Escolar Documentos
Profissional Documentos
Cultura Documentos
Eugene Bardach
Richard and Rhoda Goldman School of Publlc Policy
University of Calfornla at Berkeley
TH15 15 A HANDBOOK of concepts and rncthods for working your way rhrough
a policy anaJysis. The presumed user is a beginning practirioner preparing to
underrake a policy analysis. 1 have developed the general approach and many of
the specific suggesrions over rwenry-five years of teaching poJicy analysis worlcshops
to first- and second-year graduare students at (he Richard and Rhoda Goldman
5chool of PubJic Policy, Universiry of California, Berkeley. 1 have also found this
handbook useful in teaching an undergraduate introducrion to public policy and
for executive education groups.
The handbook assurncs a familiarity with basic econornic concepts, including
those having to do wirh rnarket failures (including rnarker imperfections). Ir ls
nor meanc ro stand alone bur should be used in conjunccion wirh orher sources.
Five of thc best texrbooks in poJicy analysis, which amplify poinrs in chis handbook,
are
Edith Srokey and Richard Zeckhauser, A Prima flr Policy Analysis (New
York: Norron, 1978).
David L. Weimer and Aidan R. Vining, Policy Analysis: Concepts and
Practice (Upper Saddle River, N.J.: Prcntice Hall, 1999).
Roben D. Behn and James W. Vaupel, Quick Analysis for BuS) Decision-
Makers (NewYork: Basic Books, 1982).
Lee 5. Friedrnan, Microeconomic Policy Analysis (New York: McGraw-
r Hill, 1984)-revised edirion forrhcoming.
Duncan MacRae Jr., and Dale Whittingron, Expert Aduice for Policy
Choice: Analysis and Discourse (Washington, D.e.: Ceorgetown Univer-
sity Press, 1997).
1 wish to acknowledge the patience of and rhe helpful rcactions frorn all the
students and fric:nds who .havc made use of this handbook, especially rhose who
pue up wirh its earlier versions. Special thanks are due to Roben Behn, Sandford
xlv A PRACTICAL GUIDE FOR POL/CY ANALYSIS INTRODUCTlON xv
pracrice and ro srudents and others who, for whatevcr reasons, are arte:mpting to The spirit in which you takc any orie of thesc sre:ps, cspecially in me: carlicst
lo ole ar rhe world rhrough rhe: e:ye:sof a practitioner. phases of your projecr, should be highly renrative. As you move: rhrough [he
problern-solving process, you will probably keep changing your problem defini-
tion, your menu of alrernarives, your ser of evaluative criteria, your sense of what
Policy Analysis: More Art than Science evidencc bears' on the problern, and so on. Wirh each successivc iceration you
Policy analysis is more art rhan scie:nce. Ir draws on inruition as much as method, will become a bit more confident that you are: on [he: righr track, rhat you are
Nevertheless, give:n [he: choice: berween advice mar irnposes too rnuch structure focusing on rhe right question, and so on. This can be a frusrrating process, bur
on rhe problern-solving proce:ss or roo lirrle, rnost be:ginning practitioners quire ir can also be re:waeding-provide:d you can leaen to enjoy rhe challe:nge of search,
reasonably pre:fer roo mucho 1 have: rherefore developed an approach 1 call me discovery, and inve:ncion.
Eightfold Parh. The primary utiliry of chis structured approach is rhar ir reminds
you of irnporranr rasks and choice:s that otherwise mighr slip your rnind; its
primary drawback is that, rakcn by itself ir can be mechanistic.
Some of the Guldelines Are Practlcal, but Most Are Conceptual
Mosr of rhe conceprs use:d will seern obvious, However, thcre are exceprions,
he Eightfold Path Fi~sr. rechnical rerrns are sornctimes employed. Sccond, sornc commonsense: tcrrns
may be used in a special way that strips rhern of cerrain connotations and perhaps
Define: thc Problem irnports others. For the rnosr part, all these concepts will bccorne inrelligible
~hrough experience and practice,
Assernble Some Evidence
These: sreps are nor ne:cessarily rakcn in precisely rhs order, nor are all of
thern necessarily significant in every problem. Howe:ver, an dforr to define me:
proble:m is usually che righe srarting place, and te:Uing thestory is almosr inevitably Your Final Product
thc e:nding poinr, Consrruceng alternatives and selecting crircria for evaluating So whar will your final produce look like? He:re is a very rough sketch of a typical
thcm musc surely come roward rh.beginning of the process. Assembling so me 'wrltren policy-analytic re:pore: In a coherenr narrative sryle you will describe some
evidence is actually a step rhar recurs chroughour the entire process, and ir applic:s problern that needs to be mitigaeed or solved. You will lay out a few alternarive
"
parricularly ro efforrs to define [he problem and ro project the outcornes of rhe courses of aceon that rnighr be taken, To each course of acron you wil! atrach
alternarives being considered. a set of projected outcomes that you think your client or audience would care
abour, suggesting the cvidentiary grounds for your projections. If no alrernative
dominates all orher aleernatves wirh re:spece [Q all rhe evaluarive crircria of intcrcsr,
Iteration 15 Continual you will indicare me nature and magnitude of me rrade-ofls implicit in differcnc
The proble:m-solving proce:ss-be:ing a proce:ss of erial and error-is iteracive, so policy choices. Depending on rhe clicnr's expccrations, you might state your own
rhat you usually must repear e:ach of rhese sre:ps, sometimos more rhan once. recornmendarion as to which alrernative should be chosen.
I
A PRACTICAL GUIDE FOR POLlCY ANALYSIS tj
e Spirit of the Eightfold Path
e spirit of me: Eightfold Path is, 1 hope, an economizing and uplifting spiric,
alyzing public policy problerns is a cornplex activiry. Ir is casy ro get los e, wasre
t
.r of rime, bccorne demoralized. Other manuals and rextbooks in polcy analysis
very concerned thar you gcr the analysis "righr," in sornc sense. I hope chis
: will hclp in that rcspecr roo. Bur, even more, I hope that this one'will help
l get ir done wirh reasonable efficiency as wcll. .' :
I PART r
The Eightfold Path
raxpayers, and social disinregrari~n. Usually you will wanr to de:re:r~e: a primary
warranr definirian as ,public roblems and r e eb le itimatd ra~~ for
problem focus; o therwise you will find mar' me analysis ge:rs out of hand. But if
amelioranon y pu IC resources?";-{r is usually hdpful to view me situaticn
rhe: proble:ms areri'r toa complcared, you may fe:d you are: willing to define more
rhan one. ' rhrough clle market [ailur lens.' In' irs simplest formularon, rnarket failure occurs
when the tecbnical properties af a good or service
the existence of discrirninarion against racial and orher rninorities: necdcd scrutiny. Scc "Projecr rhe Ourcorncs" for
definitions can easi 1y escape
che failurc of governmenr to function wdl in arcas whcrc Ir is tradition- further discussion.
a1ly expecred to act eflectively kg., in providing public schools).
rdltlons That Cause Problems Are Also Problems Common Pitfalls in problem Depnitlon-
(
le problem conditions are not experienced as troublesorne per se by citizens Problern definirion is a deceptively simple srep. Ir is a step bcset by ac least rwo
are perceived by rhern, or by analysts working on their behalf ro be causes dangeraus pitfa.lls. . .
roubles, Ir is sornerimes useful to diagnose one or more a1leged causes of this Defining the solutlon Into the "problem." Your problern definidon should
: and to define these as problema to be mitigarcd or removed, forinstance, not include an implicit solution intraduced by sernantic carelessness. Proj~~(ed
le of rhe problerns in rhe air pollurion arca is rhat srates have not been willing solutions rnust be evaluatcd empirically and not legirirnated rnerely by definicion.
orce rnotorists to keep rheir engines runed up and rheir exhaust systerns in Therefore, keep the prablc:m definition srripped down (O a rnere description, and
rer order." cavc open where you willlook for solutons.
Note mar rhis sort of problern deflnition is not merdy descriprive but is a1so
n ostic. It implicidy asserts that so me condition, which may or may nor be Don', say: "The:re is (00 lird~ sheltcr for horncless fanlilies." This fo~mula-
bling to people per se, is an irnporranr cause of 50 me orher condirion mar is don mighe inadverrenrly imply rhat "more shelter" is rhe bese so~~tlon
.ed rroubling. Problem definirions that pretend to sornc diagnostic power can and rnight inhibir you fram thinking about ways (Q prevent [amilics
rseful, but they can a1so be rreacherous. Suppose, afrer a11, rhat me causal fram becoming homeless in rhe, firsr place. Try instead: "There are (00
nosis is mistaken or mislc:ading, for exarnple, that states' unwillingness to many horneless familie:s." ".'
rce engine rnainrenance routines is not in fact a very important cause of air Don', say: "New schools are being builr too slowly. T~lS ~o:mulatlon
ution, Becausc definition in sornc conrexrs connotes legitimate arbitraririess could imply "more schools" as the solucion and could l.nhlb1Cyou :rom
1define '[ustice' to rnean ... ") the causal.!daims irnplicir in diagnoscic problem rhinking abour ways [O use existing facilities more efficlently. Try InItead
THE EIGHTFOLD PATH 7
6 A PRACTICAL GUIDE FOR POLiCY ANALYSIS
2. Some analysrs also clairn that ir is simply nor wormwhile ea define as "probltms condi-
tions rhar cannor be amelior:ued: Problems are berree ueared as oppoeeuniries foe improve-
mef\ti dcfined problems, as peablems of choice berween alrernarive me.OS ea eealiu: a given
opporruni'Y. Thc proccss of problcm definiran would rhen be one of search, crearion, .nd
inltial ex.min.rion of ideas foe salurion unril problem of choice is reachcd: O.vid Oery.
Problem Dtfinilion in Policy .nalyit (L.wrtncc: Universi'Y Pross of Kan; , 1984). 27.
THE EIGHTFOLD PATH 9
A PRACTICAL GUIDE FOR POLlCY ANALYSIS
policy analysis proce:ss. this "Assernble Some Evidence" step on th~ Eightfold
:n~' .dcnce Of these rwo activiries, thinking is by far the more important,
Path will be taken more: [han once:, but wirh a different focus each umc.
t husding data ta es much more time: reading documents, huncing in libraries, , I
ring over studies and sratisccs, intervie:wing people, traveling to interviews and
jring for appoinrrnenrs, and so on.
The real-world setrings in which policy analysis is done rarcly afford the rime: Think 8efore You Collect
. a re:se:arch effoft that would please a carc:ful academic re:se:arche:r.In fa~r, time: Thinking and colle:cting data are cornplernenrary activities: you can be: a much
.ssure is probably almost as dange:rous an e:ne:my of high-quality policy analysis more: efficienr collector of data if you think, and keep on thinking, abour whar you
polirically rnorivated bias, if not more: so. The:rc:fore, ir is essential ro econornize do and don't need (or-want) to know, and why. The principal-and e:xce:edingly
your data collection activiries, The ke:y to -cconornizing is this: Ter
to colle:ct common-mistake: rnade by be:ginne:rs and veterans alike: is to spe:nd rime: colle:cting
1 rhose: dara mat can be: turned ineo "informaron" that in rurn, c n be: data thar have linle or no porential ro be: developed into evidcnce concerning
1Ve:rted inro "evi ence: t at as so me: bearing on your prob,km. ; anyrhing you actually carc about, Pco le: often do this be:cause: runnin' around
-Por the 10gicaIly mindcd, here are: sorne definitions: Data are facts-or, some: collectin ' a ooks and fe:c:Is roductive: whe:reas Irst-rate: inking is hard and
ght say, representations of facrs-abour (he warld. Data indude: all sorts of fruscrating. Also, me: people paying for your work rend to. be: re:~sure ~t, w en
tistics bur go we:1Ibc:yond sraristics, too. Data also include, for instance, facrs they sce you busily collccting data. somehow they are: ge:mng their mone:y s worth.
cut an age:ncy manager' s ability ro deal cdnstructively with the prc:ss. Information The value of evldence. Since most evidcncc is cosdy to produce, you rnust
data that has "rneaning," in (he sense rhar ir can help you sort the world into weigh its likdy cost against its likdy value. How is its likdy value to be: estimacedi
ferent logical or e:mpirical categories. The: pre:vale:nce: of cigare:tte: smoking in The: answer may be: casr in a declsion-analytic frame:work (de:cision crees), though
e: diffe:renc counrries is data. bur these data be:come: inforrnation whe:n you you should reme:mber thar me proce:ss of making a decisin involve:s a great many
:ide: it is inreresting ro array the countries compararively (e.g., from lowesr to elernenrs prior to the: rnorncnt of actual choice, such as defining a useful problem,
;hest prcvalcnce}. Euidence is information rhat affects the cxisting ~c1ie:fs of thinking up better candidate: solurions, and selecting a usc:ful model. In general,
portant people (including yoursdf) about significant Features of [he problem me: value: of any pie:ce of evidence depcnds on
u are studying and how ir rnighc be: solved or rnirigared. Differenrial prevalencc
smoking, for instance, can becorne evidence be:aring on hypothc:sc:s about . me likc:lihood thar ir will cause: you to subsritute sorne beuer decisin for
fe:re:neial levels of conce:rn abour personal health across countries, I
whatever decision you would have rnade wirhout it (which might have
You need cvidence forthree principal purposes. One purpose is toasscss the been an "acceptable" decisin in and of itself):
:un: and extenc of rhe problem(s) you are r in to define:. A secon~ me: like:lihood rhar the substituted decisin will, dire:ctly or indire:ccly, pro-
: partlcular fe:arures o e:concrete: po i sitUarion you are: e:n ed in stud in . duce: a better policy outcorne than the outcorne that would have becn
0nsrance, y to ow-or guc:ss-about agency workloads, recen'[ produced by the original decisin;
dge:t figures, dernographic changc:s 'in a service area, the political ideology of me: magnitude of the difference in value: berween the likely-ro-be im-
: agency chief rhe competency of rhe middle-Ievd managers in the agency, prove:d ourcorne and me original outcome:.
:l me: current attirudcs of so me: other age:ncy rhat nominally coopc:rare:s wirh
Self-control. Ir is surprising how well you can do in many cases by garhering
s one on sornc problern. The third purpose is .so assess policie:s that have: been
,ught, by at leasr some people, to have: workcd cffecrivc:ly in situations appare:ndy no evidcnce at all bur simply sirring down and rhlnking sornerhing through and
ular to your own, in orhe:r jurisdictions, perh;p5, or ar other ti~.l,All thre:~ then making sorne serious educated guessc:s. There is nothing shame:ful about
acting on such gue:sstimatc:s and therehy conserving your dara-collecring time: and
[.poses are: rdevant to die: goa1 of produLng le~oje:ction;-of possible
.ley outcornes. e:nergic:s for answe:ring quesrions for which good evidence is re:ally ne:cc:ssary. See
Because each of thesc purpose:s becomes salienr in differcnr phasc:s of the Pare II, "Gathering Data for Policy Research."
A helpful check on yoursc:l{ to pn:ve:nt yoursclf from collecring useless data.
is to ask yourself the following quesrions be:fore:embarking on some: data colle:crion
Sometimos rhese siruacioru will have becn evaluarcd srarlstically, and somerirnes no" see
Parr IIl. "Sm.r< (Bese) Praccices' Research." ' venture:
-
10 A PRACTICAl GUIDE FOR POllCY ANAlYSIS
THE EIGHTFOLD PATH 11
..
"SUppOSC the data rurn OUt ro loole like so-and-so as opposed ro thus-
Use Analogies
and-such. What irnplication would chat-have for my undcrStanding of
how ro solve this problem?" Somedmes' ir pays to gathe:r data about things rhar are, on the surface, quite:
"Cornpared ro my ?est guess .about how the data willlook once I've goe unlike rhe problern you are studying but thar, under che surface, show insrructive
thcrn, how much diffcrcnr rnighr rhey look if 1 acrually took the trouble similarities. For instance, your understanding of how a rnerit pay plan for compen-
ro gee rhern?" sating managers in the public sector might work could perhaps be improved by
"How much is ir worrh ro me ro confirm -the actual di.fFercnce berween seeing how similar schernes work in me privare sector. Or if you are working on
what 1 c~~ guess and whar 1 can learn about me world by really getting the problem of how a stare can discipline, and perhaps disbar, incornperenr
rhe data? artorneys. you might usefully sperid a good deal of your rime learning abour how
the me:dical profession handles problems of physician incornperence. If you are
Ir is this son of crirical attitude abour me value of expensive data collection working.on how to reduce neighborhoods' resistance to accepring low-incorne
that often leads good and experienced policy analysts to make: do with back-of- housing projecrs, you could usefully look inro me lirerature on communiry resis-
the-e~velope. esrirnares. However. none of rhe above is meanc to be an excuse for tance ro accepting solid-waste incinerators.
shlr~(Jng the job of getting good data-and somctirnes ots of thern, at huge costs As rhese exarnples suggest. so me analogies are easicr to peroeive, and te malee
In nrne and money-.when yo~'ve convinced yourself rhac the investment really scnse or.' than cchers. le takes a lirtle imaginarion te see insrrucrive analogies .and,
wo.uldpay off. There: s an obvious and crirical diffen:nce berween jusrifiable and occasionaliy. a little daring to rey ro convince others to see both me usefulness
unjustifiable guesstimares. of the analogy and its inevitable limitations.
and sharing sornc of your own reactions with thern. You becorne more of a Try to inuent alternarives rhat rnight prove to be superior to the alterna-
partner in the process rhan an ourside observer and diagnostician-An even more ti ves currently being discussed by the key political actors. It's good to
cornplex and challenging role is for you to becorne a particular type of "parrner," brainsrorrn, to try ro be crcative=-but don't expect that you'll necessarily
a faciliraror and broker, whether by acting as a transmission belt from one person produce much berter ideas than other people have already thought abour,
to another or by convening meetings and other garherings. ! One way to coax your creaciviry is to refer ro che checklist in Appendix
A, "ThingsGovernments Do." For each entry on the lisr, ask yoursdf,
"Might ir make sense to rry so me version of this generic srraregy ro help
mirigate this problerni'" Because ir is a long and cornprehensive list, the
freeng the Captlve Mnd
answer with n:spect to any single srrategy will usually be negative. Going
In exchange for access to data and a ready-made worldview, researchers sornetirnes through the list systemarically is word~hile, however. Because the list
uncrirically accept problem definitions and preferred solutions from kindly infor- is not very long, with experience you willneed to spend only a few
manes (not to rnention from paying clients or employers). To' counter such minutes to decide whether any ideas there rnighr be worth considering
ternprations, be sure to make contact with individuals or factions whom you would further.
expecr to disagree-the more sharply the better=wirh your kindly informants. A A1ways include in your firsr approach tothe problern the alternative "Ler
time-saving, but only parcial, substituto is to ask your kindly informant, "Who present trends continue undisturbed." You nccd to do this because che
might objecr strongly to your point of view about this, and why might they world is full of natura!ly occurring change, and so me of these ongoing
do so?" changes mighr mirigare the problem on which you are working. (Nore
that 1 am not characcerizing this alrernarive as "Do Norhing.' Ir is not
possible to "do norhing.' Mosr of the trends in rnotion will probably per-
3. Construct the Alternatives sist and alter the problern, whether for better or for worse.)