Policy Analysis in 750 words: Rachel Meltzer and Alex Schwartz (2019) Policy Analysis as Problem Solving

Please see the Policy Analysis in 750 words series overview before reading the summary. This post might well represent the largest breach of the ‘750 words’ limit, so please get comfortable. I have inserted a picture of a cat hanging in there baby after the main (*coughs*) 1400-word summary. The rest is bonus material, reflecting on the links between this book and the others in the series.

Rachel Meltzer and Alex Schwartz (2019) Policy Analysis as Problem Solving (Routledge)

‘We define policy analysis as evidence-based advice giving, as the process by which one arrives at a policy recommendation to address a problem of public concern. Policy analysis almost always involves advice for a client’ (Meltzer and Schwartz, 2019: 15).

Meltzer and Schwartz (2019: 231-2) describe policy analysis as applied research, drawing on many sources of evidence, quickly, with limited time, access to scientific research, or funding to conduct a lot of new research (2019: 231-2). It requires:

• careful analysis of a wide range of policy-relevant documents (including the ‘grey’ literature often produced by governments, NGOs, and think tanks) and available datasets
• perhaps combined with expert interviews, focus groups, site visits, or an online survey (see 2019: 232-64 on methods).

Meltzer and Schwartz (2019: 21) outline a ‘five-step framework’ for client-oriented policy analysis. During each step, they contrast their ‘flexible’ and ‘iterative’ approach with a too- rigid ‘rationalistic approach’ (to reflect bounded, not comprehensive, rationality):

1. ‘Define the problem’.

Problem definition is a political act of framing, not an exercise in objectivity (2019: 52-3). It is part of a narrative to evaluate the nature, cause, size, and urgency of an issue (see Stone), or perhaps to attach to an existing solution (2019: 38-40; compare with Mintrom).

In that context, ask yourself ‘Who is defining the problem? And for whom?’ and do enough research to be able to define it clearly and avoid misunderstanding among you and your client (2019: 37-8; 279-82):

• Identify your client’s resources and motivation, such as how they seek to use your analysis, the format of analysis they favour, their deadline, and their ability to make or influence the policies you might suggest (2019: 49; compare with Weimer and Vining).
• Tailor your narrative to your audience, albeit while recognising the need to learn from ‘multiple perspectives’ (2019: 40-5).
• Make it ‘concise’ and ‘digestible’, not too narrowly defined, and not in a way that already closes off discussion by implying a clear cause and solution (2019: 51-2).

In doing so:

• Ask yourself if you can generate a timeline, identify key stakeholders, and place a ‘boundary’ on the problem.
• Establish if the problem is urgent, who cares about it, and who else might care (or not) (2019 : 46).
• Focus on the ‘central’ problem that your solution will address, rather than the ‘related’ and ‘underlying’ problems that are ‘too large and endemic to be solved by the current analysis’ (2019: 47).
• Avoid misdiagnosing a problem with reference to one cause. Instead, ‘map’ causation with reference to (say) individual and structural causes, intended and unintended consequences, simple and complex causation, market or government failure, and/ or the ability to blame an individual or organisation (2019: 48-9).
• Combine quantitative and qualitative data to frame problems in relation to: severity, trends in severity, novelty, proximity to your audience, and urgency or crisis (2019: 53-4).

During this process, interrogate your own biases or assumptions and how they might affect your analysis (2019: 50).

2. ‘Identify potential policy options (alternatives) to address the problem’.

Common sources of ideas include incremental changes from current policy, ‘client suggestions’, comparable solutions (from another time, place, or policy area), reference to common policy instruments, and ‘brainstorming’ or ‘design thinking’ (2019: 67-9; see box 2.3 and 7.1, below, from Understanding Public Policy).

Identify a ‘wide range’ of possible solutions, then select the (usually 3-5) ‘most promising’ for further analysis (2019: 65). In doing so:

• be careful not to frame alternatives negatively (e.g. ‘death tax’ – 2019: 66)
• compare alternatives in ‘good faith’ rather than keeping some ‘off the table’ to ensure that your preferred solution looks good (2019: 66)
• beware ‘ best practice’ ideas that are limited in terms of (a) applicability (if made at a smaller scale, or in a very different jurisdiction), and (b) evidence of success (2019: 70; see studies of policy learning and transfer)
• think about how to modify existing policies according to scale or geographical coverage, who to include (and based on what criteria), for how long, using voluntary versus mandatory provisions, and ensuring oversight (2019: 71-3)
• consider combinations of common policy instruments, such as regulations and economic penalties/ subsidies (2019: 73-7)
• consider established ways to ‘brainstorm’ ideas (2019: 77-8)
• note the rise of instruments derived from the study of psychology and behavioural public policy (2019: 79-90)
• learn from design principles, including ‘empathy’, ‘co-creating’ policy with service users or people affected, ‘prototyping’ (2019: 90-1)

3. ‘Specify the objectives to be attained in addressing the problem and the criteria to  evaluate  the  attainment  of  these  objectives  as  well as  the  satisfaction  of  other  key  considerations  (e.g.,  equity,  cost, equity, feasibility)’.

Your objectives relate to your problem definition and aims: what is the problem, what do you want to happen when you address it, and why?

• For example, questions to your client may include: what is your organization’s ‘mission’, what is feasible (in terms of resources and politics), which stakeholders to you want to include, and how will you define success (2019: 105; 108-12)?

In that values-based context, your criteria relate to ways to evaluate each policy’s likely impact (2019: 106-7). They should ensure:

• Comprehensiveness. E.g. how many people, and how much of their behaviour, can you influence while minimizing the ‘burden’ on people, businesses, or government? (2019: 113-4)
• Mutual Exclusiveness. In other words, don’t have two objectives doing the same thing (2019: 114).

Common criteria include (2019: 116):

1. Effectiveness. The size of its intended impact on the problem (2019: 117).
2. Equity (fairness). The impact in terms of ‘vertical equity’ (e.g. the better off should pay more), ‘horizontal equity’ (e.g. you should not pay more if unmarried), fair process, fair outcomes, and ‘intergenerational’ equity (e.g. don’t impose higher costs on future populations) (2019: 118-19).
3. Feasibility (administrative, political, and technical). The likelihood of this policy being adopted and implemented well (2019: 119-21)
4. Cost (or financial feasibility). Who would bear the cost, and their willingness and ability to pay (2019: 122).
5. Efficiency. To maximise the benefit while minimizing costs (2019: 122-3).

 

4. ‘Assess the outcomes of the policy options in light of the criteria and weigh trade-offs between the advantages and disadvantages of the options’.

When explaining objectives and criteria,

• ‘label’ your criteria in relation to your policy objectives (e.g. to ‘maximize debt reduction’) rather than using generic terms (2019: 123-7)
• produce a table – with alternatives in rows, and criteria in columns – to compare each option
• quantify your policies’ likely outcomes, such as in relation to numbers of people affected and levels of income transfer, or a percentage drop in the size of the problem, but also
• communicate the degree of uncertainty related to your estimates (2019: 128-32; see Spiegelhalter)

Consider using cost-benefit analysis to identify (a) the financial and opportunity cost of your plans (what would you achieve if you spent the money elsewhere?), compared to (b) the positive impact of your funded policy (2019: 141-55).

• The principle of CBA may be intuitive, but a thorough CBA process is resource-intensive, vulnerable to bias and error, and no substitute for choice. It requires you to make a collection of assumptions about human behaviour and likely costs and benefits, decide whose costs and benefits should count, turn all costs and benefits into a single measure, and imagine how to maximise winners and compensate losers (2019: 155-81; compare Weimer and Vining with Stone).
• One alternative is cost-effectiveness analysis, which quantifies costs and relates them to outputs (e.g. number of people affected, and how) without trying to translate them into a single measure of benefit (2019: 181-3).
• These measures can be combined with other thought processes, such as with reference to ‘moral imperatives’, a ‘precautionary approach’, and ethical questions on power/ powerlessness (2019: 183-4).

 

5. ‘Arrive at a recommendation’.

Predict the most likely outcomes of each alternative, while recognising high uncertainty (2019: 189-92). If possible,

• draw on existing, comparable, programmes to predict the effectiveness of yours (2019: 192-4)
• combine such analysis with relevant theories to predict human behaviour (e.g. consider price ‘elasticity’ if you seek to raise the price of a good to discourage its use) (2019: 193-4)
• apply statistical methods to calculate the probability of each outcome (2019: 195-6), and modify your assumptions to produce a range of possibilities, but
• note Spiegelhalter’s cautionary tales and anticipate the inevitable ‘unintended consequences’ (when people do not respond to policy in the way you would like) (2019: 201-2)
• use these estimates to inform a discussion on your criteria (equity, efficiency, feasibility) (2019: 196-200)
• present the results visually – such as in a ‘matrix’ – to encourage debate on the trade-offs between options
• simplify choices by omitting irrelevant criteria and options that do not compete well with others (2019: 203-10)
• make sure that your recommendation (a) flows from the analysis, and (b) is in the form expected by your client (2019: 211-12)
• consider making a preliminary recommendation to inform an iterative process, drawing feedback from clients and stakeholder groups (2019: 212).

 

 

Policy analysis in a wider context

Meltzer and Schwartz’s approach makes extra sense if you have already read some of the other texts in the series, including:

1. Weimer and Vining, which represents an exemplar of an X-step approach informed heavily by the study of economics and application of economic models such as cost-benefit-analysis (compare with Radin’s checklist).
2. Geva-May on the existence of a policy analysis profession with common skills, heuristics, and (perhaps) ethics (compare with Meltzer and Schwartz, 2019: 282-93)
3. Radin, on:

• the proliferation of analysts across multiple levels of government, NGOs, and the private sector (compare with Meltzer and Schwartz, 2019: 269-77)
• the historic shift of analysis from formulation to all notional stages (contrast with Meltzer and Schwartz, 2019: 16-7 on policy analysis not including implementation or evaluation)
• the difficulty in distinguishing between policy analysis and advocacy in practice (compare with Meltzer and Schwartz, 2019: 276-8, who suggest that actors can choose to perform these different roles)
• the emerging sense that it is difficult to identify a single client in a multi-centric policymaking system. Put another way, we might be working for a specific client but accept that their individual influence is low.

4. Stone’s challenge to

• a historic tendency for economics to dominate policy analysis,
• the applicability of economic assumptions (focusing primarily on individualist behaviour and markets), and
• the pervasiveness of ‘rationalist’ policy analysis built on X-steps.

Meltzer and Schwartz (2019: 1-3) agree that economic models are too dominant (identifying the value of insights from ‘other disciplines – including design, psychology, political science, and sociology’).

However, they argue that critiques of rational models exaggerate their limitations (2019: 23-6). For example:

• these models need not rely solely on economic techniques or quantification, a narrow discussion or definition of the problem, or the sense that policy analysis should be comprehensive, and
• it is not problematic for analyses to reflect their client’s values or for analysts to present ambiguous solutions to maintain wide support, partly because
• we would expect the policy analysis to form only one part of a client’s information or strategy.

Further, they suggest that these critiques provide no useful alternative, to help guide new policy analysts. Yet, these guides are essential:

‘to be persuasive, and credible, analysts must situate the problem, defend their evaluative criteria, and be able to demonstrate that their policy recommendation is superior, on balance, to other alternative options in addressing the problem, as defined by the analyst. At a minimum, the analyst needs to present a clear and defensible ranking of options to guide the decisions of the policy makers’ (Meltzer and Schwartz, 2019: 4).

Meltzer and Schwartz (2019: 27-8) then explore ways to improve a 5-step model with insights from approaches such as ‘design thinking’, in which actors use a similar process – ‘empathize, define the problem, ideate, prototype, test and get feedback from others’ – to experiment with policy solutions without providing a narrow view on problem definition or how to evaluate responses.

Policy analysis and policy theory

One benefit to Meltzer and Schwartz’s approach is that it seeks to incorporate insights from policy theories and respond with pragmatism and hope. However, I think you also need to read the source material to get a better sense of those theories, key debates, and their implications. For example:

1. Meltzer and Schwartz (2019: 32) note correctly that ‘incremental’ does not sum up policy change well. Indeed, Punctuated Equilibrium Theory shows that policy change is characterised by a huge number of small and a small number of huge changes.

• However, the direct implications of PET are not as clear as they suggest. Baumgartner and Jones have both noted that they can measure these outcomes and identify the same basic distribution across a political system, but not explain or predict why particular policies change dramatically.
• It is useful to recommend to policy analysts that they invest some hope in major policy change, but also sensible to note that – in the vast majority of cases – it does not happen.
• On his point, see Mintrom on policy analysis for the long term, Weiss on the ‘enlightenment’ function of research and analysis, and Box 6.3 (from Understanding Public Policy), on the sense that (a) we can give advice to ‘budding policy entrepreneurs’ on how to be effective analysts, but (b) should note that all their efforts could be for nothing.

2. Meltzer and Schwartz (2019: 32-3) tap briefly into the old debate on whether it is preferable to seek radical or incremental change. For more on that debate, see chapter 5 in the 1^st ed of Understanding Public Policy in which Lindblom notes that proposals for radical/ incremental changes are not mutually exclusive.
3. Perhaps explore the possible tension between Meltzer and Schwartz’s (2019: 33-4) recommendation that (a) policy analysis should be ‘evidence-based advice giving’, and (b) ‘flexible and open-ended’.

• I think that Stone’s response would be that phrases such as ‘evidence based’ are not ‘flexible and open-ended’. Rather, they tend to symbolise a narrow view of what counts as evidence (see also Smith, and Hindess).
• Further, note that the phrase ‘evidence based policymaking’ is a remarkably vague term (see the EBPM page), perhaps better seen as a political slogan than a useful description or prescription of policymaking.

 

Finally, if you read enough of these policy analysis texts, you get a sense that many are bunched together even if they describe their approach as new or distinctive.

• Indeed, Meltzer and Schwarz (2019: 22-3) provide a table (containing Bardach and Patashnik, Patton et al, Stokey and Zeckhauser, Hammond et al, and Weimer & Vining) of ‘quite similar’ X-step approaches.
• Weimer and Vining also discuss the implications of policy theories and present the sense that X-step policy analysis should be flexible and adaptive.
• Many texts – including Radin, and Smith (2016) – focus on the value of case studies to think through policy analysis in particular contexts, rather than suggesting that we can produce a universal blueprint.

However, as Geva-May might suggest, this is not a bad thing if our aim is to generate the sense that policy analysis is a profession with its own practices and heuristics.