P a g e | 1

The Impact of Bureaucratic Structure on Government Eligibility Decisions

Lael R. Keiser
Department of Political Science and Truman School of Public Affairs
University of Missouri
Columbia, MO 65211
[email protected]

* prepared for presentation at the Public Management Research Association Conference, June
2011
** Draft do not cite without permission

P a g e | 2

Abstract
One central function of bureaucracies is to determine eligibility in many federal and state programs. The
structure of the organization in which decisions are made may affect whether the organization has a
predisposition to be more lenient or more stringent in eligibility determinations. In this paper I compare
the determinants of eligibility decisions in Social Security Disability across two different bureaus within
the Social Security Administration (Administrative Law Judge offices and state government Disability
Determination Services). I test the hypothesis that structural differences between the two units explain
why each unit responds differently to signals from the environment, to the ability to gather information,
and to cognitive biases created from the ideological environment where decisions are made. I find that
ALSs are more responsive to signals from U.S. Congress whereas DDS are more responsive to intra-
agency signals and the ability of the bureaucracy to gather information. I also find that differences in the
responses of DSS and ALJs to the task environment are predicted by the fact that only ALJs actually see
claimants face to face. Finally, both DDSs and ALJs decisions reflect the ideological context where
bureaucrats process information, which supports the argument that cognitive biases influence bureaucratic
decision-making regardless of organizational structure.
P a g e | 3

Bureaucracies play a central role in making public policy by applying program rules to
individual cases (Lipsky 1980; Prottas 1977). In so doing, they create the public policy that the
public actually experiences. An understanding of public policy, therefore, requires an
understanding of the determinants of bureaucratic behavior. While the dominant paradigm for
understanding bureaucratic behavior focuses our attention on how the incentive structures
elected officials create constrain bureaucratic behavior (principal-agent theory), scholars have
recently argued that political scientists refocus their attention on the central bureaucratic task of
information processing and return to the behavioral approach to understanding bureaucratic
organizations (Jones 2003; Workman, Jones and Jochim 2010). This approach draws from both
early work in the behavioral tradition in public administration (see March and Simon 1958;
Simon 1947) and the work on social construction and issue definition in public policy (see
Rochefort and Cobb 1994; Stone 1984).
Jones and his colleagues argue for a new theory to guide research on public
bureaucracies, which they label as an information processing theory of bureaucracy. To best
understand why public bureaucracies implement policy the way they do, we should focus on the
limited attention span of bureaucracies, the use of heuristics and assumptions to simplify the task
environment, and the distribution of issues across different sub-units within the bureaucracy
(Workman, Jones and Jochim 2010). Agencies charged with implementing programs are not
monolithic black boxes but instead are made up of sub-units all with their own structures and
cultures (Simon 1947). In order to understand why bureaucracies shape public policy the way
they do through policy implementation; we must therefore pay attention to how different units
within the bureaucracy respond differently to information in the task and political environment
and recognize that information is often ambiguous. Policy implementation by any federal
P a g e | 4

agency should reflect, therefore, how the organizational design of sub-units within the
bureaucracy shape how people within that sub-unit respond to environmental stimuli that is
filtered through the structure of the sub-unit. In what follows, I use insights from the
Information processing theory of bureaucracy to generate hypotheses about how the
organizational structure of different units within the Social Security Administration shape how
decision making in the Social Security Disability program in these units reflects various
responses to information from the political and task environment. More specifically, I compare
the impact of the task and political environment on two sub-units within the Social Security
Disability system, namely Disability Determination Offices (DDSs) and Administrative Law
Judge offices (ALJs) across the fifty U.S. states. Although both of these bureaus determine
eligibility for the Social Security Disability program, they differ in three important respects.
First, DDS offices are state government agencies whereas administrative law judge offices are
federal agencies. Second, Administrative Law Judges have more administrative independence
from the Social Security Administration than do Disability Examiners (decision-makers within
DDS). Third, Administrative Law Judges have face to face contact with clients whereas
Disability Examiners do not. Organizational theory allows the generation of hypotheses about
how these two organizations will respond differently to environmental signals and how these
signals will influence policy outputs in ways that are predicted from differences in their
organizational structure.
The Importance of Bureaucratic Structure
According to open systems theory, bureaucratic organizations interact with their
environment (Perrow 1972; Thompson 1967; Rourke 1984). However, how this interaction
takes place and what its impact is on bureaucratic decisions is mediated by the structure of the
P a g e | 5

organization. The organizational structure influences what types of stimuli from the
environment reaches individual bureaucrats and it sets constrains on bureaucratic decisions and
actions (Scott 1992). Organizational arrangements determine what types of information is
collected by the bureaucracy and who pays attention to it (Simon 1947; Jones 1991; Jones 1993).
The characteristics and goals of the organization predict what types of environmental
stimuli are salient (Scott 1992). Structure includes the rules of operation, standard operating
procedures, bureaucratic culture, levels of hierarchy and a variety of other factors. These factors
affect how bureaucracies respond to environmental pressure. Some structures should make the
bureaucracy more open to environmental influences than other structures. For example, rules
that allow interest groups and citizens to comment on bureaucratic rulemaking make bureaucratic
decision-making more responsive to interest groups (Yackee and Yackee 2006). Hiring practices
and requirements determine which profession most influence bureaucratic organizations, which
in turn influences how people within the bureaucracy make decisions (Eisner and Meier 1990;
Eisner 1993). In so far as structure mediates the relationship between the environment and the
bureaucracy, differences in bureaucratic structure should explain variation in the responsiveness
of different sub-units to similar information from the environment.
The Social Security Disability Program
The Social Security Administration implements Disability Insurance program
(DI). DI is a federal social insurance program that provides cash assistance to the
disabled. In order to qualify for DI due to a disability, potential clients must be assessed
as unable to work in the same range of jobs as non-disabled people given their age,
education and work experience (Roth 1987). In addition to the medical criteria for
P a g e | 6

disability, claimants must have contributed to the system for a particular number of
quarters. The federal government provides all of the funding for the program and
determines program rules (Dolgoff, Feldstein and Skolnik 1993).
Although the federal government provides all the funding for the program and
dictates program rules, state governments play a role by making determinations on initial
applications. State offices are referred to as Disability Determination Services (DDSs).
State governments have this administrative responsibility because program advocates
during the creation of the Social Security Disability program in 1954, mandated state
involvement as a way to appease the American Medical Association and conservatives
who did not want federal officials meddling with doctor-physician relationships and who
raised fears of socialized medicine. (Stone 1984). For a variety of reasons, the federal
government has been reluctant to take full responsibility of the eligibility determination
process despite the domination of federal financing even though the Social Security
disability program is not currently considered as socialized medicine. (Berkowitz 1987).
1

Although state governments are involved in determining eligibility, the federal
government does limit state influence by giving another bureau the power to review
appealed cases. Claimants who are denied at the initial level by the state can appeal their
decisions to administrative law judges (A.L.J.s) who work for the federal government.
ALJs hear cases de nova and are not constrained by DDS decisions. Additional evidence
is often gathered at this point (Lewin Group 2001). If a claimant is dissatisfied, they can

1
Current reformers who wish to centralize the determination process face opposition from policy actors
hesitant to increase the size of the federal bureaucracy. Hiring federal bureaucrats as disability examiners
would also displace current state employees and may increase the cost of the program because federal pay
scales are, for the most part, higher than state pay scales (Berkowitz 1987).

P a g e | 7

take their case to the Appeals Council, a centralized administrative office that reviews
A.L.J. decisions. Finally, claimants have appeal rights to federal district court.
The structure of the disability eligibility system allows us to test the hypothesis that
variation in bureaucratic structure will help to explain variation in how bureaucratic
organizations respond to their environment. By comparing the impact of environmental factors
on both Administrative Law Judge and Disability Examiner decisions, we can test the hypothesis
that structure of decision making affects how organizations respond to different information from
the environment. Both bureaus produce the same output, e.g. decisions that award or deny
applications for disability, but both have different structures.
Although many cases are clear denials or allowances, a substantial proportion of cases are
ambiguousneither clear allowances or denials (Mashaw 1983). Whether these decisions are
errors or not is not empirically verifiable and is a question more of politics and values than
science (Mashaw 1983; Stone 1984). The concept of disability resists precise definition and
measurement, giving street-level bureaucrats and bureaucracies high levels of discretion in
decision-making. (Berkowitz 1987; Derthick 1990). High levels of uncertainty exist in medical
diagnostic decisions that determine whether a patient is considered disabled.
In addition to the subjective medical information, examiners must also take into account a
claimants age and occupational history. What kinds of jobs a claimant can work in given their
physical limitations is not always easy to determine and these decisions also contain subjective
elements (Stone 1984). In sum, the statutory definition of disability is not sharp enough for us to
say with confidence that any particular decision is or is not a correct application of the statute
(Mashaw 1983). Substantial levels of disagreement exist between bureaucrats determining
P a g e | 8

eligibility. A survey of DDS examiners in three states revealed that examiners believe that
reasonable examiners might disagree in about 20% of cases (Keiser 2010).
DDS and ALJ decisions on social security provide a good case study for whether
variation in structure affects responsiveness to signals from the environment because DDS and
ALJ offices are located across the fifty states and therefore variation exists in environmental
stimuli. Decisions vary between DDS offices and ALJ offices across the fifty states. For
example, in 2000, DDS examiners in Louisiana rejected 68% of claims while DDS examiners in
Utah rejected only 38%. Similar variation exists between ALJ offices with award rates ranging
from a high of 83% in Maine to a low of 42% in Hawaii (unpublished data from the SSA). The
award rates (percent cases decided that were awarded) in both DDS and ALJ offices for the years
1990 to 1999 are dependent variables in this study. I create one model predicting award rates in
DDS offices and one predicting award rates in ALJ offices.
Since DDS and ALJ offices are geographically placed in similar environments, the
variation in decisions provides a research design to test the theory that bureaucratic structure
affects a bureaucracys openness to environmental influence. Because DDS offices and ALJ
offices differ in their structure, variables that measure environmental influence should have
different impacts. DDS offices are state agencies, have no face to face contact with claimants,
and work under a quality assurance system run by the Social Security Administration. In
contrast, ALJ offices are federal, have face to face contact with claimants, and have some
administrative independence from the Social Security Administration due to the Administrative
Procedures Act. I describe these differences in more detail below.
Important Structural Differences between Decision-making in DDS and ALJ Office
P a g e | 9

Bureau Independence
The Administrative Procedures Act (APA) in 1946 created the position administrative
law judge in the federal agencies in response to concerns that personnel adjudicating cases were
not independent of the agencies in which the ALJs worked and therefore were not fair (Lubbers
1981). To ensure more objectivity on the part of adjudicators, the Administrative Procedures
Act gave the power to the Civil Service Commission (now the Office of Personnel Management)
to determine the initial qualifications and the compensation of ALJs. Evaluations of ALJS by
OPM cannot take into account how their home agencies have rated them. ALJs can only be
removed for good cause as determined by the Merit Systems Protection Board. The APA also
required that ALJs be assigned cases in rotation to ensure their objectivity (Lubbers 1981) and
places on ALJs the expectation that they will acquire and consider all relevant information
needed to reach decisions in individual cases (Kritzer 1998). SSA employs over half of all ALJs
in the federal government (Lubbers 1981).
The protections and requirements in the APA create limitations on the ability of federal
agencies to systematically evaluate and influence the decision making of ALJs because any
systematic evaluation of ALJS potentially conflicts with ALJ independence. The federal courts
have ruled that the congressional intent of the APA was to safeguard the impartiality of the
agency hearing process but that independence was not as full as that provided to the judicial
branch by Article III of the Constitution. The APA requires in general that ALJs cannot be
placed under political pressure to compromise their integrity or impartiality but they do not
have as many protections as federal judges (Cofer 1985).
P a g e | 10

Conflict and ambiguity exists, therefore, in the extent to which ALJs in the Social
Security Administration are subject to evaluation by SSA. Throughout the 1970s and 1980s,
the SSA has attempted to impose a variety of performance evaluation criteria on ALJs to which
the ALJs have objected and threatened legal action. They have occasionally followed through
with those threats and won concessions from the SSA (see Cofer 1985).
One way in which federal agencies oversee street-level implementers is through
reviewing street-level decisions through quality assurance programs. In quality assurance
programs, employees in a quality assurance unit review a sample of street level decisions to
measure their accuracy. Inaccurate decisions are returned to the street-level for re-processing
(the Lewin Group 2001). Currently, ALJS have no formal internal quality assurance program.
Regional Chief ALJs can educate ALJs for future decisions but cannot order an ALJ to change
a decision (the Lewin Group 2001). The Appeals Council does review a random sample of ALJ
decisions separate from those that come from claimant appeals, although the percent ranged over
time from 5% to 1%. If the Appeals Council feels the ALJs decision is not supported, the case
is remanded back to the ALJ. There are no penalties assessed but remands do add to the
workload of ALJs (Lewin Group 2001).
As mentioned earlier, DDSs have some independence from the Social Security
Administration by the fact that they are state rather than federal agencies. Examiners in DDS
offices, do not, however, have the same legal right to independence as we find with the ALJs.
DDSs offices are subject to both internal and external quality assurance programs and the SSA
has greater ability to use these quality assurance programs to affect the DDSs than SSA has with
the ALJ offices.
P a g e | 11

SSA monitors DDS decisions through their Quality Assurance program. Ten regional
SSA offices review a random sample of DDS decisions in each regional Offices territory. SSA
designed the system so that Quality Assurance personnel review 70 allowances and 70 denials
per quarter, which results in about 2% of all DDS decisions (The Lewin Group 2001). Regional
SSA offices calculate accuracy rates for each DDS office based on the number of incorrect or
insufficiently supported decisions Quality Assurance find for each DDS office. (The Lewin
Group 2001). The regional offices return those cases to DDS offices for corrective action. If a
DDS offices accuracy rating falls below 90.6%, SSA must provide management and
performance support to the DDS office. Although no state DDS office has lost control over
initial decision making, if a DDS offices performance accuracy fails to improve, SSA can take
over the management of that state office (The Lewin Group 2001).
Prior to 1980, the structure of the oversight system did not favor review of either denials
or allowances. Following a period of rising allowance rates in the Social Security Disability
program, however, Congress mandated that SSA conduct Pre-effectuation Reviews (PER) of all
DDS allowances in 1980 (The Lewin Group 2001). This changed the structure of the system to
focus on allowance errors rather than denial errors since more allowances are reviewed than
denials. PER reviews are not included, however, in the calculation of each DDS offices
accuracy rates for either denials or allowances. Following the creation of the PER review,
accuracy rates for allowances in the Quality Assurance program rose but denial accuracy rates
fell (The Lewin Group 2001).
Taken as a whole, the Social Security Administration must deal with two sub-units with
varying levels of independence. While the DDSs have independence due to their status as state
agencies, they do not have as much legal independence that the ALJs enjoy due to the
P a g e | 12

professional rights given to ALJs through the APA. This independence should affect how
responsive these two sub-units are to signals from the political environment. In so far as bureau
independence protects bureaucrats from political pressure, we should expect that ALJS will be
less affected by political principals than are DDSs.
Much literature in political science and public administration has focused on the impact
of elected officials on bureaucratic behavior. A rich literature exists that explores how elected
officials can create contracts with public bureaucracies to control them (see Wood 2010 for a
review of these studies). A substantial amount of literature exists showing that bureaucratic
outputs shift with changes in which party controls Congress or the Presidency (see Wood and
Waterman 1994, for example). Bureaucracies are responsive to political signals from elected
officials.
While an information processing approach to studying the bureaucracy focuses on how
the characteristics of information and information processing shape bureaucratic behavior, it
does not contradict per se, the more dominant principal agent approach, which focuses on
political control (see Wood 2010 for a review). The main difference is that an information
processing approach treats signals from elected officials, and the incentives structures they
develop, as one piece of information competing for the attention of bureaucrats (Workman, Jones
and Jochim 2010). Rather than focusing exclusively on those signals as attempts of political
control that bureaucrats necessarily pay attention to, an information processing approach views
these signals as disturbances in the environment of bureaucracies that highlight the possibility
that something is amiss. Bureaucracies may or may not pay attention to these signals (Workman,
Jones and Jochim 2010).
P a g e | 13

A content analysis of congressional hearings between 1990 and 2000 reveals a partisan
difference in the signals that Congress sends regarding the DI program. All congressional
hearings between 1990 and 2000 were reviewed to identify statements that indicated members of
Congress believed bureaucrats were being too restrictive or too lenient in granting claims in the
program. The content analysis revealed that Republicans are much more focused on the
possibility that the bureaucracy is awarding benefits to people who should not receive them.
Republicans made statements 115 times during this time period during hearings that focused on
benefits being awarded to people who did not deserve them while Democrats only made such
statements 5 times. In contrast, Democrats expressed concern that deserving applicants were
being denied 9 times whereas Republicans expressed such concern only 3 times. While neither
Republicans nor Democrats made very many statements about unwarranted decisions, the
difference between the two is quite large. As signals in Congress become more conservative,
therefore, signals are being sent that decision makers should take care not to award cases with
ambiguous evidence. Decision makers within the SSA may or may not be paying attention to
this information. The greater independence of ALJs should, however, make them less responsive,
than the DDSs. This leads to the first hypothesis.
H1: Changes in the ideology of congressional oversight committees will have a
greater impact on DDS eligibility decisions than ALJ eligibility decisions.
To measure the signals that the bureaucracy receives from Congress, the ideology of one
of the SSAs primary oversight committees, e.g. the subcommittee on Social Security in the
Ways and Means Committee is used.
2
The ideology of the committee is measured by the average

2
Other committees are also relevant for the Social Security Administration such as the Senate
Finance Committee and the Senate Judiciary Committee. The ideology of members on Ways
P a g e | 14

score for members serving on that committee using Poole and Rosenthals Common Space
Scores (McCarty and Poole 1995). As the common space scores become more conservative
(positive), I expect that allowance rates will decrease as conservatives are more concerned with
program size than access.
The structural independence of ALJs relative to DDSs should also make ALJs less
responsive to signals from intra-agency actors. The DDSs receive three signals from intra
agency actors about their performance. First, as mentioned before, SSA uses a quality assurance
program to evaluate the DDS offices and have the ability to take over a DDS office if accuracy
rates fall below a particular level. There is very little variation in accuracy rates for allowances
but there is variation for denials. We should expect that DDSs that have low accuracy for denials
the prior year to have higher allowance rates the following year. Second, ALJs do review
decisions by DDSs. Although there are no requirements that DDSs pay attention to the number
of cases that ALJs reverse, prior research has shown that DDS examiners who are aware of the
number have higher allowance rates than those who do not (Keiser 2010). We should expect,
therefore, that DDSs that have had a greater portion of their cases overturned the prior year, will
grant more claims the following year. To test for this, the ALJ award rate the prior year is
included in the model. Third, in 1996 the Social Security Administration adopted new rules, e.g.
Process Unification, to try to reduce the discrepancy between ALJ decisions and DDS decisions.
These new rules tried to make treatment of pain more consistent between DDSs and ALJs. A
dummy variable that is coded zero prior to 1996 and 1 thereafter is included in the model to
reflect the adoption of the Process Unification Standards. This variable is logged to reflect a lag

and Means is highly correlated with other measures, as well as the ideology scores of the House,
Senate, or Congress as a whole.

P a g e | 15

for implementation timing. I also include the percent of cases that ALJs reversed the prior year
to reflect signals from ALJs that the DDSs are making erroneous decisions (source: Freedom of
Information Act Request from Social Security Administration).
Like the DDSs, ALJs also receive signals from intra-agency actors that indicate whether
they are being too generous or too restrictive in their eligibility decisions. First, applicants
denied by ALJs can appeal to a unit within the Social Security Administration called the Appeals
Council, which then sends cases back to ALJs if they disagree with the decision (or allow the
case outright). The Appeals Council also reviews some ALJ decisions by random selection and
these are also returned to the ALJs (Lewin Group 2001). Although Appeal Council remands are
not part of the performance evaluation of ALJs, they do increase the workload of the ALJs,
giving them an incentive to try to minimize them. To capture this, I include a measure of the
percent of cases appealed to the Appeals Council that the Appeals Council remands to ALJ
offices (source: FOIA request from SSA). Compared to the DDSs, however, the signals that
ALJs receive are less salient because they are not part of any performance review. Second, ALJs
also receive a signal that they should try to reduce the number of reversals they make of DDS
decisions when the SSA adopted the Process Unification standards (dummy variable described
above).
Although intra agency signals exist in the decision-making environment of the ALJs, the
structural independence of the ALJs, especially as compared to the DDS, should make us expect
that decision-making in the ALJs will be less responsive to these signals coming from intra-
agency actors than are DDS offices. This leads to the second hypothesis.
P a g e | 16

H2: At the DDS level, eligibility decisions will reflect intra-agency signals awards
but ALJ decisions will not due to the structural independence of ALJs.
Interactions with Claimants
Another structural difference between DDS offices and ALJ offices is that disability
examiners in DDS offices have no face to face interaction with claimants whereas administrative
law judges physically see claimants at administrative hearings. Face to face interactions between
clients and street level bureaucrats can lead to more responsiveness to clients. Many people
within the Social Security Administration and observers of the system have speculated that one
of the reasons for the high reversal rate of ALJs of DDS decisions is that Disability Examiners
do not see the claimant in person whereas ALJs do. Face to face interactions allow ALJs to
better gauge the legitimacy of an applicant's claim of disability and inability to work. ALJs can
also illicit more information about the claimants medical condition during the hearing. Face to
face interactions may also provide an opportunity for the claimant to evoke sympathy and to
provide more evidence to support a positive ruling.
The difference between the level of bureaucrat-claimant interaction between the DDSs
and the ALJ should result in differences between how the two sub-units respond to the task
environment, e.g. the unemployment rate in the state, the number of minorities in the potential
applicant pool and the level of self reported disability within the state.
According to the law, disability decisions are to be solely based on whether or not a
person can work in a hypothetical job in the national economy not whether the applicant could
actually get a job. However, norms to award citizens benefits when few job opportunities exist
may increase the number of awards. Policymakers have long been concerned that the disability
program would turn into an unemployment program during periods of economic downturns
P a g e | 17

(Berkowitz 1987; Stone 1984). Economic downturns make it more difficult for the marginally
disabled to find work and make it more likely that individuals previously working will become
aware of their limitations and seek benefits (Howards, Brehm, and Nagi 1980). Furthermore,
street level bureaucrats may face political pressure from state and federal elected officials to
loosen the eligibility requirements during times of economic downturn. Not surprisingly, past
research has found support for the link between the economy and disability rates (Halpern 1979;
Howards, Brehm, and Nagi 1980; Keiser 1999). We should expect that higher unemployment
rates will lead to higher award rates (Source: U.S. Statistical Abstract).
However, the face to face interaction between applicants and bureaucrats in ALJ offices
and the lack of such interaction in DDS offices should mediate the effect of unemployment on
bureaucratic outputs. Without the face to face interaction, examiners in DDS offices should be
more isolated from feelings of sympathy due to high unemployment. Bureaucratic values to
implement the law as intended could lead to a reduction in award rates in response to high
unemployment because unemployment increases the marginal cases in the caseload. A weak
economy drives more people to apply for benefits than would have in a strong economy (see
Soss and Keiser 1999). A large proportion of claimants during an economic downturn should
have more marginal cases because more potential applicants would be able to find and perform
work in a strong economy than in a weak one. To control for economic downturns, the
unemployment rate is included in the model. High unemployment should translate into lower
award rates because the caseload includes more marginal cases than it does under low
unemployment. In contrast to DDS decisions, ALJ decisions may be influenced by the face to
face interaction that takes place between applicant and bureaucrat and we should expect the
P a g e | 18

opposite relationship. ALJ offices in states with high unemployment should have higher award
rates than ALJ offices in areas with low unemployment. This leads to the 3
rd
hypothesis.
H3: In U.S. states with high unemployment, DDS award rates will be lower than in states
with low unemployment. ALJ award rates will be higher in states with high unemployment
and lower in states with lower unemployment.
In so far as the face to face interaction helps to illuminate disability beyond that
communicated in the paper record, ALJ decisions may be more responsive to how state residents
identify as disabled. Identifying as disabled is subjective and one person may find their
medical condition disabling while another person with the same condition will not (Stone 1984).
The rules for eligibility for DI are actually very strict because claimants must be deemed as
unable to work in hypothetical jobs given their education and age. They do not actually need to
be able to be hired to work in that job. Presumably, claimants with ambiguous claims who
believe that their medical conditions are disabling will be able to make a more convincing case in
person than they would on a paper application. We should expect that states with a higher
proportion of people who believe they have a disabling condition will have a greater number of
people with marginal cases apply for benefits.
The lack of face to face interactions at the DDS level should make it more likely that
DDS Examiners will interpret these claims more stringently than they will ALJs. I include the
percent of the state population who report a disabling condition to test this hypothesis. The
number of people in a state with a self reported disabling condition is a measure of disability in
the fifty states created by the Cornell University Rehabilitation Research and Training Center on
Disability Demographics and Statistics (StatsRRTC), which measures whether people have
P a g e | 19

disabilities that limit their ability to work.
3
Because of the lack of face to face interactions at the
DDS level, we should not expect DDS offices in states with high self reported disability to have
higher allowance rates. At the ALJ level, however, claimant beliefs in their disability should
translate to a higher allowance rate because ALJs will respond to the demeanor and physical
condition that they can visibly see (or not see) in the hearing. This leads to the fourth
hypothesis.
H4: DDS decisions in U.S. states with high self reported disabilities should be lower
or no different than states with lower self reported disabilities. ALJ decisions in states with
high self reported disabilities should be higher than in states where residents are less likely
to report disabilities.
Finally, the face to face encounter may also affect the impact of racial composition of
state residents on disability decisions. Face to face interactions with clients should be more
likely to bring the issue of race and ethnicity to the forefront. In other words, it should highlight
race as a client characteristic more than does a paper application. Minorities also
systematically have less access to health care, and therefore, may have less well developed
medical records. Since ALJs physically see claimants, they may be more likely to ask the
claimant to provide medical information. Because the face to face interaction of between
claimants and bureaucrats at the ALJ level, we should expect that the racial composition of the
state population should have an impact on decision-making at the ALJ level but not the DDS

3
This measure was created using data from the Current Population Survey (CPS) conducted by
the Census Bureau and Bureau of Labor Statistics (Houtenville, Erickson, and Lee 2005). This
measure is based on responses to a question asking whether someone in the household had a
disability that prevents them for working or limits the kind or amount of work they can do. This
question has appeared consistently on the CPS since 1981 and provides a reliable measure of
work limitation disability over time and across the fifty U.S. states (Burkhauser, Houtenville, and
Wittenburg 2002). This measure corresponds to other commonly used measures of disabled
populations (Burkhauser, Daly, Houtenville and Nargis 2002).

P a g e | 20

level. There are reasons to expect, however, that the impact will be either positive or negative.
In so far as ALJs are discriminatory, a larger minority presence in the caseload should reduce the
award rate. If, however, the face to face conflict results in more sympathy for the claimant or a
desire to gather more information from people with poor quality medical records, we should
expect a larger minority presence in the caseload to increase award rates. To measure the
minority presence in the state, I use the percent of state residents who are African American or
Hispanic (source: U.S. Statistical Abstract).
H5: Award rates in DDSs offices should be lower in states with high minority populations
but higher in ALJ offices.
The Role of Cognitive and Organizational Biases
In addition to bureaucratic structure, how information from the external
environment is perceived is influenced by the cognitive biases held by decision-makers. These
include ideology, professional identification and material interests (Jones 2001; Workman, Jones
and Jochim 2010). Street-level bureaucrats implementing many federal programs like the Social
Security Disability Insurance program interpret program rules and collect and process
information while located in different state political environments. The jurisdictional boundaries
of U.S. states shape the identities of the people who live within them (Gimpel and Schuknecht
2004, 6). People living in the same state develop shared understandings of the role of
government in citizens lives (Elazar 1984). Just as the U.S. state in which individuals live
affects their political behavior and attitudes, the location that bureaucrats live may affect their
policy behavior (Kaufman 1960).
P a g e | 21

The state political environment not only shapes the attitudes and behaviors of individual
bureaucrats, it also helps shape the attributes of bureaucratic offices as a whole (Thompson 1967;
Goggin et al. 1990). Individuals interacting within bureaucratic organizations create standard
operating procedures and communication patterns that affect all individuals working within that
organization regardless of their individual attributes (Simon 1947). Patterns of information
processing within bureaucracies can make individuals working in one bureaucracy react
differently to the same information as do individuals working in a different bureaucracy (Simon
1947; Jones 2001). Bureaucracies develop cultures, which James Q. Wilson (1989, 93)
defines as patterned and enduring differences among systems of coordinated action that lead
those systems to respond in different ways to the same stimuli.
When bureaucracies implementing federal programs are located in ideological
environments, those patterned and enduring differences should reflect the ideological
environment and shape how street-level bureaucrats interpret program rules and view clients.
In other words, bureaucrats operating in different U.S. states may respond differently to the same
objective information (Goggin et al. 1990). One important factor in the environment for social
welfare bureaucracies is the general ideology of the state. Ideology shapes preferences toward
the size of the welfare state and whether people are more or less concerned with denying
deserving applicants or allowing undeserving ones. Although attitudes about the welfare
state are complex (Feldman and Zaller 1992), actors in more liberal environments should be
more likely to prefer to give the benefit of the doubt to claimants than those in conservative
environments since liberals tend to be less concerned with program size and more concerned
with access. In contrast, actors in conservative environments should be more concerned with
P a g e | 22

program size and less concerned with access (Stone 1984; Schneider and Jacoby 2005). These
state political environments should create biases in the bureaucracies that operate within them.
It is important to note that the ideology of the state in which the DDSs and the ALJs
make decision is not a measure of political control. State governments across the board have an
interest to maximize awards because doing so increases the number of state residents with
disposable income and relieves demand on programs for which state governments pay.
Maximizing award rates helps to provide state residents with assistance without costing state
governments anything. The ideology of the citizens within the state is more reflective,
therefore, of general orientations toward social welfare programs rather than the utility
maximizing behavior of state governments to affect program generosity. All state governments
have an incentive to increase award rates, and therefore, this incentive is held constant across the
U.S. states.
Whether or not we should expect both ALJs and DDSs decision making to reflect
environmental biases toward the welfare state depends on the extent to which the
professionalization of legal personnel reduce the impact of such biases. This is lively debate
within scholarship on judicial politics and jurisprudence. In so far as the professional
socialization of legal personnel minimizes the impact of the local political culture, we should
expect that decision making at the ALJ level will not reflect state political culture whereas it will
at the DDS level. To measure the political ideology of state citizens, I use Berry et al.s (1998)
measure of citizen ideology is used to measure the ideological environment. This variable is
subtracted from 200 so that high values indicate that a state is conservative rather than liberal so
that different measures of ideology are all consistent in the paper. The citizen ideology of the
state serves as a proxy for general political environment in each state.
P a g e | 23

H6: DDS offices in states with more conservative political ideology will have lower
allowance rates than those in more liberal states. No differences will exist between ALJs
operating in different ideological environments.
The Ability to Gather Information
Finally, in addition to the task and political environment, as well as biases within
decision-makers, decision-making within bureaucracies will reflect the ability of the bureaucracy
to gather information (Simon 1947). The rules of the disability program and information from
personal interviews suggest that when decisions are made with less attention and more quickly,
bureaucrats are more likely to deny claims. Many applicants of Social Security Disability have
multiple ailments that together make it impossible for the applicant to work. There are multiple
characteristics that street-level bureaucrats may investigate if they choose (personal interview
with DDS medical consultant, October 15, 2000). Insofar as awards require more information,
we might expect that examiners who take longer to process claims will have higher award rates
because those decisions were reached with less information. This leads to the final hypothesis.
H7: The more decisions made per decision maker the lower the award rate. This should
apply to both DDS and ALJ decisions.
To measure the ability to collect information, I use the number of decisions made per
decision maker and expect that more decisions per decision-maker will lead to fewer awards.
Because the ALJs are reviewing decisions already processed by the DDSs, I also control for the
number of cases decided per examiner at the DDS level in the ALJ model. ALJs that review less
developed cases from the DDSs should be more likely to reverse those decisions because they
were decided on less complete information. To capture information collection ability I use the
P a g e | 24

number of decisions per FTE at the DDS level and number of dispositions per ALJ in the ALJ
model (source: FOIA request from the Social Security Administration).
Research Design
I use a cross sectional time series model to estimate the impact of decision- making
resources on program generosity across the U.S. states for the years 1990 to 2000.
4
Time Series
Cross Section data analysis is notoriously challenging because the modeler must deal with
problems associated with cross-sectional units as well as dynamic units (Beck and Katz 1995).
Given the properties of the data, I present the O.L.S. model with a lagged dependent variable and
panel corrected standard errors, which correct for groupwise heteroscedasticity and
contemporaneous correlation of the errors (Beck and Katz 1995). The use of a lagged dependent
variable is appropriate because organizational theory strongly suggests that bureaucratic behavior
is dynamic since bureaucracies develop standard operating procedures and have programmed
decision rules that prevent bureaucrats from radically changing their behavior (March and Simon
1958; Simon 1947; Jones 2001). Because decisions at the ALJ level are shaped by earlier
decisions at the DDS level, I include a measure of the percent of claimants denied at the DDS
level who appeal. This should control for the fact that ALJs in areas where DDSs are very
generous will face a different applicant pool than states where the DDSs were very strict in their
interpretation of eligibility rules. This variable was used rather than the award rate of the DDSs
because the award rate of the DDSs is endogenous with many other variables in the model.
Findings and Implications
(Table One about here)

4
Because of missing data, North Carolina, South Carolina and South Dakota are excluded from
the analysis.
P a g e | 25

Table one presents the preliminary findings for both the DDS and ALJ models. The
findings do not support the argument that the structural independence of the ALJs isolates them
from signals from the political environment. Although the DDSs do not seem to respond to
political signals from Congress, the ALJs do. The more conservative the oversight
subcommittee, the lower the award rate for ALJs. The variable measuring congressional
subcommittee ideology fails to reach significance in the DDS model and is much substantively
smaller than in the ALJ model.
When it comes to signals from intra-agency sources, however, the DDSs are more
responsive. The coefficient for the accuracy rate and the change in rules following Process
Unification all reach statistically significance in the DDS model. In the ALJ model, in contrast,
intra-agency signals from the adoption of Process Unification and remands from the Appeals
Council fail to reach statistical significance. In sum, decision making at the DDS level seems
unresponsive to signals from elected officials but responsive to intra-agency signals whereas
decision-making at the ALJ level seems responsive to signals from elected officials but not from
intra-agency actors.
The findings do show support for the hypothesis that the face to face interactions at the
ALJ level and the lack thereof at the DDS level will affect how both units respond to the task
environment with some caveats. The size of the minority population depresses award rates at the
DDS level but raises them at the ALJ level. This supports the view that ALJS are able to use
their face to face encounters to overcome some of the disadvantages that minority claimants have
in case development and/or that ALJs are more sympathetic to the historic discrimination of
minorities in the workplace. The variable measuring the self reported disability rate of residents
in the states also meets expectations. ALJs in states with more residents with self reported
P a g e | 26

disabilities grant a higher percent of claims whereas DDSs report a lower percent. Unlike the
variables measuring minorities and self reported disability, the unemployment rate failed to meet
expectations. It does not reach statistical significance in the ALJ or DDS models suggesting that
the SSA has been successful in ensuring the Disability Insurance program does not turn into an
unemployment program.
The findings also show support for the hypothesis that the cognitive biases from the
political environment where bureaucrats made decisions has an impact on award rates. DDS
offices in states with more conservative ideology, have lower award rates than those in more
liberal states. Despite the professionalization of ALJs to treat cases individually without bias,
the political ideology of the state also seems to affect ALJ decisions. Taken together, the results
support the view that the cognitive biases of decision makers in the bureaucracy affect how they
interpret information from their caseload and that ALJs are not immune to that.
The results also support the argument that the ability of the bureaucracy to gather
information affects bureaucratic decisions. In DDS offices where examiners are deciding more
cases each year, award rates are lower. Although a similar pattern exists for ALJ decision-
making, the variable measuring decisions per ALJ fails to reach statistical significance.
Interestingly, ALJ decision making is responsive to the ability of bureaucrats at the DDS level to
gather information. More cases that were initially denied at the DDS level are approved at the
ALJ level in states where examiners are deciding more cases per person than when they are
deciding less. This suggests that ALJs are overturning cases that were perhaps decided without
full documentation at the lower level.
P a g e | 27

Overall the findings support the argument that different units within a bureaucracy
respond to signals from the environment differently and that some of those differences are what
we would expect given the structural characteristics of the program. The amount of face to face
interaction with clients seems to affect how responsive bureaucrats are to particular factors in the
task environment such as the size of the minority population and the self reported disability rate
of residents. Similarly, I find no evidence that ALJs are responsive to signals from intra-agency
actors while I do find such evidence for DDSs, which is consistent with the greater independence
that ALJs have from SSA due to the protections from the APA.
While the findings regarding the impact of signals from elected officials was unexpected
(ALJ responsiveness to signals from Congress and a lack of responsiveness in the DDS), it is
consistent with the purpose of the APA, which was to protect ALJs from influence from the
agencies in which they worked (Cofer 1985). It is possible that the ALJs are responsive to
Congress so that they can find allies outside the agency to protect themselves from interference
from SSA. While the APA provides protection, it is ambiguous and the ALJs have had to appeal
to outside allies like the federal courts and Congress to protect their independence.
It is important to note that these findings are preliminary and should be interpreted with
caution because several methodological issues are yet to be resolved. First, decision making at
the ALJ level is affected by DDS decision-making. While ALJs hear cases de nova and collect
new evidence, the applications they review are affected by what decisions the DDSs reached
earlier. Although this was controlled for by using the percent of DDS denials that claimants
appealed, that variable may have introduced endogeneity into the model. Future research should
better deal with the connection between DDS and ALJ decisions in the design of the study.
These preliminary results suggest, however, that units within federal agencies respond differently
P a g e | 28

to similar signals from the environment in ways that are, in some cases, predicted by differences
in the structural characteristics of those units.


P a g e | 29

References

Beck, Nathanial and Jonathan n. Katz. "What to Do and (Not Do) with Time Series Cross
Section Data." American Political Science Review 89 (1995): 6334-647.

Berkowitz, Edward. Disabled Policy. Cambridge: Cambridge University Press, 1987.

Cofer, Donna Price. Judges, Bureaucrats, and the Question of Independence. Westport, CT:
Greenwood Press, 1985.


Derthick, Martha. Agency under Stress: The Social Security Administration in American
Government. Washington D.C: The Brookings Institution, 1990.

Dolgoff, Ralph, Donald Feldstein, and Skolnik. Louise. Understanding Social Welfare. 3rd ed.
New York: Longman, 1993.

Dow, Michael G. , and Timothy L. Boaz. "Assisting Clients of Community Mental Health
Centers to Secure Ssi Benefits: A Controlled Evaluation." Community Mental Health
Journal 30, no. 5 ( 1994): 429-38.

Eisner, Mark Allen, and Kenneth. J. Meier. "Presidential Control Versus Bureaucratic Power:
Explaining the Reagan Revolution in Antitrust."." American Journal of Political Science
34, no. 1 (1990): 269-187.

Elazar, Daniel J. American Federalism: A View from the States. 3rd ed. New York: Harper &
Row, 1984.

Feldman, Stanley, and John Zaller. " The Political Culture of Ambivalence: Ideological
Responses to the Welfare State." American Journal of Political Science 36, no. 1 ( Feb.,
1992): 268-307.

Goggin, Malcom L., Ann Bowman, James P. Lester, and Laurence J. Jr. O'Toole.
Implementation Theory and Practice: Toward a Third Generation. Glenview, Illinois:
Foresman/Little, Brown Higher Education, 1990.

Jones, B. D. (2003). "Bounded Rationality and Political Science: Lessons from Public
Administration and Public Policy." Journal of Public Administration Research and
Theory 13(4): 395-412.
Keiser, Lael R. "Understanding Street-Level Bureaucrats' Decision Making: Determining
Eligibility in the Social Security Disability Program." Public Administration Review 70,
no. 2: 247-57.

P a g e | 30


Keiser, Lael R. "State Bureaucratic Discretion and the Administration of Social Welfare
Programs: The Case of Social Security Disability." Journal of Public Administration
Research and Theory 9, no. 1 (1999): 87-106.

Kritzer, B.

Lewin Group. "Evaluation of Ssa's Disability Quality Assurance (Qa Processes) and
Development of Qa Options That Will Support the Long Term Management of the
Disability Program." 2001.

Lipsky, Michael. 1980. Street Level Bureaucrats: Dilemmas of the Individual in Public Services.
New York: Russell Sage Foundation.

Lubbers.

Mashaw, Jerry L. Bureaucratic Justice: Managing Social Security Disability Claims. New
Haven: Yale University Press., 1983.

McCarty, Nolan M., and Keith T. Poole. "Veto Power and Legislation: An Empirical Analysis of
Executive and Legislative Bargaining from 1961 to 1986." Journal of Law and
Economic Organization 11, no. 2 (1995): 282-312.

Perrow, Charles. Complex Organizations. Glenview, IL: Scott, Foresman, 1972.

Prottas, Jeffrey Manditch. 1979. People-Processing: The Street-Level Bureaucrat in Public
Service Bureaucracies. Lexington: Lexington Books.

Rochefort, David A. and Roger W. Cobb. The Politics of Problem Definition: Shaping the Policy
Agenda. Lawrence, KS: University Press of Kansas, 1994.

Roth, William. "Disabilities: Physical." In Encyclopedia of Social Work. Silver Spring, MD:
NASW, 1987.

Rourke, Francis E. Bureaucracy, Politics and Public Policy. 3rd ed. Boston: Little, Brown, 1984.

Scott, Richard and John W. Meyer. "The Organization of Societal Sectors: Propositions and
Early Evidence." In The New Institutionalism in Organizational Analysis, edited by
Walter W. Powell and Paul J. DiMaggio. Chicago: Chicago University Press, 1991.


Simon, Herbert A. Administrative Behavior: A Study of Decision-Making Processees in
Administrative Organizations. New York: The Free Press, 1947.

P a g e | 31

Stone, Deborah. The Disabled State. Philadelphia, PA: Temple University Press, 1984.


Thompson, J. D. (1967). Organizations in Action: Social Science Bases of Administrative
Theory. New York, McGraw-Hill Book Company.

Wilson, James Q. Bureaucracy: What Government Agencies Do and Why They Do It: Basic
Books, 1989.

Wood, B. Dan. "Agency Theory and the Bureaucracy " In Handbook of American Bureaucracy,
edited by Robert F. Durant, 181-206. New York: Oxford University Press, 2010.

Wood, B. Dan, and Richard W. Waterman. Bureaucratic Dynamics: The Role of Bureaucracy in
a Democracy. Boulder, CO: Westview Press, 1994.

Workman, Samual, Bryan D. Jones, and Ashley E. Jochim. 2010. "Policymaking, Bureaucratic
Discretion, and Overhead Democracy." In Handbook of American Bureaucracy, edited
by Robert F. Durant, 612-37. New York Oxford University Press.



P a g e | 32

Table One
DDS Award
Rate
ALJ Award
Rate

Variables coefficient S.E. Coefficient S.E
Political
Signals

Congressional
Oversight
-3.40 6.57 -27.47 7.27**

Task
Environment

Unemployment -0.24 0.23 -0.56 0.16
% with
Disability

-0.56

0.13**

0.28

0.14**
% Minority -0.08 0.03** 0.05 0.03**

Intra-Agency
Signals

Process
Unification

1.89

1.12*

Denial
Accuracy

0.86

0.33

N/A

Appeals
Council
Remands


N/A


Cognitive
Biases

Citizen
Ideology

-0.52

0.01**

-0.47

0.03*

Quality of
Information

Decisions per
decision-maker
(DDS)


0.05


0.02**


0.05


0.02**
Decisions per
decision-maker
(ALJ)


N/A





-0.009


0.008

Award Rate
t-1
0.76 0.57** 0.75 0.67**
N 423
R2 .83
** p<.05
*p<.10

P a g e | 33