thomas J. Rothenberg, Chapter 5, edited by J.H. Stock and D.W.K. Andrews. Cambridge: Cambridge University Press, 2005, 80-108.
Sweeney, M. "Does the Public Provision of Kidneys Crowd Out Living Kidney Donation? A Regression Discontinuity Analysis.'" Unpublished Manuscript, Cornell University, 2010.
Young, A., M. Karpinski, D. Treleaven, A. Waterman, C. R. Parikh, H. Thiessen-Philbrook, R.C. Yang, and A. X. Garg. "Differences in Tolerance for Health Risk to the Living Donor among Potential Donors, Recipients, and Transplant Professionals." Kidney International, 73(10), 2008, 1159-66.
Zlatoper, T. J. "Models Explaining Motor Vehicle Death Rates in the United States." Accident Analysis and Prevention, 21, 1989, 125-54.
JOSE M. FERNANDEZ, DAVID H. HOWARD and LISA STOHR KROESE *
* We wish to thank Partha Deb. Lan Shi, and seminar/session participants at the University of Louisville: the Midwest Economic Association Meetings, Evanston, IL; the American Society of Health Economist Conference, Ithaca, NY; and the American Economic Association Meetings in Denver, CO.
Fernandez: Assistant Professor, Department of Economics, College of Business, University of Louisville. Louisville, KY 40292. Phone +1 502 852 4861, Fax +1 502 852 7672, E-mail email@example.com
Howard: Associate Professor, Department of Health Policy and Management, Rollins School of Public Health, Emory University, Atlanta, GA 30322. Phone +1 404 727 3907, Fax +1 404 727 9198. E-mail firstname.lastname@example.org
Stohr Kroese: Department of Economics, College of Business, University of Louisville, Louisville, KY 40292. Phone +1 502 852 4861, Fax +1 502 852 7672. E-mail email@example.com
(1.) On the basis of OPTN data as of March 19, 2012. See http://www.optn.org/.
(2.) Currently, 13 states have legislation awarding a $10,000 state income tax credit to LDs. Another six states allow LDs to receive paid leave of absence from work for up to 30 days. See the following website for more information http://www.transplantliving.org/livingdonation/financial aspects/statetax.aspx. The use of websites, such as http:// www.matchingdonors.com, to find potential donors has also aided the supply of CDs. Finally, Mandelbrot (2007) reports on changes in donations policies that have expanded the number of viable donors including donations from non-blood relatives and from cardiac deaths.
(3.) See Becker and Ellias (2007), Epstein (2008), Howard (2007), Roth (2007).
(4.) Panel reactive antibody (PRA) is a blood test determining the likelihood of kidney rejection. Individuals with low PRA levels are given less priority on the publicly available supply of organs.
(5.) A secondary set of instruments used is demographic information for non-medical injury deaths gathered from the compressed mortality files. The sub-sample of injury deaths are associated with deaths where the internal organs are not damaged as a result of medication and thus are good candidates for donation.
(6.) Only seven states maintained the national speed limit on rural highways: Connecticut, Maryland, Massachusetts, New Jersey, New York, Pennsylvania, and Rhode Island. Delaware, the District of Columbia, and Hawaii did not have highways classified as rural.
(7.) We do not report dates for Alaska, Idaho, Montana, or Wyoming as these states are not used in this study due to lack of organ donation data for these states.
(8.) These laws cover front-seat occupants only, although belt laws in 21 states and the District of Columbia cover all rear seat occupants as well. See Table 1 for current belt laws by state.
(9.) The OPTN publicly provides these data on their website www.OPTN.org. Both individual level and OPO level data are available, but only the OPO level data have geographic information.
(10.) Howard (2011) and Dickert-Conlin, Elder, and Moore (2011) use the OPO level data to exploit geographic variation in donations rates, but these studies use a shorter panel than the sample used in this article.
(11.) Organ donation data are available for Washington, DC, but demographic data over the same time period are missing. These counts exclude paired transplants or transplant chains where organ recipients arrange for one of their organs to be transplanted into a different recipient in exchange for receiving an organ themselves. We only considered transplanted kidneys although more than the reported number of kidneys are recovered, but lost to disease or spoil.
(12.) Medical expenses for the donor are covered by the donee's insurance. About 80% of the transplant cost is covered by Medicare and the remaining 20% is covered by private insurance.
(13.) Marital status data are missing for less populated states between 1988 and 1994. This reduces the sample to 933 state year observations. However, missing values are imputed using linear interpolation in the regression. The estimated parameters are not sensitive to inclusion of these imputed values.
(14.) U.S. Obesity Trends by State from 1985 to 2008 http://www.cdc.gov/obesity/data/trends.html (last accessed June 2009). Twenty-one observations are lost because of missing obesity data for Arkansas, Colorado, the District of Columbia, Maryland, New Jersey, and Pennsylvania in a few years.
(15.) Linear interpolations are used to fill in missing values between survey years.
(16.) See the following article for more information: SEOPF/UNOS, "Organ and Tissue Donation: A Reference Guide for Clergy," 4th ed., edited by M. L. Cooper and G. J. Taylor. Richmond: 2000.
(17.) These events present a problem during estimation in that the use of year fixed effects may confound the effect of speed limit changes.
(18.) All organ transplant recipients have 80% of their cost covered by Medicare Part B regardless of age.
(19.) Kidney grafts from LDs have a 95.1% survival rate versus 89% for grafts from CDs.
(20.) See Epstein (2008) for a discussion on the effects of altruism on organ supply.
(21.) We also consider an autoregressive error, but do not find any evidence that either living or cadaveric donations follow a random walk or explosive process.
(22.) We did attempt to proxy for altruism by using average charitable contributions per tax return within a state, but found no effect on living kidney donations.
(23.) As a rule of thumb, an F-stat of 10 or greater indicates strong instruments (Stock, Wright, and Yogo 2002).
(24.) Only New Hampshire did not have belt law during this time period and all other states simply switched between primary or secondary enforcement seat belt laws.
(25.) These results still hold after first-differencing the data to control for autoregressive shocks in the error term.
(26.) Elasticity can be evaluated at the mean as [epsilon] = ([delta]y/[delta]x)(x/y)|[sub.x=[bar.x], y=[bar.y]].
(27.) These results are also robust to different specifications of the error term including an AR(1) process and a random effects model.
(28.) Religious beliefs concerning death differ with respect to cardiac death versus "brain death." Orthodox Judeo-Christian beliefs define death as cardiac death, but cadaveric donations primarily occur during "'brain death."
(29.) Scientific Registry of Transplant Recipients Annual Data Report, Table 2.9; (http://www. srtr.org/annual_reports/2010/209 don-rel-ty_dc.htm?o=2& g=5&c=5).
TABLE 1 Motor Vehicles Safety Laws Helmet Law Seat Belt Law State Universal Partial Primary Secondary Alabama 6/23/1976 2/9/1999 7/18/1991 Arizona 5/27/1976 1/1/1991 Arkansas 6/29/1967 1997 6/30/2009 7/15/1991 California 1/1/1992 1/1/1985 1/1/1993 1/1/1986 Colorado 7/1/2007 5/20/1977 7/1/1987 Connecticut 1/1/1990 1/1/1986 Delaware 6/10/1978 6/30/2003 1/1/1992 Florida 9/13/1967 7/1/2000 6/30/2009 7/1/1986 Georgia 7/6/1969 7/1/1996 9/1/1988 Hawaii 6/7/1977 12/16/1985 Illinois 7/3/2003 1/1/1988 Indiana 1/1/1984 7/1/1998 7/1/1987 Iowa 7/l/1986 Kansas 7/1/1979 7/1/1986 Kentucky 6/13/1968 7/15/1998 7/20/2006 7/15/1994 Louisiana 1/1/1982 8/15/1999 9/1/1995 7/1/1986 Maine 8/15/2004 7/1/1980 9/20/2007 12/26/1995 Maryland 10/1/1992 7/1/1979 10/1/1997 7/1/1986 Massachusetts 5/22/1967 2/1/1994 Michigan 7/29/1969 4/1/2000 7/1/1985 Minnesota 4/6/1977 6/9/2009 8/1/1986 Mississippi 3/28/1974 5/27/2006 7/1/1994 Missouri 9/28/1967 9/28/1985 Nebraska 1/1/1989 1/1/1993 Nevada 1/1/1972 7/1/1987 New Hampshire repeal 8/7/1977 No Law 9/30/1995 New Jersey 1/1/1968 5/1/2000 3/1/1985 New Mexico 3/31/1978 1/1/1986 New York 1/1/1967 12/1/1984 North Carolina 1/1/1968 12/1/2006 10/1/1985 North Dakota 7/1/1977 7/14/1994 Ohio 7/10/1978 5/6/1996 Oklahoma 5/21/1976 11/1/1997 2/1/1987 Oregon 6/16/1988 10/4/1977 12/7/1990 Pennsylvania 7/15/1968 9/4/2003 11/23/1987 Rhode Island 7/1/1992 6/18/1991 South Carolina 6/16/1980 12/9/2005 7/1/1989 South Dakota 7/1/1977 1/1/1995 Tennessee 6/4/1967 7/1/2004 4/21/1986 Texas 9/1/1989 9/1/1997 9/1/1985 Utah 5/10/1977 4/28/1986 Vermont 3/6/1968 1/1/1988 Virginia 6/26/1970 1/1/1994 Washington 6/7/1990 7/1/1987 7/l/2002 6/11/1986 West Virginia 5/25/1971 9/1/1993 Wisconsin 3/19/1978 6/30/2009 12/1/1987 Speed Limits (mph) State Rural Urban Alabama 70 65 Arizona 75 65 Arkansas 70 55 California 70 65 Colorado 75 65 Connecticut 65 55 Delaware 65 55 Florida 70 65 Georgia 70 65 Hawaii 60 60 Illinois 65 55 Indiana 70 55 Iowa 70 55 Kansas 75 75 Kentucky 70 65 Louisiana 75 70 Maine 75 65 Maryland 65 65 Massachusetts 65 65 Michigan 70 65 Minnesota 70 65 Mississippi 70 70 Missouri 70 60 Nebraska 75 65 Nevada 65 65 New Hampshire 75 65 New Jersey 65 55 New Mexico 75 75 New York 65 65 North Carolina 70 70 North Dakota 75 75 Ohio 70 65 Oklahoma 75 70 Oregon 65 55 Pennsylvania 65 55 Rhode Island 65 55 South Carolina 70 70 South Dakota 75 75 Tennessee 70 70 Texas 80 75 Utah 80 65 Vermont 65 55 Virginia 70 70 Washington 70 60 West Virginia 70 55 Wisconsin 65 65 Notes: The Insurance Institute for Highway Safety http://www.iih. org/laws/ provides the effective dates. Maximum limit may apply only to specified segments of interstate. Universal coverage requires that all riders wear helmets. Partial coverage requires only riders under a certain age must wear helmets. Primary enforcement allows a vehicle to be stopped solely for this offense, but secondary enforcement requires an alternative offense as the reason for stopping the vehicle. TABLE 2 Summary Statistics for Donation Types, Obesity Levels, and Traffic Safety Laws Variables M SD Minimum Living donations 93.72 113.89 0.00 Living donations per 100,000 1.51 1.11 0.00 Cadaveric donations 177.73 205.15 0.00 Cadaveric donations per 100,000 2.74 1.09 0.00 Blood-related donations 68.80 79.45 0.00 Blood-related donations per 100,000 1.11 0.74 0.00 ESRD prevalence (non-diabetic) (a) 819.73 172.51 424.02 ESRD prevalence (diabetic) 437.06 180.75 101.95 % Obese (BMI [greater than or equal 0.18 0.06 0.06 to] 30) Secondary seat belt laws 0.60 0.49 0.00 Primary seat belt laws 0.30 0.46 0.00 Helmet law (full) 0.47 0.50 0.00 Helmet law (partial) 0.42 0.49 0.00 Rural speed limit (mph) 66.6 4.46 55.0 75 mph (rural) 0.11 0.32 0.00 70 mph (rural) 0.22 0.41 0.00 65 mph (rural) 0.58 0.49 0.00 60 mph (rural) 0.04 0.20 0.00 55 mph or less (rural) 0.04 0.21 0.00 Urban speed limit (mph) 59.6 6.14 55.0 75 mph or greater (urban) 0.03 0.18 0.00 70 mph (urban) 0.10 0.29 0.00 65 mph (urban) 0.23 0.42 0.00 60 mph (urban) 0.04 0.19 0.00 55 mph (urban) 0.60 0.49 0.00 Variables Maximum N Living donations 743 966 Living donations per 100,000 8.07 966 Cadaveric donations 1292 966 Cadaveric donations per 100,000 7.13 966 Blood-related donations 500 966 Blood-related donations per 100,000 5.54 966 ESRD prevalence (non-diabetic) (a) 1261.6 966 ESRD prevalence (diabetic) 1174.0 966 % Obese (BMI [greater than or equal 0.33 966 to] 30) Secondary seat belt laws 1.00 966 Primary seat belt laws 1.00 966 Helmet law (full) 1.00 966 Helmet law (partial) 1.00 966 Rural speed limit (mph) 75.0 966 75 mph (rural) 1.00 966 70 mph (rural) 1.00 966 65 mph (rural) 1.00 966 60 mph (rural) 1.00 966 55 mph or less (rural) 1.00 966 Urban speed limit (mph) 75.0 966 75 mph or greater (urban) 1.00 966 70 mph (urban) 1.00 966 65 mph (urban) 1.00 966 60 mph (urban) 1.00 966 55 mph (urban) 1.00 966 Notes: Data for Alaska, Idaho, Montana, and Wyoming are unavailable. (a) ESRD patients alive on December 31 of each year, per million people. TABLE 3 Summary Statistics for State Demographics M SD Minimum Living variables % White 0.77 0.15 0.28 % Black 0.11 0.09 0.00 %a Hispanic 0.07 0.09 0.00 % Other 0.05 0.09 0.00 % Female 0.51 0.01 0.49 % Married (a) 0.59 0.04 0.23 % Divorced (a) 0.10 0.02 0.06 %a Widowed (a) 0.08 0.02 0.04 % Separated (a) 0.02 0.01 0.00 Never married (a) 0.18 0.03 0.09 % Partnered (a) 0.03 0.01 0.00 Age between 0 and 19 years old 0.28 0.02 0.23 Age between 20 and 34 years old 0.22 0.02 0.17 Age between 35 and 54 years old 0.28 0.02 0.20 Age between 55 and 74 years old 0.16 0.02 0.11 Age 75 years old or greater 0.06 0.01 0.03 % No insurance 0.12 0.04 0.05 % Private insurance 0.64 0.06 0.45 % Medicaid 0.10 0.03 0.02 % Medicare 0.12 0.02 0.07 % Military 0.04 0.02 0.01 % Catholic 0.20 0.12 0.02 % Jewish 0.01 0.02 0.00 % Other religion 0.30 0.15 0.09 State population (100,000) 58.62 60.81 5.50 State income per capita ($10,000) 2.69 0.55 1.57 Death due to injury (b) variables % White 0.83 0.12 0.29 Black 0.13 0.11 0.00 % Other 0.04 0.10 0.00 % Female 0.32 0.16 0.00 Ages between 0 and 19 years old 0.13 0.03 0.05 Ages between 20 and 34 years old 0.25 0.04 0.14 Ages between 35 and 54 years old 0.29 0.05 0.17 Ages between 55 and 74 years old 0.15 0.02 0.10 Ages greater than 75 years old 0.18 0.05 0.07 Maximum N Living variables % White 0.98 966 % Black 0.37 966 %a Hispanic 0.45 966 % Other 0.61 966 % Female 0.52 966 % Married (a) 0.69 933 % Divorced (a) 0.16 933 %a Widowed (a) 0.17 933 % Separated (a) 0.08 933 Never married (a) 0.42 933 % Partnered (a) 0.07 933 Age between 0 and 19 years old 0.40 966 Age between 20 and 34 years old 0.28 966 Age between 35 and 54 years old 0.33 966 Age between 55 and 74 years old 0.21 966 Age 75 years old or greater 0.09 966 % No insurance 0.24 966 % Private insurance 0.77 966 % Medicaid 0.20 966 % Medicare 0.18 966 % Military 0.12 966 % Catholic 0.63 966 % Jewish 0.10 966 % Other religion 0.76 966 State population (100,000) 367.59 966 State income per capita ($10,000) 5.34 966 Death due to injury (b) variables % White 1.00 874 Black 0.43 874 % Other 0.70 874 % Female 0.95 874 Ages between 0 and 19 years old 0.22 874 Ages between 20 and 34 years old 0.37 874 Ages between 35 and 54 years old 0.40 874 Ages between 55 and 74 years old 0.22 874 Ages greater than 75 years old 0.32 874 Note: Data for Alaska. Idaho. Montana, and Wyoming are unavailable. (a) Marital status is missing for Delaware and Rhode Island. Missing values are imputed using linear interpolation for n = 966. (b) Compressed mortality file is available for 1988-2006. Only deaths resulting from non-medical injury are used. TABLE 4 Least Squares Regression of Living Donor per 100,000 in State Population OLS l Living Donations per 100,000 [beta] SE Cadaveric donations per 100,000 0.01 (0.04) Prevalence ESRD (diabetic) (a) -4e-4 (5e-4) Prevalence ESRD (non-diabetic) (a) 4e-3 * (5e-4) % White -6.16 (4.15) % Black 23.6 * (5.57) % Hispanic 6.28 (5.07) % Female -32.7 ** (18.7) % Obese 0.86 (1.15) (% Obese) (b) Income per capita ($10,000) 0.01 (0.03) Population (100,000) -0.01 * (3e-3) Age: 0-19 -30.5 * (9.14) Age: 20-34 -39.5 * (8.34) Age: 35-54 -30.1 * (9.37) Age: 55-74 -24.2 * (8.80) % Private insured -0.81 (1.03) % Medicaid -4.92 * (1.29) % Medicare -2.10 (2.03) % Military -2.70 (1.94) % Catholic -2.63 * (0.97) % Jewish -18.4 * (5.84) % Other religion 1.25 ** (0.71) % Married % Divorced % Widowed % Separated % Partnered [R.sup.2] 0.87 No. of observations 966 State and year fixed effects Yes OLS 2 Living Donations per 100,000 [beta] SE Cadaveric donations per 100,000 0.01 (0.04) Prevalence ESRD (diabetic) (a) -5e-4 (6e-4) Prevalence ESRD (non-diabetic) (a) 4e-3 * (5e-4) % White -5.61 (4.17) % Black 24.1 * (5.67) % Hispanic 6.32 (5.02) % Female -32.5 ** (18.8) % Obese 0.87 (1.15) (% Obese) (b) Income per capita ($10,000) 0.02 (0.03) Population (100,000) -0.01 * (3e-3) Age: 0-19 -29.8 * (9.19) Age: 20-34 -39.0 * (8.35) Age: 35-54 -29.3 * (9.42) Age: 55-74 -23.4 * (8.87) % Private insured -0.69 (1.02) % Medicaid -4.78 * (1.27) % Medicare -1.90 (2.02) % Military -2.49 (1.94) % Catholic -2.73 * (0.97) % Jewish -17.1 * (6.09) % Other religion 1.06 (0.73) % Married 0.05 (0.80) % Divorced -2.15 (1.43) % Widowed -0.81 (1.52) % Separated 0.10 (3.34) % Partnered 2.81 (2.30) [R.sup.2] 0.87 No. of observations 966 State and year fixed effects Yes OLS 3 Living Donations per 100,000 [beta] SE Cadaveric donations per 100,000 0.03 (0.04) Prevalence ESRD (diabetic) (a) -3e-4 (6e-4) Prevalence ESRD (non-diabetic) (a) 4e-3 * (5e-4) % White -3.70 (4.15) % Black 24.1 * (5.73) % Hispanic 4.16 (5.23) % Female -36.6 *** (18.7) % Obese 14.5 * (3.20) (% Obese) (b) -36.2 * (8.46) Income per capita ($10,000) 0.01 (0.03) Population (100,000) -0.01 * (3e-3) Age: 0-19 -26.4 * (9.17) Age: 20-34 -36.9 * (8.25) Age: 35-54 -30.5 * (9.32) Age: 55-74 -25.2 * (8.70) % Private insured -1.09 (1.03) % Medicaid -5.12 * (1.29) % Medicare -1.40 (2.00) % Military -2.14 (1.91) % Catholic -2.84 * (0.96) % Jewish -12.3 ** (6.31) % Other religion -0.33 (0.85) % Married -0.48 (0.80) % Divorced -2.38 ** (1.71) % Widowed -1.17 (1.58) % Separated -0.60 (3.37) % Partnered 2.08 (2.38) [R.sup.2] 0.88 No. of observations 966 State and year fixed effects Yes Notes: All estimation samples consist of 46 states from 1988 to 2008. The unit of observation is state x year. All observations are weighted by the state's population in the given year. (a) Signifies ESRD prevalence per million by state population. (b) Standard errors are estimated using the Huber/White/Sandwich estimator of variance. * p<.01, ** p<.10, *** p<.05. TABLE 5 Least Squares Regression of Cadaveric Donations per 100,000 on Traffic Laws OLS 4 OLS 5 Cadaveric Donations per 100,000 [beta] SE [beta] SE Helmet (full) -1.11 * (0.19) -1.12 * (0.18) Helmet (partial) -0.82 * (0.17) -0.84 * (0.16) Speed limit urban (a) -0.21 (0.33) Speed limit rural (a) 0.90 ** (0.51) Primary seat belt laws Secondary seat belt laws ESRD (diabetic) (b) -3e-3 * (6e-4) -3e-3 * (6e-4) ESRD (non-diabetic) (b) 3e-3 * (7e-4) 3e-3 * (7e-4) % White -4.73 (5.39) -2.73 (5.60) % Black -0.82 (8.98) 0.19 (9.14) % Hispanic 2.96 (6.77) 4.13 (6.90) % Female 63.0 * (21.3) 66.5 * (21.7) Obese -6.81 (4.21) -6.15 (4.14) (% Obese) (2) 16.8 (10.4) 15.3 (10.3) Income per capita ($10,000) -5e-3 (0.04) -4e-3 (0.04) Population (100,000) -6e-3 (4e-3) -4e-3 (4e-3) Age: 0-19 -34.7 * (12.3) -32.8 * (12.5) Age: 20-34 -34.5 * (11.4) -32.1 * (11.5) Age: 35-54 -36.6 * (12.5) -34.7 * (12.6) Age: 55-74 -51.7 * (11.2) -49.1 * (11.5) Private insured -0.76 (1.38) -0.79 (1.38) % Medicaid -0.82 (1.59) -1.01 (1.59) % Medicare 5.53 *** (2.44) 5.68 *** (2.45) % Military -4.56 ** (2.45) -4.31 ** (2.44) % Catholic -4.43 * (1.24) -4.26 * (1.24) % Jewish -23.0 * (8.64) -21.2 *** (8.85) % Other religion -3.56 * (1.10) -3.63 * (1.13) % Married 0.41 (0.94) 0.45 (0.95) % Divorced 6.44 * (1.72) 6.40 * (1.71) % Widowed 3.65 *** (1.84) 3.63 *** (1.84) % Separated -5.07 (3.56) -4.93 (3.57) % Partnered -3.53 (2.64) -3.30 (2.63) [R.sup.2] 0.77 0.77 No. of observations 966 966 Additional controls Injury death controls No No Fatality rates No No State and year fixed effects Yes Yes OLS 6 OLS 7 Cadaveric Donations per 100,000 [beta] SE [beta] SE Helmet (full) -1.12 * (0.18) -1.34 * (0.22) Helmet (partial) -0.83 * (0.16) -1.05 * (0.20) Speed limit urban (a) -0.18 (0.33) -0.24 (0.36) Speed limit rural (a) 0.89 ** (0.52) 0.96 ** (0.55) Primary seat belt laws 0.03 (0.13) -0.06 (0.12) Secondary seat belt laws -0.02 (0.11) -0.06 (0.11) ESRD (diabetic) (b) -3e-3 * (6e-4) -3e-3 * (8e-4) ESRD (non-diabetic) (b) 3e-3 * (7e-4) 3e-3 * (8e-4) % White -2.07 (5.83) -1.04 (6.93) % Black 0.33 (9.17) 4.43 (10.8) % Hispanic 4.65 (7.03) 6.64 (8.40) % Female 68.9 * (23.0) 74.7 * (25.3) Obese -5.74 (4.17) -5.35 (4.58) (% Obese) (2) 14.3 (10.4) 10.7 (12.1) Income per capita ($10,000) -2e-3 (0.04) -0.02 (0.05) Population (100,000) -5e-3 (4e-3) -3e-3 (5e-3) Age: 0-19 -32.8 * (12.6) -27.5 ** (14.1) Age: 20-34 -31.6 * (11.8) -24.6 ** (13.5) Age: 35-54 -34.1 * (12.9) -33.2 *** (14.7) Age: 55-74 -48.9 * (11.6) -40.0 * (13.2) Private insured -0.77 (1.38) 0.53 (1.44) % Medicaid -0.91 (1.60) -0.29 (1.68) % Medicare 5.65 *** (2.46) 6.86 * (2.47) % Military -4.28 ** (2.45) -4.77 ** (2.59) % Catholic -4.24 * (1.25) -6.66 * (1.58) % Jewish -21.4 *** (8.87) -14.5 (11.0) % Other religion -3.55 * (1.12) -3.73 * (1.38) % Married 0.43 (0.95) 0.14 (1.01) % Divorced 6.40 * (1.72) 5.39 * (1.86) % Widowed 3.52 ** (1.86) 3.28 ** (1.91) % Separated -5.10 (3.56) -4.08 (3.60) % Partnered -3.27 (2.64) -2.52 (2.67) [R.sup.2] 0.77 0.78 No. of observations 966 874 Additional controls Injury death controls No Yes Fatality rates No No State and year fixed effects Yes Yes OLS 8 Cadaveric Donations per 100,000 [beta] SE Helmet (full) -0.97 * (0.23) Helmet (partial) -0.73 * (0.19) Speed limit urban (a) Speed limit rural (a) Primary seat belt laws 0.57 * (0.18) Secondary seat belt laws 0.47 * (0.17) ESRD (diabetic) (b) -6e-3 * (1e-3) ESRD (non-diabetic) (b) 3e-3 * (1e-3) % White -8.60 (6.80) % Black -24.1 *** (9.48) % Hispanic -2.77 (8.33) % Female 9.90 (27.6) Obese 2.55 (6.15) (% Obese) (2) -2.27 (14.7) Income per capita ($10,000) 0.02 (0.05) Population (100,000) 0.01 ** (6e-3) Age: 0-19 -87.2 * (19.3) Age: 20-34 -95.9 * (19.0) Age: 35-54 -108 * (20.8) Age: 55-74 -105 * (20.2) Private insured -0.98 (1.50) % Medicaid 0.77 (1.77) % Medicare 3.68 (2.92) % Military -7.32 *** (2.86) % Catholic -6.08 * (1.84) % Jewish -15.9 (16.7) % Other religion -1.05 (1.84) % Married -1.90 (1.88) % Divorced 4.37 (3.65) % Widowed -3.96 (3.69) % Separated -7.43 (5.76) % Partnered -6.45 ** (3.33) [R.sup.2] 0.81 No. of observations 644 Additional controls Injury death controls No Fatality rates Yes State and year fixed effects Yes Notes: Robust standard errors are estimated using the Huber/White/Sandwich estimator of variance. All observations are weighted by the state's population in the given year. Observations are lost due to missing years in the mortality files and fatality rates. (a) Indicates the natural logarithm of the variable is used. (b) Signifies ESRD prevalence per million by state population. * p < .01, ** p <. 10, *** p <. 05. TABLE 6 Least Square Regression of MVA donations versus Non-MVA donations per 100,000 Non-MVA Donations per 100,000 (1) (2) (3) Fatality rate per 100 million -0.20 -0.16 -0.20 vehicle miles traveled (0.19) (0.19) (0.20) Full helmet laws 0.06 0.09 0.09 (0.20 (0.20 (0.20 Partial helmet laws -0.11 -0.07 -0.08 (0.17) (0.17) (0.17) Speed limit urban (a) -0.03 -0.07 (0.41) (0.43) Speed limit rural (a) -0.43 -0.33 (0.78) (0.81) Seat belt law (primary) -0.05 (0.09) [R.sup.2] 0.94 0.94 0.94 N 494 494 494 MVA Donations per 100,000 (4) (5) (6) Fatality rate per 100 million 0.08 * 0.07 0.05 vehicle miles traveled (0.05) (0.05) (0.06) Full helmet laws -0.14 ** -0.12 *** -0.12 * (0.05) (0.05) (0.07) Partial helmet laws -0.12 ** -0.12 ** -0.14 *** (0.04) (0.04) (0.07) Speed limit urban (a) 0.32 *** 0.36 * (0.13) (0.19) Speed limit rural (a) -0.54 ** (0.17) Seat belt law (primary) -3e-3 (0.01) [R.sup.2] 0.81 0.81 0.81 N 494 494 494 MVA Donations per 100,000 (7) (8) Fatality rate per 100 million 0.07 0.07 vehicle miles traveled (0.06) (0.05) Full helmet laws -0.12 * -0.12 *** (0.07) (0.05) Partial helmet laws -0.14 *** -0.12 ** (0.067) (0.044) Speed limit urban (a) 0.32 *** (0.13) Speed limit rural (a) -0.25 -0.534 ** (0.33) (0.17) Seat belt law (primary) -1e-3 -1e-3 (0.01) (0.02) [R.sup.2] 0.81 0.81 N 494 494 Notes: t statistics in parentheses. State and year fixed effects are used in all regressions. The standard errors are robust to arbitrary heteroskedasticity and are clustered by state. The sample is restricted to the years 1995-2008, when the OPTN provided data about the mechanism of death for cadaveric donations. State characteristics are also included as additional controls. Indicates the natural logarithm of the variable is used. * p < .10, ** p < .01, *** p < .05. TABLE 7 Instrumental Variables Estimation of Living Donations per 100,000 IV 1 IV 2 Living Donations per 100,000 [beta] SE [beta] SE Cadaveric donations per -0.53 * (0.19) -0.46 * (0.17) 100,000 ESRD (diabetic) (a) -2e-3 *** (7e-4) -1e-3 *** (7e-4) ESRD (non-diabetic) (a) 5e-3 * (7e-4) 5e-3 * (6e-4) %a White -4.62 (5.02) -4.51 (4.83) % Black 25.7 * (7.87) 25.5 * (7.46) Hispanic 7.26 (6.30) 6.91 (6.01) Female -4.72 (23.4) -8.27 (22.0) % Obese 8.68 *** (4.33) 9.33 *** (4.06) (% Obese) (2) -21.6 *** (11.0) -23.3 *** (10.4) Income per capita ($10,000) 1e-3 (0.04) 0.01 (0.04) Population (100,000) -0.01 * (4e-3) -0.01 * (4e-3) Age: 0-19 -45.4 * (13.5) -43.3 * (12.7) Age: 20-34 -53.7 * (12.3) -51.9 * (11.6) Age: 35-54 -44.9 * (12.2) -43.3 * (11.7) Age: 55-74 -53.0 * (14.1) -49.9 * (13.2) % Private insured -1.68 (1.20) -1.61 (1.15) % Medicaid -5.97 * (1.47) -5.87 * (1.42) % Medicare 1.62 (2.71) 1.29 (2.56) % Military -4.58 *** (2.32) -4.31 ** (2.20) % Catholic -5.45 * (1.49) -5.16 * (1.39) % Jewish -25.2 * (8.98) -23.8 * (8.51) % Other religion -1.97 ** (1.10) -1.79 ** (1.03) % Married -0.10 (0.89) -0.14 (0.86) % Divorced 1.63 (2.11) 1.18 (1.97) % Widowed 0.98 (1.87) 0.74 (1.78) % Separated -3.13 (3.64) -2.84 (3.53) % Partnered 0.41 (2.55) 0.60 (2.48) Centered [R.sup.2] 0.82 0.83 No. of observations 966 966 Instruments Helmet laws Yes Yes Speed limits No Yes Seat belt laws No No Injury death controls No No Fatality rates No No Identification test Under-identification 20.82 * 23.67 * Weak-instruments (b) 17.49 ([dagger]) 10.35 ([dagger]) Over-identification 3.98 *** 6.01 State and year fixed effects Yes Yes IV 3 IV 4 Living Donations per 100,000 [beta] SE [beta] SE Cadaveric donations per -0.50 * (0.17) -0.23 ** (0.12) 100,000 ESRD (diabetic) (a) -2e-3 *** (7e-4) -1e-3 *** (6e-4) ESRD (non-diabetic) (a) 5e-3 * (6e-4) 6e-3 * (6e-4) %a White -4.58 (4.96) -2.59 (4.81) % Black 25.6 * (7.68) 33.3 * (7.29) Hispanic 7.14 (6.16) 8.79 (5.81) Female -5.93 (22.4) -13.0 (20.1) % Obese 8.90 *** (4.15) 12.1 * (3.67) (% Obese) (2) -22.2 *** (10.6) -31.8 * (10.2) Income per capita ($10,000) 0.01 (0.04) 2e-3 (0.03) Population (100,000) -0.01 * (4e-3) -0.01 * (4e-3) Age: 0-19 -44.6 * (13.1) -33.0 * (10.8) Age: 20-34 -53.1 * (11.9) -42.0 * (9.66) Age: 35-54 -44.3 * (12.0) -35.6 * (10.5) Age: 55-74 -51.9 * (13.5) -31.7 * (10.5) % Private insured -1.66 (1.18) -0.61 (1.05) % Medicaid -5.94 * (1.45) -5.27 * (1.35) % Medicare 1.51 ('2.61) 0.14 (2.18) % Military -4.49 *** (2.23) -2.63 (2.02) % Catholic -5.35 * (1.42) -5.07 * (1.47) % Jewish -24.7 * (8.76) -6.19 (6.58) % Other religion -1.91 ** (1.05) -1.07 (1.00) % Married -0.12 (0.88) -0.99 (0.74) % Divorced 1.47 (2.01) -1.52 (1.54) % Widowed 0.90 (1.82) -0.78 (1.56) % Separated -3.03 (3.59) -1.77 (3.15) % Partnered 0.48 (2.52) 2.33 (2.29) Centered [R.sup.2] 0.82 0.87 No. of observations 966 874 Instruments Helmet laws Yes Yes Speed limits Yes Yes Seat belt laws Yes Yes Injury death controls No Yes Fatality rates No No Identification test Under-identification 24.94 * 37.50 * Weak-instruments (b) 7.04 ([double 4.94 daggger]) Over-identification 12.56 *** 33.38 *** State and year fixed effects Yes Yes IV 5 Living Donations per 100,000 [beta SE Cadaveric donations per -0.54 *** (0.22) 100,000 ESRD (diabetic) (a) -4e-3 * (e-3) ESRD (non-diabetic) (a) 6e-3 * (1e-3) %a White -8.77 (6.66) % Black 16.3 (11.6) Hispanic -3.16 (8.30) Female -20.0 (25.6) % Obese 14.2 * (5.34) (% Obese) (2) -29.2 *** (13.1) Income per capita ($10,000) 0.04 (0.05 Population (100,000) -0.01 (6e-3) Age: 0-19 -88.7 * (24.7) Age: 20-34 -103 * (25.4) Age: 35-54 -107 * (26.2) Age: 55-74 -122 * (27.0) % Private insured -0.18 (1.44) % Medicaid -3.92 *** (1.75) % Medicare 3.34 (3.02) % Military -6.65 *** (2.93) % Catholic -6.90 * (2.28) % Jewish -13.9 (15.5) % Other religion -2.49 ** (1.51) % Married -0.92 (1.94) % Divorced -0.03 (3.17) % Widowed -2.89 (3.40) % Separated -0.11 (6.30) % Partnered 0.45 (3.41) Centered [R.sup.2] 0.84 No. of observations 644 Instruments Helmet laws Yes Speed limits No Seat belt laws Yes Injury death controls No Fatality rates Yes Identification test Under-identification 10.12 ** Weak-instruments (b) 6.41 Over-identification 2.043 State and year fixed effects Yes Notes: Standard errors are estimated using the Huber/White/Sandwich estimator of variance. All observations are weighted by the state's population in the given year. Observations are lost due to missing years in the mortality files and fatality rates. Underidentification test uses the Kleibergen-Paap rk LM statistic, Weak instruments test uses Kleibergen-Paap rk Wald F statistic, and the Overidentification test uses the Hansen J statistic. (a) Signifies ESRD prevalence per million by state population. (b) Cragg and Donald (1993) F statistic. * p < .01, ** p < .10, *** p < .05; Stock-Yogo critical values: ([dagger]) 10%, ([double dagger]) 20%. TABLE 8 Disaggregating the Effect of Cadaveric Donations on Sub-Samples of Living Donors Models: Marginal Effect of Cadaveric Donations per 100,000 People on Living Blood-Related Non-Blood-Related Donations Donors Donors Instruments: Helmet laws and -0.09 -0.21 * speed limits (0.06) (0.05) [R.sup.2] = 0.84 [R.sup.2] = 0.82 Instruments: Helmet laws, -0.09 -0.21 * speed limits, and seat belt (0.06) (0.05) laws [R.sup.2] = 0.84 [R.sup.2] = 0.82 Controls: Include (% Obesity) -0.05 -0.19 * (a) Instruments: Helmet laws and (0.06) (0.04) speed limits [R.sup.2] = 0.85 [R.sup.2] = 0.82 Controls: Include (% Obesity) -0.05 -0.19 * (a) Instruments: Helmet laws, seat (0.06) (0.04) belt laws, and speed limits [R.sup.2] = 0.85 [R.sup.2] = 0.82 Instruments: Helmet laws, -0.06 -0.19 * speed limits, and injury (0.06) (0.05) death demo [R.sup.2] = 0.86 [R.sup.2] = 0.83 Instruments: Helmet laws, -0.06 -0.19 * speed limits, seat belt (0.06) (0.05) laws, and injury death demo [R.sup.2] = 0.86 [R.sup.2] = 0.83 Models: Marginal Effect of Cadaveric Donations per 100,000 People on Living Anonymous Spouse + Friends Donations Donors Donors Instruments: Helmet laws and -7e-3 -0.20 * speed limits (4e-3) (0.04) [R.sup.2] = 0.46 [R.sup.2] = 0.81 Instruments: Helmet laws, -8e-3 ** -0.20 * speed limits, and seat belt (4e-3) (0.04) laws [R.sup.2] = 0.46 [R.sup.2] = 0.81 Controls: Include (% Obesity) -3e-3 -0.18 * (a) Instruments: Helmet laws and (4e-3) (0.04) speed limits [R.sup.2] = 0.45 [R.sup.2] = 83 Controls: Include (% Obesity) -3e-3 -0.18 * (a) Instruments: Helmet laws, seat (4e-3) (0.04) belt laws, and speed limits [R.sup.2] = 0.45 [R.sup.2] = 0.83 Instruments: Helmet laws, -7e-3 -0.20 * speed limits, and injury (4e-3) (0.04) death demo [R.sup.2] = 0.46 [R.sup.2] = 0.81 Instruments: Helmet laws, -8e-3 ** -0.21 * speed limits, seat belt (4e-3) (0.04) laws, and injury death demo [R.sup.2] = 0.46 [R.sup.2] = 0.81 Notes: All estimation samples consist of 46 states from 1988 to 2008. The unit of observation is state x year. All observations are weighted by the state's population in the given year. (a) Standard errors, in parentheses, are estimated using the Huber/White/Sandwich estimator of variance. * p < .01, ** p < .10.
|Printer friendly Cite/link Email Feedback|
|Author:||Fernandez, Jose M.; Howard, David H.; Kroese, Lisa Stohr|
|Date:||Jul 1, 2013|
|Previous Article:||The effects of increased access to emergency contraception on sexually transmitted disease and abortion rates.|
|Next Article:||Cooperation spillovers and price competition in experimental markets.|