Sunteți pe pagina 1din 27

LAUREN E.

SILVER
Causal Conditions Associated with Variation in State Adoption of Body Mass Inde
Measure!ent Le"islation#
$roposal for a %ualitati&e Co!parati&e Analysis
Lauren Silver
December 17, 2008
1
LAUREN E. SILVER
'.( IN)R*+UC)I*N AN+ UN+ERS)AN+IN, *- ).E $R*BLEM
Prominent federal and private entitiesincluding the ffice of the Surgeon !eneral, the
"nstitute of #edicine, the $enters for Disease $ontrol and Prevention, and a number of ph%sician
specialt% groups and ma&or health associationshave declared over'eight and obesit% in both
children and adults a ma&or public health problem in the (nited States )*oplan, Liverman,
*raa+, 200,- (.S. Department of /ealth and /uman Services, 2001- (.S. Department of /ealth
and /uman Services, 20080. 1oth the prevalence and incidence of over'eight and obesit% in the
(.S. has have been increasing for the past three decadesif current trends persist, nearl% one2
half of adults and one23uarter of children in the (.S. 'ill be obese b% 201, )4ang 5 1e%doun,
20070.
$ompared to their normal2'eight counterparts, over'eight and obese children and adults
are more li+el% to develop cardiovascular diseases, pulmonar% disease, and t%pe "" diabetes
)!regg et al., 200,- *uman%i+a et al., 20080. "n addition, obesit% has been lin+ed to certain t%pes
of cancer, osteoarthritis, gastroesphageal reflu6 disease, liver disease, emplo%ee abseentism, and
reduced 3ualit% of life )*uman%i+a et al., 20080. 7s a result, per capita spending on health care
for obese people is appro6imatel% 87 percent higher than spending among those 'ho are normal2
'eight )9horpe, :lorence, /o'ard, ;os+i, 200<0. 9hus, the rise in obesit% prevalence and related
rise in health care spending have been ma&or factors contributing to a significant increase in
overall health care spending in the (.S., accounting for appro6imatel% 27 percent of the overall
gro'th in spending from 1=87 to 2001 )9horpe, et al., 200<0.
'.' State/Le&el $olicy Efforts to Co!0at Childhood *0esity
2
LAUREN E. SILVER
1% fiscal %ear 2010, <8 states are e6pected to face serious budget deficits, estimated to
reach a total of >100 billion. )#c?ichol 5 Lav, 20080 $onsidering that state2level estimates of
annual medical e6penditures attributable to obesit% range from >87 to >1.7 billion ):in+elstein, et
al., 200<0, cutting health care costs accruing from the impact of obesit% and co2morbid
conditions presents a significant opportunit% to reign in health spending over the long term and
ultimatel% help alleviate state budget shortfalls.
"ndeed, man% states have considered implementing a range of obesit%2related policies and
interventions addressing rising childhood obesit% rates. :rom 2008 to 200,, 88 states considered
a total of appro6imatel% 700 childhood obesit% prevention bills and passed about 17 percent
)1oehmer, Lu+e, /aire2;oshu, 1ates, 1ro'nson, 20080. Legislation addressing 'al+ing or bi+ing
trails, safe routes to school, model school policies, and state'ide studies or tas+ forces comprised
the ma&orit% of bills that states passed )1oehmer, et al., 20080.
'.1 Body Mass Inde Measure!ent in $u0lic Schools
Schools offer an ideal setting in 'hich to encourage health% lifest%le choices among
children and adolescents that 'ill remain 'ith them throughout adulthood )/a%ne, et al., 200<0.
9o that end, measuring the bod% mass inde6 )1#"0 of public school students on a regular basis is
%et another polic% that has been proposed although adopted
1
on a much smaller scale compared
to alternative initiatives )1ro'nson, 1oehmer, /aire2;oshu, Dreisinger, 2007- $a'le% 5 Liu,
20080. :rom 2008 through 200,, 1#"2related legislation comprised onl% , percent of the total
number of childhood obesit% prevention bills introduced in state legislatures, and of those
introduced, onl% 22 percent 'ere enacted )1ro'nson, et al., 20070. 4ithout in2school 1#"2
1
9he terms @enactA, @passA, and @adoptA are used interchangeabl% to impl% legislation that has officiall% been passed
b% a state legislature, regardless of 'hether the legislation actuall% has been implemented as intended.
8
LAUREN E. SILVER
related measurement initiatives in place, states 'ill lac+ the necessar% data to lin+ changes in
childhood obesit% to changes in school2 and communit%2level initiatives and programs. 9hus, this
t%pe of legislation is necessar% to ensure that ph%sical activit% and nutrition programs in schools
are evaluated and modified accordingl% to improve effectiveness. "t is un+no'n, ho'ever,
'hether certain bill2 and state2level characteristics are associated 'ith this t%pe of legislation, per
se. "f such factors can be identified, then polic%ma+ers 'ill be in a better position to craft
legislationat opportune timesthat ma6imiBes chances for adoption.
1.( *B2EC)IVES *- $R*$*SE+ S)U+3
1.' ,aps in the Literature
Cesearch to date addressing variabilit% in state enactment of childhood obesit% prevention
legislation is sparse at best. $a'le% and Liu )20080 used probit regression techni3ues to anal%Be
state2level factors associated 'ith the enactment of childhood obesit% bills addressing ph%sical
education and nutrition in schools, 1#" reporting b% schools, and health education. Dmplo%ing a
frame'or+ that categoriBed independent variables b% state health, political, and socioeconomic
characteristics, the authors found that )p2value E 0.0, or stronger0 childhood obesit% prevention
legislation is more li+el% to be enacted in states 'ith a Democratic governor, a legislature not
controlled b% Cepublicans, and in states 'ith adults reporting a larger gap bet'een their current
and reported ideal 'eight )$a'le% 5 Liu, 20080. $hildhood obesit% prevention legislation 'as
less li+el% to be enacted in states 'ith high adult obesit% rates and high per capita incomes
)$a'le% 5 Liu, 20080. Li+e'ise, 1oehmer, et al. )20080 e6amined state2level as 'ell as bill2
level factors associated 'ith enactment. (sing multi2level logistic regression anal%ses, the
authors conclude that enactment is more li+el% in states 'ith Democratic control of both
<
LAUREN E. SILVER
chambers of the state legislature and 'ith high high school non2completion rates- and if a bill has
bipartisan support, addresses 'al+ingFbi+ing trails, model school policies, and safe routes to
school, and is proposed as part of a larger state budget bill )p2value E 0.0, or stronger0.
Studies addressing factors associated 'ith state enactment of legislation re3uiring 1#"
measurement in schools, per se, are virtuall% non2e6istent. 9hus, the current stud% proposes to
use 3ualitative comparative anal%sis )G$70 to determine the state2 and bill2level characteristics
associated 'ith state enactment of this t%pe of legislation.
1.1. )heory +e&elop!ent
Gualitative comparative anal%sis is a useful approach for refining and developing ne'
h%potheses regarding social and political phenomena that are characteriBed b% causal comple6it%
)Cihou6, 2008- Comme, 1==,- $#P7SSS0. 9he proposed stud% see+s to develop refined
h%potheses regarding variation in state enactment of 1#" measurement legislation b% identif%ing
the uni3ue combinations of causal conditions
2
that result in enactment. "n contrast to the
3uantitative studies described above, this stud% does not see+ to identif% the distinct impact of
multiple independent variables, but see+s to inductivel% arrive at combinations of causal
conditions that result in the outcome of interest.
4.(. +A)A AN+ ME).*+S
9he current stud% proposes t'o, related 3ualitative comparative anal%ses that 'ill
identif% combinations of conditions, at the state2 and bill2level, that are associated 'ith state
adoption of 1#" measurement legislation. 7ll anal%ses 'ill be carried out 'ith the latest version
of G$7 soft'are.
4.' %ualitati&e Co!parati&e Analysis
2
"n 'idel% accepted G$7 terminolog%, causal conditions are s%non%mous 'ith independent variables.
,
LAUREN E. SILVER
Gualitative comparative anal%sis )G$70, developed b% $harles Cagin, is a cross2case
research approach that utiliBes the logic of 1oolean algebra to decode the causal comple6it% b%
'hich social and political phenomena unfold. 9o that end, G$7 see+s to minimiBe the inherent
difficulties in dra'ing causal inferences from observational datamost notabl% that @naturall%
occurring data lac+ sufficient variet% to permit e6periment2li+e comparative anal%sesA )Cagin,
1=870. 7s a result, researchers usuall% are able to e6amine onl% a subset of all the possible
logical combinations of causes that lead to an outcome of interest )Cagin, 1=870.
7dditionall%, G$7 is a fitting research approach 'hen there is a small or intermediate
total number of casesthat is, 'hen there are too fe' cases for appl%ing statistical techni3ues
)Cagin,1=87- Cihou6, 200H0 and too man% cases for conducting in2depth 'ithin2case anal%ses.
Since it can handle an intermediate number of cases, G$7 presents the opportunit% to produce
generaliBations, a shortcoming of alternative case2oriented methods that rel% on onl% a fe' cases
)Cihou6, 200H0.
8.1.1 G$7 7pplied to the $urrent Stud%
1ased on the causal assumptions regarding variation in state adoption of 1#"
measurement legislation )described belo'0, the current stud% proposes using 3ualitative
comparative anal%sis to address the research 3uestion of interest. Specificall%, G$7 is a useful
approach for the current stud% because it accounts for the causal comple6it% inherent in 'h%
some states adopt and others do not b% identif%ing combinations of conditions related to a
dichotomous outcome, 'hich 'ill provide useful information for polic%ma+ers, health
organiBations, and other sta+eholders 'ho are loo+ing to target their obesit% prevention efforts to
receptive polic% communities.
H
LAUREN E. SILVER
Cagin )1=870 argues that changes in social and political phenomena result not from
changes in an% one independent variable, but from the intersection of multiple conditions
'hereb% an% alteration in the resulting causal combination fails to produce the phenomenon of
interest )Cagin, 1=870. 4hen a range of different combinations of conditions result in social or
political change, then that phenomenon results from con&unctural causationa +e% feature of
causal comple6it% )Cagin, 1=870. Cagin )1=870 stresses that G$7, therefore, is @consistent 'ith
commonsense notions about ho' the 'orld 'or+s.A D6isting empirical and theoretical
+no'ledge regarding 'h% states adopt an% t%pe of childhood obesit% prevention legislation
paints a comple6 picture that is consistent 'ith con&unctural causation. #ultiple causal
conditions interact at both the state2 and bill2level. 9herefore, it is highl% li+el% that the causal
conditions leading to state adoption of 1#" measurement legislation are no less comple6 in their
interactionsand are li+el% even more comple6 given the controversial nature of 1#"
measurement proposals compared to other, more neutral obesit% prevention initiatives.
8.1.2 Stages of G$7 7nal%sis
Scholars present a number of slightl% different variations on the steps involved in conducting a
G$7 anal%sis )$overdill 5 :inla%, 1==,- Cihou6, 2008, Comme, 1==,, Cagin, 1=8<0. 9he
sections belo' present the proposed stud% according to a modified se3uence that borro's from
each variation. /o'ever, the proposed process for the current stud% is not framed in terms of
e6plicit stages given the iterative nature of G$7 )$#P7SSS0.
4.1 $roposed %CA Analysis
8.2.1 Developing Preliminar% $oncepts and /%potheses
7
LAUREN E. SILVER
Prior to conducting a G$7 anal%sis, researchers must carr% out an e6tensive amount of
@pre2G$7 'or+A )$overdill 5 :inla%, 1==,- $#P7SSS0. 9heoretical and empirical +no'ledge
of the phenomenon under investigation should drive the selection of causal conditions, thus, it is
imperative to search the literature for studies as 'ell as begin to develop an @intimac%A 'ith
potential cases )$#P7SSS0. Cesearchers should limit the total number of causal conditions
based on their relevance and the number of available cases )$#P7SSS0.
7dcoc+ and $ollier )20010 assert that measurement validit% should be assessed according
to the level of agreement bet'een s%stematiBed concepts, indicators of those concepts, and
coding outcomes of cases. "n other 'ords, ho' concepts are operationaliBed ultimatel% impacts
to 'hat e6tent the coding of cases validl% represents the concepts the researcher intends to
measure. 9o that end, this stud% dra's on the statistical anal%ses conducted b% 1oehmer et al.
)20080 and $a'le% and Liu )20080, in addition to multiple sources in the literature to develop
s%stematiBed concepts at the state2 and bill2level of anal%sis. 7 preliminar% scan of available
cases provided conceptual information as 'ell. (ltimatel%, causal conditions 'ere selected based
on e6isting h%potheses found in the literature and on information gleaned from e6amining a
select number of cases demonstrating all possible outcomes.
8.2.1.1 State-level concepts
*0esity as a statewide pu0lic health pro0le!. 9o date, there is no evidence that states
'ith high adult obesit% rates are more li+el% to enact childhood obesit% prevention legislation
)1oehmer et al., 2008- $a'le% 5 Liu, 20080. 9his stud%, ho'ever, incorporates the prevalence of
adult obesit% in addition to state2level obesit%2attributable medical e6penditures for both the
stateIs total population and #edicaid population into the larger concept of @obesit% as a state'ide
public health problem.A "n the realm of public health, polic%ma+ers often perceive a public
8
LAUREN E. SILVER
problem in terms of the cost accrued b% the state )e.g., spending on illness resulting from specific
diseases0. 4hen 7r+ansas public officials passed the first bill re3uiring state'ide 1#"
measurement in schools in 2008, the rise in the stateIs #edicaid spending 'as cited as one of a
fe' ma&or contributing factors )C%an, et al., 200H0. "ncluding state spending attributable to
obesit% in the proposed anal%sis attempts to capture this dimension of ho' polic%ma+ers ma%
consider obesit% an agenda2'orth% problem. "ndeed, in t'o other 'ell2+no'n public health
issuestobacco control and seat belt enforcementestimates of state costs attributable to
tobacco use and hospital inpatient charges for non2belted passengers have influenced
polic%ma+ersI decisions to increase cigarette prices and enact primar% enforcement seatbelt la's
)Dconomos, et al., 20010. "n fact, Dconomos, et al. )20010 claim that @it could be argued 'hether
an% social transformationJcan ever be successful unless the prevention benefits can be 3ualified
in dollar terms.A verall, the @problem environmentA influences man% polic% decisions )?ice,
1==<0.
$artisan control. Prior research demonstrates that political part% d%namics impact the
li+elihood of states adopting childhood obesit% prevention legislation. ?amel%, states 'ith
legislatures controlled b% Democrats andFor 'ith a Democratic governor are more li+el% to adopt
)1oehmer et al., 2008- $a'le% 5 Liu, 20080. 9hese findings are consistent 'ith broader themes
from the literature. 4eber )1===0 contends that the @politico2economicA environment of a state is
one of three crucial internal state factors affecting polic% decisionsCepublican control results
in policies that reduce the scope of government 'hile Democratic control results in policies that
e6pand the scope of government. 9hus, both theor% and empirical +no'ledge indicate that
partisan control is li+el% to influence 'hether states adopt 1#" measurement legislation.
=
LAUREN E. SILVER
$olicy entrepreneur. *ingdon )1=8<0 defines polic% entrepreneurs as @advocates for
proposals or for prominence of ideas.A 9heir uni3ue role relative to others in the polic%
communit% is their 'illingness to invest their time, energ%, reputation, and even mone% in the
hope of a future return. "n short, the% are @central figures in the dramaA and enhance a proposalIs
chances for moving up on the polic% agenda )*ingdon, 1=8<0.
"n 2008, 7r+ansas became the first state to pass legislation re3uiring that schools measure
studentsI bod% mass inde6 and include the information on report cards sent home to parentsthe
most @radicalA provision of the legislation )C%an et al., 200H0. 9he republican governor at the
time, #i+e /uc+abee, had lost more than 100 pounds as part of a personal crusade to improve
his health after being diagnosed 'ith t%pe "" diabetes, and became a fervent champion of the
legislation. 9he impact of his personal e6perience on polic% decisions 'as @perceived b% man% to
be reflected in the advancement of subse3uent health2related initiativesA in the state, )C%an et al.,
200H0. ?o', five %ears later, the idea of measuring studentsI bod% mass inde6 in schools remains
a divisive issue, often splitting supporters and opponents over the proper role of government.
#oreover, despite its importance for program and polic% monitoring and improvement,
1#" measurement in schools has received mi6ed revie's from parents, legislators, and public
health researchers concerned about the potential negative impact on children. pponents cite a
number of reasons 'h% such policies should be implemented onl% 'ith e6treme caution,
includingK the intrusive nature of such programs, the potential increase in stigmatiBation to'ard
obese children- the ris+ that children 'ill attempt unsafe 'eight loss practices or the ris+ that
parents 'ill impose such practices on their children 'ithout medical advice- inade3uate
resources in schools to implement such programs- and the potential transfer of school resources
from ph%sical activit% and nutrition programs to 1#"2related programs )?ihiser et al, 20070.
10
LAUREN E. SILVER
9hus, polic% entrepreneurs ma% be crucial for overcoming these rifts in opinion that other'ise
'ould th'art adoption.
State socioecono!ic condition. Some scholars argue that the social construction of
target populations influences the nature of the policies that public officials support )liver, 200H-
Schneider 5 "ngram, 1==8- *ersh 5 #orone, 20020. "n particular, *ersh and #orone )20020
assert that government regulation of personal behavior )e.g., diet and e6ercise, prohibition,
famil% planning, etc.0 originates from certain @triggersA that stimulate a political response,
transforming personal behavior into a public problem. ne such trigger occurs 'hen those
labeled responsible for the public problem are perceived as @demon usersAmost often the poor
and racial and ethnic minorities )*ersh 5 #orone, 20020. ?umerous empirical studies
demonstrate that the poor and racial and ethnic minorities comprise a disproportionate share of
the obese population in the (.S. :urthermore, children, as a group, tend to be vie'ed b% public
officials as @dependentsA 'ho are more helpless and need% than other constituents )Schneider 5
"ngram, 1==80. 1oth groups are +no'n for lo' political participation as 'ell )Schneider 5
"ngram, 1==80. 9his combination of neediness, class bias, and lac+ of political influence creates a
target population that ma% encourage public officialsin lo'er socioeconomic statesto
rationaliBe that more paternalistic policies, such as 1#" measurement, are 'arranted to produce
desired change 'ithout inciting a politicall% threatening reaction from the target population.
State pu0lic health pro&ision. 1oehmer, et al. )20080 demonstrate that states spending
less on public health initiatives are more li+el% to enact childhood obesit% prevention legislation.
9he current stud%, ho'ever, posits the opposite h%pothesis. 4here states spend mone%, in large
part, reflects the spending priorities of state governments )4eber, 1===0. 1% 'a% of e6ample,
states spending more than others on tobacco control, in general, also spend more on funding for
11
LAUREN E. SILVER
3uitline operations, a population2'ide tobacco cessation initiative )*eller, et al., 20070.
9herefore, assuming that spending priorities reflect issue priorities, states spending more on
population health also should be more li+el% to adopt 1#" measurement legislation.
8.2.1.2 Bill-level concepts
Bipartisan support. Prior research sho's that states are more li+el% to adopt childhood
obesit% prevention legislation 'hen legislators from both parties support the measure )1oehmer,
et al., 20080. #ore generall%, multiple sponsorship increases a billIs chances for success in state
legislatures )1ro'ne, 1=8,0. 9he current stud% h%pothesiBes that bi2partisan sponsorship 'ould
further strengthen those chances b% signaling that the issue cuts across part% lines.
Visi0ility. $hildhood obesit% prevention measures that are included as part of state
budget bills ma% have a higher chance of adoption than single2topic bills )1oehmer, et al., 20080.
#ore generall%, omnibus billsbills 'ith multiple measures from unrelated polic% areas
provide a vehicle through 'hich polic% proposals unli+el% to pass on their o'n can be adopted
)*rutB, 2001- *rutB, 20000. Due to the siBe and variation in topics covered, measures attached to
omnibus legislation are less visible to legislators, and therefore less visible to potential opponents
'ho ma% th'art adoption. "ndeed, *rutB )20000 argues that @members2at2large are seldom a'are
of the details contained in omnibus bills.A 7s a result, 1#" measurement initiatives included in
omnibus bills ma% stand a higher chance of adoption than 1#" bills that stand on their o'n.
Scope of 0ill. 9he scope of the bill refers to 'hether it calls for 1#" screening or
surveillance. 1#" surveillance refers to @the s%stematic and anon%mous collection, anal%sis, and
interpretation of data from a census or representative sampleA of students )?ihiser, et al, 20070.
"n contrast, 1#" screening programs @are designed to assess the 'eight status of individual
12
LAUREN E. SILVER
students to detect those at ris+ for 'eight2related health problemsA )?ihiser, et al, 20070.
Screening programs also send studentsI results home to their parents. 1#" surveillance programs
see+ to monitor trends in obesit% and the impact of school2based prevention initiatives, 'hereas,
1#" screening programs aim to raise a'areness among children and their parents and motivate
both to ma+e lifest%le changes )?ihiser, et al., 20070. 7ccordingl%, 1#" screening proposals
li+el% 'ill dra' more opposition since these programs are considered more intrusive, especiall%
b% those 'ho, ideologicall%, favor minimal government intervention in personal matters. 9hus,
bills that call for surveillance are li+el% to pass more than bills calling for screening.
8.2.2 #easuring and $oding $ausal $onditions and utcomes
Scholars emplo%ing the G$7 approach have used a number of different strategies for
dichotomiBing and coding causal conditions and outcomes )Cagin, 1=8<- /ic+s, #isra 5 ?g,
1==,- $overdill 5 :inla%, 1==,0. D6amples include the most simple form in 'hich the
presenceFabsence of a single condition representing one concept is coded as )1,00. ther
strategies include basing a conceptIs presence or absence on 'hether a case e6hibits a minimum
number of causal conditions that represent a particular concept- splitting the possible coding of a
concept across t'o or more conditions such that there are four or more combinations of 0 and 1
for that concept- dichotomiBing ordinal measures based on theoretical and empirical +no'ledge
of the phenomenon under investigation- and constructing scales in 'hich scores on multiple
conditions are summed for each case and a threshold is set for coding the presence or absence of
the concept. 9he measures and coding schemes for both the state2 and bill2level anal%ses are
described belo' and summariBed in :igure 1.
18
LAUREN E. SILVER
8.2.2.1 Causal conditions
7t the state2level, five causal conditions are included in the anal%sis. 9he prevalence of
adult obesit% in the state, annual state spending on obesit% among the total state population, and
annual spending on obesit% among the #edicaid population connote 'hether obesity is a
statewide public health problem. Polic%ma+ers li+el% 'ill compare their state to others along
these dimensions, thusfor each casethis measure is coded as 1 if a state e6hibits values on
t'o out of three conditions that e6ceed the national average for those conditions. ther'ise, it is
coded as 0.
Partisan control represents the degree to 'hich Democrats e6erted partisan control over
state legislative decision2ma+ing at the time public officials considered the legislation. 9hat is,
'hether Democrats held enough seats in the state legislature to overrule Cepublican opposition
to bills andFor 'hether the governor 'as a Democrat. "t is coded as 1 if one or both conditions
are present- other'ise it is coded as 0.
9he influence of a policy entrepreneur indicates 'hether a +e%, influential actor 'ithin
the polic% communit% contributed to the development and adoption of the legislation in such a
'a% that other ma&or actors involved doubt the legislation 'ould have been adopted 'ithout this
personIs support. 1ased on semi2structured intervie's 'ith other +e% pla%ers involved in the
legislation and a scan of the print media, it is coded as 1 if an entrepreneur e6erted influence, and
0 if not.
State socioeconomic condition is represented b% three socioeconomic indicators of state
residentsK per capita income, percent of the population in povert%, and high school non2
completion rate. "f a case e6hibits t'o out of three that indicate a socioeconomic status lo'er
1<
LAUREN E. SILVER
than the national average )i.e., lo'er than average per capita income and higher than average
povert% and high school non2completion rate0, it is coded as 1- and is coded as 0 other'ise.
State public health provision represents the amount of resources the state devotes to
population health 10 relative to all other spending prioritiesmeasured b% #edicaid
e6penditures as a percent of total state e6pendituresand 20 relative to health spending onl%
measured b% total e6penditures on population health as a percent of total health spending. "t is
coded as 1 if both conditions e6ceed the national average. 9his stud% places a higher standard for
this conditionIs presence based on prior statistical anal%ses demonstrating that states spending
less on population health are more li+el% to adopt childhood obesit% prevention legislation
)1oehmer, et al., 20080.
7t the bill2level, three causal conditions are included in the anal%sis. Bipartisanship
indicates 'hether a bill received bipartisan support among state legislators and is measured b%
'hether a bill had t'o or more sponsorsat least one from both the Democratic and Cepublican
part%. "t is coded as 1 if a bill had bipartisan support and coded as 0 if it did not.
Visibility represents ho' detectable a bill 'as during the legislative process. Lisibilit% is
considered lo' if the bill 'as included as part of a larger ombnibus bill, such as a state budget
bill- and is considered high if it the bill addressed the issue of 1#" measurement alone or along
'ith other related health polic% matters. "t is coded as 1 for lo' visibilit%, and 0 for high
visibilit%.
4hether a bill calls for 1#" surveillance or screening represents the scope of the bill.
1#" screening refers to measuring the 1#" of individual students and notif%ing parents of the
results, and therefore is considered broader in scope than 1#" surveillance. "t is coded as 1 if the
official bill te6t calls for screening, and 0 if it does not.
1,
LAUREN E. SILVER
8.2.2.2 Outcomes
:or the state2level anal%sis, states that have adopted 1#" measurement legislation 'ill be
coded as 1, and states that have not introduced an% 1#" measurement legislation to date 'ill be
coded as 0. :or the bill2level anal%sis, states that have adopted 1#" measurement legislation 'ill
be coded as 1, and situations in 'hich states have introduced but not adopted a particular 1#"
measurement bill 'ill be coded as 0. 1% comparing adopted bills to introduced but not adopted
bills, the bill2level anal%sis 'ill focus specificall% on measuring bill2level factors onl% )state2
level conditions 'ill be controlled for as described belo'0.
8.2.2.8 Potential challenges with measurement and access to data
Some measurement challenges are anticipated. :irst, the dichotomiBation and coding of
ordinal data ma% not reflect, in realit%, the thresholds that state polic%ma+ers respond to 'hen
considering proposing andFor adopting polic% proposals. Li+e'ise, despite developing the semi2
structured intervie' protocols based on 'ell2established theor% about polic% entrepreneurs )i.e.,
*ingdon, 1=8<0, the 3uestions and additional probes ma% fail to uncover 'hether a polic%
entrepreneur 'as, in fact, present and effective. 9hird, the operationaliBation of conceptssuch
as @obesit% as public health problemA and @state public health provisionAma% not reflect as
accuratel% as possible ho', in practice, public officials frame the obesit% problem or prioritiBe
population health relative to other state priorities.
Some difficulties accessing data are anticipated as 'ell. State legislative 'ebsites var%
'ith respect to the amount of legislative information and detail that is made publicl% available
online, thus, coding for some cases ma% re3uire more @detective 'or+A than others, such as
1H
LAUREN E. SILVER
obtaining data b% telephone. 7dditionall%, gaining access to the @rightA +e% actors for semi2
structured intervie's 'ill most li+el% pose the greatest measurement challenge of all.
Figure 1. Summary of Measures and Data Sources
Systematized Concept Operationalization Indicator(s) Data Source(s)
State-e!el
Causal Conditions
Obesity as a statewide public
health problem
State adult obesity prevalence
State spending on obesity
Percent of state adult population with
BMI 30 greater than national
average
State spending on obesity greater
than national average among total
population
State spending on obesity greater
than national average among
Medicaid population
B!"SS#
"in$elstein% et al% &00'
Partisanship (as there partisan control of
the state government at the
time of bill enactment
)emocratic ma*ority in the
legislature
)emocratic governor
+,S-
Policy entrepreneur -egislation introduction and.or
outcome significantly
influenced by actions of policy
entrepreneur
!esponses of $ey actors w.in the
policy community to semi/structured
interviews indicative of presence
!epetition of $ey actor0s1 by print
media
Semi/structured
interviews w.$ey actors#
Scan of print media
coverage
State socioeconomic condition Per capita income# percent
population in poverty# high
school non/completion rate
Per capita income below national
average
Percent population in poverty above
national average
2igh school non/completion rate
above national average
34S4 ,ensus
State public health provision State e5penditures on
population health and
Medicaid
Medicaid e5penditures as a percent
of total e5penditures greater than
national average
6otal e5penditures on population
health as a percent of total health
spending greater than national
average
+7SBO State
85penditure !eports#
+7SBO State 2ealth
85penditure !eport
Outcomes
-egislation on BMI
measurement in public schools
Bill introduced and adopted
OR no BMI measurement bills
introduced to date
Bill te5t# official record of bill
enactment# lac$ of record of
legislation introduction to date
,),% +,S-% 6rust for
7merica9s 2ealth% -e5is
+e5is State ,apitol%
state legislative
websites
"ill-e!el
Causal Conditions
Bipartisanship Bipartisan support for the bill More than one sponsor from both State legislative
17
LAUREN E. SILVER
parties indicated by bill te5t websites
:isibility Bill passed as part of omnibus
legislation
Bill te5t State legislative
websites
Scope of bill )oes the bill call for BMI
screening of individual
students and notifying parents
of the results
Bill te5t State legislative
websites
Outcomes
-egislation on BMI
measurement in public schools
Bill introduced and adopted
OR bill introduced and not
adopted
Bill te5t# record of bill status ; history ,),% +,S-% 6rust for
7merica9s 2ealth% -e5is
+e5is State ,apitol%
state legislative
websites
B!"SS% Behavioral !is$ "actor Surveillance System# +,S-% +ational ,onference of State -egislatures# +7SBO% +ational
7ssociation of State Budget Officers# ,),% ,enters for )isease ,ontrol and Prevention
8.2.8 $onstructing 9ruth 9ables and Selecting $ases
nce causal conditions and outcomes are operationaliBed into binar% form, the ra' data
collected on state 1#" measurement legislation must be coded and reconstructed into a truth
table ):igure 20. 9he truth table enables the use of 1oolean algebra as the tool for anal%sis. 9he
total number of ro's in a truth table is determined b% the number of causal conditions )e.g., for
the state2level anal%sis, there are five causal conditions, so the total number of ro's 'ill be 2
,
,
'hich e3uals 82 ro's0. Dach ro' in the truth table represents one of 82 possible logical
combinations of the five causal conditions, including the presenceFabsence of the outcome. :or
the more straightfor'ard use of 1oolean anal%sis that the current stud% proposes, the fre3uenc%
of each combination of conditions can be ignored )Cagin, 1=870. 9he focus is on different t%pes
of situations that lead or do not lead to states adopting 1#" legislation, and not the fre3uenc% at
'hich these situations occur.
Figure #. $artial %rut& %a'le for State-e!el Factors
State O $ ( S )
1 0 0 < 0 0 0
# < < 0 0 < <
* 0 0 < < 0 0
+ < < < < < <
, < < 0 < < <
- 0 0 0 0 0 0
18
LAUREN E. SILVER
. < 0 0 0 < 0
/ < < < 0 < <
,ausal conditions= O>Obesity as a statewide public health problem# P>Partisan control# 8>presence of policy
entrepreneur# S>State socioeconomic condition# 2>State public health provision
Outcome= If ->0% state has not introduced BMI measurement legislation to date# if -><% state adopted BMI
measurement legislation
$ase selection, therefore, is based on identif%ing cases that are consistent 'ith as man%
ro's in the truth table as possible. :or the current stud%, cases 'ill be selected on the t'o
possible outcomes for the state2level anal%sisK states that, to date, have introduced no legislation
regarding 1#" measurement in schools )coded as 00, and states that not onl% have introduced,
but also have passed such legislation )coded as 10.
$ases 'ill be identified using a number of sources )see :igure 10, including the $enters
for Disease $ontrol and Prevention online ?utrition, Ph%sical 7ctivit% and besit% Legislative
Database, the ?ational $onference of State Legislature annual updates of childhood obesit%
legislative options, 9rust for 7mericaIs /ealth annual @: as in :atA reports )9rust for 7mericaIs
/ealth, 200<220080, Le6is ?e6is State $apitol files, and individual state legislative 'ebsites. :or
efficienc% purposes, cases 'ill initiall% be identified using the first four sources, and then 'ill be
confirmed b% locating the full te6t and official status of the bill on individual state legislative
'ebsites. Should the current stud% fail to identif% cases that are consistent 'ith all 82 possible
combinations of the truth table, there are various strategies for dealing 'ith missing cases
)Cagin, 1=870.
8.2.< Simplif%ing and 7ssessing $ausal $ombinations
nce as man% ro's in the truth table as possible have been filled in using the process described
above, these @primitive e6pressionsA should be reduced into shorter e6pressions )Cagin, 1=870.
9his process is referred to as @minimiBationA )Cagin, 1=870. 9he ob&ective is to eliminate causal
conditions that are irrelevant to the outcome. 9his tas+ mirrors #illIs method of similarit% in that
1=
LAUREN E. SILVER
if t'o e6pressions share the same outcome and the same presenceFabsence of causal conditions
but for one, then that one condition can be eliminated as a causal condition. 9he e6ample of
minimiBation belo' is based on the fabricated ra' data in :igure 2. States < and , have enacted
1#" measurement legislation, but the% differ b% one conditionthe presence of a polic%
entrepreneur )capital letters indicate the presence of a condition- lo'ercase letters indicate the
absence of a condition0K
9he t'o primitive e6pressions can be reduced to a simpler e6pressionK PS/ L. Li+e'ise,
states 2 and 8 share the same outcome, but also differ regarding the presence of a polic%
entrepreneurK
9hese e6pressions reduce to Ps/ L.
9he t'o reduced e6pressions from each pair of states can be reduced further since the% share the
same outcome, but differ 'ith respect to one condition, state socioeconomic statusK
Dliminating causal conditions that are irrelevant to particular combinations continues in
this manner until all possible combinations that result in the presenceFabsence of L have been
reduced to their simplest form. 9hese combinations then comprise the final 1oolean e3uation per
State <K PDS/ L
State ,K PeS/ L
State 2K Pes/ L
State 8K PDs/ L
PS/ L
Ceduces to L M P/
Ps/ L
20
LAUREN E. SILVER
outcome )Cagin, 1=870. Suppose, for instance, the onl% other reduced e6pression that results in L
is pD. 9he final 1oolean e3uation is L M P/ N pD. 9hus, 1#" measurement legislation
results from t'o different combinations of causal conditionsK the first combination includes
obesit% perceived as a state'ide public health problem, Democratic partisan control of state
government, and state public health provision that is higher than the national average- 'hile the
second combination includes obesit% perceived as a state'ide public health problem, the absence
of Democratic partisan control, and the influence of a polic% entrepreneur. 9he minimiBation
process see+s to identif% the simplest combinations of necessar% and sufficient conditions for
each potential outcome, that is, success or failure of a state to adopt 1#" measurement
legislation )Cagin, 1==10.
4.4 Bill/Le&el Analysis
9he same G$7 process 'ill be carried out for the bill2level anal%sis. "n order to control
for the effect of state2level factors, ho'ever, case selection 'ill differ. 9here are t'o strategiess
for selecting cases for the bill2level anal%sis- 'hether both are used depends on ho' man% cases
can be identified using the first strateg%, 'hich offers the most control. 9he first strateg% entails
identif%ing states in 'hich a bill has been introduced, but not adopted as 'ell as a bill that has
been adopted. 1% focusing on comparing bills 'ith different outcomes 'ithin the same state, the
ma6imum amount of control is achieved for potential confounding factors at the state2level. "f
the bill2level anal%sis re3uires additional cases after e6hausting the strateg% described above,
states then 'ould be matched according to the five state2level conditions e6amined in the stud%.
Pairs of states sharing the same coding across these five conditions 'ould be investigated for
instances of bill introduction 'ithout adoption and bill introduction 'ith adoption. State2level
21
LAUREN E. SILVER
factors 'ill be controlled for to the e6tent that these cases are consistentl% and accuratel% coded
across state2level conditions.
5.( S)REN,).S AN+ LIMI)A)I*NS *- ).E $R*$*SE+ S)U+3
5.' Stren"ths
Cagin )1=8<0 asserts that G$7 @should prove valuable in an% investigation in 'hich there
is reason to suspect that features of multivariate statistical anal%sis inhibit the identification of
descriptivel% meaningful, but subtle, patterns of interaction.A 4hile 1oehmer, et al. )20080 and
$a'le% and Liu )20080 provide a useful foundation for h%pothesiBing the degree to 'hich
potential individual causal conditions are associated 'ith state adoption of 1#" measurement
legislation, these studies fail to capture the inherent causal comple6it% that in all li+elihood
characteriBes this particular phenomenonin large part because the outcome of this t%pe of
legislation depends on the decision2ma+ing processes of public officials. !iven its focus on
simplif%ing causal comple6it%, G$7 is a particularl% suitable approach for capturing the
comple6it% of decision2ma+ing. #oreover, b% focusing on con&unctural causation that results in
the presence or absence of state adoption of 1#" measurement legislation, G$7 presents
findings in a manner that is more meaningful to polic%ma+ers and other sta+eholders 'ho are
loo+ing to target obesit% prevention efforts at receptive state environments.
Li+e'ise, the proposed t'o2level G$7 approach 'ill provide a basis for interpreting the
separate effects of state2 and bill2level factors on state adoption of 1#" measurement legislation.
9his 'ill further aid polic%ma+ers and sta+eholders in choosing ho' and 'here to target their
efforts.
5.1 Li!itations
22
LAUREN E. SILVER
Despite its strengths as a method relative to traditional statistical techni3ues and alternative
3ualitative, case2oriented approaches, G$7 presents a number of limitations in general and
regarding this particular stud%, in particular. :irst, an% stud% involving the use of G$7 runs the
ris+ of overloo+ing an important causal condition. "f a missing, but significant causal condition is
identified, the results of the initial G$7 no longer 'ill appl% as the introduction of &ust one
condition 'ill alter the truth table, the minimiBation process, and final 1oolean e3uations.
Cegarding the current stud%, it is possible that potentiall% important causal conditions are missing
from the proposed anal%sis, including the impact of state diffusion of polic% innovation )Lolden,
200H0, the %ear in 'hich states considered 1#" measurement legislation )$a'le% 5 Liu, 2080,
and the impact of public opinion and interest groups.
:inall%, dichotomiBing causal conditionsespeciall% conditions that originall% 'ere in
ordinal formcould result in a potential loss of information )Cagin, 1=870.
6.( AN)ICI$A)E+ -IN+IN,S
"n terms of specific causal conditions, the current stud% anticipates that certain state2 and
bill2level characteristics that prior research has demonstrated are significantl% associated 'ith
state adoption of childhood obesit% legislation, in general, also 'ill demonstrate a causal role
follo'ing 3ualitative comparative anal%sis. Specificall%, it is anticipated that the presence of
state2level characteristics of partisan control and state socioeconomic condition 'ill lead to
outcomes reflecting legislation adoption, as 'ell as the bill2level characteristic of bi2partisan
support. 9hese characteristics demonstrated the strongest statisticall% significant association 'ith
state adoption of childhood obesit% legislation, in general.
#ore broadl%, the current stud% 'ill provide a stronger empirical basis for theoriBing
about ho' state2 and bill2level characteristics are li+el% to affect 'hether states adopt legislation
28
LAUREN E. SILVER
calling for 1#" measurement in public schools. 7lthough prior statistical anal%ses )1oehmer, et
al., 2008- $a'le% 5 Liu, 20080 provide the foundation for forming preliminar% h%potheses about
'hich state and bill characteristics ma% matter, the proposed G$7 anal%sis 'ill develop an
empiricall% stronger theor% b% accounting for the causal comple6it% behind the decision2ma+ing
processes of public officials in one state compared to other states. "n short, the multiple causal
paths b% 'hich this t%pe of legislation succeeds and fails 'ill be identified.
7dditionall%, state andFor bill characteristics previousl% h%pothesiBed to pla% important
roles in the outcome of 1#" measurement legislation either 'ill be confirmed as +e%
characteristics or 'ill be eliminated via the minimiBation process described above. 9his 'ill
further clarif% for polic%ma+ers, researchers, health organiBations, and other sta+eholders 'here
the% should focus childhood obesit% prevention efforts and future areas of related research.
RE-ERENCES
7dcoc+, C. 5 $ollier, D. )20010. #easurement validitK 7 shared standard for 3ualitative and
3uantitative research. The American Political Science eview, =,, ,2=2,<H.
1oehmer, 9.*., Lu+e, D.7., /aire2;oshu, D.L., 1ates, /.S., 1ro'nson, C.$. )20080. Preventing
childhood obesit% through state polic%K Predictors of bill enactment. American !ournal of
Preventive "edicine# $%, 88828<0.
1ro'ne, 4.P. )1=8,0. #ultiple sponsorship and bill success in (.S. state legislatures. &egislative
Studies 'uarterly, 10, <882<88.
1ro'nson, C.$., 1oehmer, 9.*., /aire2;oshu, D., Dreisinger, #.L. )20070. Patterns of childhood
obesit% prevention legislation in the (nited States. Preventing $hronic Disease, <, 88828<0.
1ro'nson, C.$., ?e'schaffer, $.;. 5 7li27barghoui, :. )1==70. Polic% research for disease
preventionK $hallenges and practical recommendations. American !ournal of Public (ealth, 87,
78,278=.
$a'le%, ;. 5 Liu, :. )20080. $orrelates of state legislative action to prevent childhood obesit%.
Obesity, 1H, 1H221H7.
2<
LAUREN E. SILVER
$#P7SSS. $omparative methods for the advancement of s%stematic cross2case anal%sis and
small2n studies. "nventor% of good practices in G$7. Cetrieved December 18, 2008 from
httpKFF'''.compasss.orgFDidactic.htm
$overdill, ;.D. 5 :inla%, 4. )1==,0. (nderstanding #ills via #ill2t%pe methodsK 7n application
of 3ualitative comparative anal%sis to a stud% of labor management in southern te6tile
manufacturing. 'ualitative Sociology, 18, <,72<78.
Dconomos, $.D., 1ro'nson, C.$., De7ngelis, #.7., ?ovelli, P., :oerster, S.1., :oreman, $.9.,
et al. )20010. 4hat lessons have been learned from other attempts to guide social changeO
)utrition eviews, ,=, S<02S,H.
!regg, D.4., $heng, P.;., $ad'ell, 1.L., "mperatore, !., 4illiams, D.D., :legal, *.#., ?ara%an,
L., 4illiamson, D.:. )200,0 Secular trends in cardiovascular disease ris+ factors according to
bod% mass inde6 in (.S. adults. !A"A# *+$, 18H8Q187<.
/a%ne, $.L., #oran, P.7. 5 :ord, #.#. )200<0. Cegulating environments to reduce obesit%.
!ournal of Public (ealth Policy, 2,, 8=12<07.
:in+elstein, D.7., :iebel+orn, ".$. 5 4ang, !. )200<0. State2level estimates of annual medical
e6penditures attributable to obesit%. Obesity esearch, 12, 1822<.
/ic+s, 7., #isra, ;. 5 ?g, 9.?. )1==,0. 9he programmatic emergence of the social securit% state.
American Sociological eview, H0, 82=28<=.
;acob%, 4.!. 5 Schneider, S.*. )20010. Lariabilit% in state polic% prioritiesK 7n empirical
anal%sis. The !ournal of Politics, H8, ,<<2,H8.
*eller, P.7., *oss, ;.*., 1a+er, 9.1., 1aile%, L.7. 5 :iore, #.$. )20070. Do state characteristics
matterO State level factors related to tobacco cessation 3uitlines. Tobacco Control, 1H, i7,2i80.
*ersh, C. 5 #orone, ;. )20020. /o' the personal becomes politicalK Prohibitions, public health,
and obesit%. Studies in American Political ,evelopment, 1H, 1H2217,.
*ingdon, ;.4. )1=8<0. Agendas# alternatives# and public policies. /arper $ollins Publishers.
*oplan, ;.P., Liverman, $.9., *raa+, L.". )Dds.0 )200,0. Preventing childhood obesit%K /ealth in
the balance. 4ashington, D.$.K ?ational 7cademies Press.
*rutB, !.S. )20010. 9acital maneuvering on omnibus bills in congress. American !ournal of
Political Science, <,, 2102228.
*rutB, !.S. )20000. !etting around gridloc+K 9he effect of omnibus utiliBation on legislative
productivit%. &egislative Studies 'uarterly, 2,, ,882,<=.
*uman%i+a, S.*., barBane+, D., Stettler, ?., 1ell, C., :ield, 7.D., :ortmann, S.P., :ran+lin,
1.7., !illman, #.4., Le'is, $.D., Poston, 4.$., Stevens, ;., /ong, P. )20080.
2,
LAUREN E. SILVER
Population2based prevention of obesit%K 9he need for comprehensive promotion of healthful
eating, ph%sical activit%, and energ% balanceK 7 scientific statement from 7merican /eart
7ssociation $ouncil on Dpidemiolog% and Prevention, "nterdisciplinar% $ommittee for
Prevention ):ormerl% the D6pert Panel on Population and Prevention Science0. Circulation, --.,
<282<H<. Cetrieved September H, 2008, from
httpKFFcirc.aha&ournals.orgFcgiFcontentFfullF118F<F<28
#c?ichol, D. 5 Lav, ".;. )2008, December0. State budget troubles 'orsen. Cetrieved December
12, 2008, from httpKFF'''.cbpp.orgF=28208sfp.htm )$enter on 1udget and Polic% Priorities0
?ice, D.$. )1==<0. Policy innovation in state government. 7mes, "7K "o'a State (niversit%
Press.
?ihiser, 7.;., Lee, S.#., 4echsler, /., #c*enna, #., dom, D., Ceinold, $., 9hompson, D.,
!rummer2Stra'n, L. )20070. 1od% mass inde6 measurement in schools. !ournal of School
(ealth# // 0-12, H,12H71.
liver, 9.C. )200H0. 9he politics of public health polic%. Annual eview of Public (ealth, */,
1=,2288.
Cagin, $.$. )Dd.0 )1==10. 3ssues and alternatives in comparative social research. ?etherlandsK
D.;. 1rill.
Cagin, $.$. )1=870. The comparative method4 "oving beyond 5ualitative and 5uantitative
strategies. 1er+ele% and Los 7ngeles, $7K (niverist% of $alifornia Press.
Cagin, $.$., #a%er, S.D. 5 Drass, *.7. )1=8<0. 7ssessing discriminationK 7 1oolean approach.
American Sociological eview, <=, 221228<.
Cihou6, 1. )20080. 1ridging the gap bet'een the 3ualitative and 3uantitative 'orldsO 7
retrospective and prospective vie' on 3ualitative comparative anal%sis. 6ield "ethods, 1,, 8,12
8H<.
Cihou6, 1. )200H0. Gualitative comparative anal%sis )G$70 and related s%stematic comparative
methodsK Cecent advances and remaining challenges for social science research. 3nternational
Sociology, 21, H7=270H.
Comme, 7.!.L. )1==,0. 1oolean comparative anal%sis of 3ualitative data, a methodological note.
'uality and 'uantity, 2=, 817282=.
C%an, *.4., $ard2/igginson, P., #c!arth%, S.!., ;ustus, #.1., 9hompson, ;.4. )200H0.
7r+ansas fits fatK 9ranslating research into polic% to combat childhood and adolescent obesit%.
(ealth Affairs# *70%2, ==22100<.
Schneider, 7. 5 "ngram, /. )1==80. Social consruction of target populationsK "mplications for
politics and polic%. American Political Science eview, 87, 88<28<7.
2H
LAUREN E. SILVER
9horpe, *.D., :lorence, $.S., /o'ard, D./., ;os+i, P. )200<0. 9he impact of obesit% on rising
medical spending. (ealth Affairs# (ealth Trac8ing Trends 9eb :;clusive.
9rust for 7mericaIs /ealth. )200<, 200,, 200H, 2007, 20080. : as in fatK /o' obesit% policies are
failing in 7merica. Cetrieved December 12, 2008 from
httpKFFhealth%americans.orgFreportsFobesit%2008F
(.S. Department of /ealth and /uman Services, $enters for Disease $ontrol and Prevention
)20080. ver'eight and obesit%K "ntroduction. Cetrieved September 8, 2008, from
httpKFF'''.cdc.govFnccdphpFdnpaFobesit%F
(.S. Department of /ealth and /uman Services, Public /ealth Service, ffice of the Surgeon
!eneral. )20010. 9he Surgeon !eneralRs call to action to prevent and decrease
over'eight and obesit%. Cetrieved September, 7, 2008 from
httpKFF'''.surgeongeneral.govFtopicsFobesit%FcalltoactionFcover.htmScitation.
Lolden, $. )200H0. States as polic% laboratoriesK Dmulating success in the $hildrenIs /ealth
"nsurance Program. American !ournal of Political Science, ,0, 2=<2812.
4ang, P., 5 1e%doun, #.7. )20070. 9he obesit% epidemic in the (nited Statesgender, age,
socioeconomic, racialFethnic, and geographic characteristicsK 7 s%stematic revie' and meta2
regression anal%sis. :pidemiologic eviews, *+, H228.

4eber, C.D. 5 1race, P. )1===0. American state and local politics. ?e' Por+, ?PK $hatham
/ouse Publishers of Seven 1ridges Press, LL$.
27

S-ar putea să vă placă și