Impact Analysis

What would have happened to those receiving intervention if they had not received the intervention? In scientific terms, you need a control group. Remember that a control group is not a population that remains unchanged. A control group is a population that is subject to everything the target group is subject to except the intended intervention. The important thing to know here is that there are many, many ways to end up with useless data, as this Economist article about randomized evaluations discusses. A randomized study showed that giving away mosquito nets for free was far more effective that charging anything.

You might conclude that the trial showed that they should always be given away. Yet it turns out that millions of nets were already in use in the part of Kenya where the field trial took place, so their value was known. The experiment guaranteed supplies, so it did not test the assertion that you need to charge something to encourage reliable suppliers. And the recipients were pregnant women, whereas the point of giving bednets away is to provide anti-malaria treatment universally. The evidence from western Kenya was clear. But it hardly settled the question of whether the government should give bednets away across the country.

As an aside, it seems like we could very profitably spend a few weeks on the scientific method directly, rather than orbiting it with alternate language.

Experiments

The best experiment possible: fully blind, randomized, large sample size, repeated.

Since this is rarely possible in economics and social science, especially at larger scales such as national development, we can use alternative methods:

• natural experiment. For example, when different states in the US which are otherwise similar adopt different policies.

  An example of a natural experiment occurred in Helena, Montana during the period from June 2002 to December 2002 when a smoking ban was in effect in all public spaces in Helena including bars and restaurants. Helena is geographically isolated and served by only one hospital. It was observed that the rate of heart attacks dropped by 60% while the smoking ban was in effect.—Wikipedia

• matching
• statistical equated controls
• non-experiments

Since we are still talking about cost-benefit analysis and how to apply it to situations with hard-to-value outcomes, here's an interesting article in the New York Times:

The Bush administration is about to propose far-reaching new rules that would give people with disabilities greater access to tens of thousands of courtrooms, swimming pools, golf courses, stadiums, theaters, hotels and retail stores.

... The Justice Department acknowledged that some of the changes would have significant costs. But over all, it said, the value of the public benefits, estimated at $54 billion, exceeds the expected costs of $23 billion.

If you've seen my sidewalks photo-essay, you'll know that Singapore isn't great with accessibility. This is not scientifically collected data, but I do see very very few disabled people, such as people in wheelchairs. in public here.

About half an hour before class was over, a camera crew from "Corporate Communications" came in to film the lecture in progress for some unspecified purpose. A few minutes later, we came to this slide in the lecture:

Design contamination refers to the situation where participants know that they are being observed (tested) and act differently because of it.

In the slightly stunned silence after they left, someone muttered, "this is contamination."

Evaluation

Programs convert inputs to outputs, which lead to outcomes. Process evaluation is a descriptive analysis, performed after implementation, which measures the efficiency of inputs to outputs. Impact analysis measures the relationship between outputs and outcomes and seeks causes.

Reading

Rossi, P., M. Lipsey and H. Freeman (2004) Evaluation: A Systematic Approach. 7th Edition. Sage Publications, Chapter 1—An Overview of Program Evaluation, pp 1-28

Rossi, P., M. Lipsey and H. Freeman (2004) Evaluation: A Systematic Approach. 7th Edition. Sage Publications, Chapter 12—The Social Context of Evaluation, pp 373-419

Discount Rates

A thousand dollars today is not the same as a thousand dollars in ten years. Cost-benefit analysis must account for changes in the value of money, i.e., inflation. This affects both costs and benefits.

Dealing with uncertainty

Several tools to account for uncertainty in cost-benefit analysis.

Sensitivity Analysis

Adjust some of the variables and see how much the projected outcomes change. Ignores relationships, especially non-linear relationships, between variables. This can be addressed by bundling various "consistent combinations" of changes into scenarios and comparing scenarios.

Monte Carlo Analysis

Estimate the probabilities of the different values of the key variables, including probabilities relative to other variables' changes. Use a computer to simulate thousands of different outcomes and see which are most likely. (Example)

Reading

Zerbe, Richard O., Jr., and Allen S. Bellas. 2007. A Primer for Benefit-Cost Analysis. Chapter 9-10, pp. 215-289. Edward Elgar Pub

This book says benefit-cost instead of cost-benefit. The difference has me counting syllables and emphases to figure out why it sounds worse. I think cost-benefit is iambic, or nearly so, as "cost" is de-emphasized. And analysis is purely iambic, so putting them all together is magical: "cost ben-e-fit a-nal-y-sys". But be-ne-fit-cost a-nal-y-sis sounds terrible.

• p 250. Using inflation numbers like 5% or 10% for discounting, when extended over a human lifespan or longer, produces a discount multiplier close to zero. That is, it says that the present value of something 70 or 100 years in the future is nothing. Taken at face value, this makes long-term cost-benefit analysis useless, and raises ethical questions. Instead, other rates such as "Social Rate of Time Preference" and "Shadow Price of Private Capital" can be used; charts on page 251 suggest values for time periods measured in decades and centuries.
• p 256: Risk is typically ignored in cost-benefit analysis. That's been very clear to me as a project manager doing estimates, and the key reason is that the users of the analysis don't want the risk. They just want a number. So I try to present a range instead of a number, but I think it just gets collapsed to the side of the range that the user wants to hear.

Cost-benefit analysis

Financial Analysis

Starts with the cash flow, the direct, measurable flow of money as a consequence of a policy alternative. Then broaden to include indirect costs, such as opportunity costs. IBM, for example, calls these green dollars and blue dollars. I once worked with a CIO who said that a particular policy option would be "free" because they could use their existing staff without extra training or having to hire or rent experts. This fallacy reflects a failure to understand "blue dollars".

Financial analysis vs cost-benefit analysis

To get to a true cost-benefit analysis, the scope of analysis must be widened even further. In addition to direct cash flow and indirect costs, complete cost-benefit analysis includes broader social costs and externalities.

Ex ante CBA is performed during planning, to inform decision-making. Ex post CBA is performed after a project is complete, to evaluate the outcome and add to general knowledge. It is also possible to do CBA in the middle of a project, and to compare ex post and ex ante CBAs to see how accurate the ex ante analysis was.

Kaldor-Hicks Criterion (an improvement on Pareto optimality): a policy should be adopted if the gainers could, in theory, compensate the losers and still be better off.

What's better, a cheap project with a very high benefit/cost ratio, or an expensive project with a lower benefit/cost ratio? They have different scales, and cannot be directly compared without more context.

Standing is very important: who and what should be counted in CBA?

Include only the changes in costs and benefits which are attributable to the alternative, i.e., the difference between baseline and the alternative. Exclude sunk outcomes. Exclude costs which are shared across all alternatives. Exclude transfer payments (not to be confused with transfer pricing), because they don't change the net cost or benefit, just the distribution. Treat taxes and subsidies case by case. Include true opportunity cost of government costs, not arbitrary prices. Avoid double-counting. Consider changes in asset value. Include externalities. Consider secondary outcomes. Include unpriced outcomes.

Reading

Boardman, A. E., D. H. Greenberg, A. R. Vining &D. L. Weimer. "Cost Benefit Analysis Concepts and Practice. Chapter 1

• p 5: Discussion of US federal requirements for performing CBA, starting in 1981. I think it's important to note (though the authors don't) that CBA is often a partisan political tool. A few ways it can be manipulated: require CBA for disfavored projects while not performing CBA on favored projects; control the scope of CBA to include or exclude important factors; selectively use or ignore the results of CBA.
• p 18: three perspectives, based on role. Analysts do real CBA. Guardians (e.g., OMB) tend to do only financial analysis. Spenders (in "service or line departments") tend to see all expenditures as benefits. They also weight their CBA heavily by political support.
• p 22: "CBA is often taught in a way that is completely divorced from political reality. We wish to avoid this mistake."

Sinder, J. A. & D. J. Thampapillai, "Introduction to Benefit-Cost Analysis", Chapter 4 & 5

• p 52: exclude international outcomes. Um. It's hard for me to get on board with that while still feeling like an ethical person.
• p 61: handy checklist of private vs social perspectives on various costs.
• p 62: Review questions. This looks like good practice.

Decision Matrices

We're looking at this chart on page 249 of "Expert Advice for Policy Choices". For some reason, the authors assert that both of the choices, pushing welfare recipients to get jobs, and "child support enforcement", are strongly supported by Conservatives, but "jobs" is only weakly supported by Liberals, and enforcement is actively opposed. Our overall topic is how to set up decision matrices to evaluate outcomes, but this example shows pretty clearly that this can be a very thin exercise. Let's catalog how it can go wrong:

• Are all of the important criteria included?
• Are all of the criteria weighted accurately?
• Can each criterion be accurately converted into a proportionate numerican value?
• Can you accurately determine the value of each alternative according to each criterion?
• Are all of the criteria mutually independent?

At the very least, it seems like malpractice to me to make and present measurements like this as single numbers instead of ranges.

Anyway. Three kinds of matrices:

• Alternatives by criteria.
• affected parties: what is the benefit or cost on each affected party for each alternative?
• political matrix: which constituencies support or oppose which alternatives?

Once you have a decision matrix, how do you reduce heterogeneous criteria scores to comparables?

• Qualitative:
  • Remove dominated alternatives. If an alternative is worse or tied with another alternative on every criterion, remove it. This can be combined with any method, qualitative or quantitative.
  • Rank the criteria. Whichever alternative is best by the #1 criterion wins. Ties go to the #2 criteria, etc. Assumes that the #1 criterion is more important than all others combined.
  • Pairwise. Compare two alternatives across all criteria and judge the winner. Compare the winner to the next alternative. Repeat. There can be only one.
• Quantitative
  • Satisficing. For each criterion, determine a minimum level to be acceptable. Score each alternative pass or fail for each criterion. This effectively quantifies everything, but does not address the relative value of different criteria.
  • Weighted criteria. Assign each criterion a value proportionate to its relative merit, and convert ratings within criteria to fractions of that value.

Another possibility is to combine elements of several alternatives.

Analytical reports should do as much as possible to simplify the decision for the decision-maker. If some alternatives are clearly excluded, they should be identified as such. If more information is needed to rank tied alternatives, that information should be precisely specified.

Readings

Bardach Part I, pp 47-59. Appendix A, pp 107-121

Weimer, D. & A. Vining (1999). Policy Analysis: Concepts and Practice, Chapter 1, "Review: The Canadian Pacific Salmon Fishery," pp 1-26

Weimer, D. & A. Vining (1999). Policy Analysis: Concepts and Practice, Chapter 11, "Goals/Alternatives Matrices: Some Examples from CBO Studies"

Class time is lagging quite a bit behind the syllabus, so I'm going to restructure each blog entry to contain a complete topic rather than match exactly what we did in class on a particular day.

Selecting criteria

Before you can choose between options, you have to assess which options are better or worse. Before you can do that, you need to define "better", but this is often very complicated. Also, the focal goal of a policy may bear little relation to its outcome and consequences. Criteria may come from many sources, from policy-makers to analysts to the public. Criteria may include:

• consequential criteria
  • good or bad outcome
  • who is affected?
  • efficiency
    • maximum output with minimum output
    • achieve objective with lowest possible cost
    • maximize net benefit for society
• moral criteria
  • right or wrong
  • Definitions of equality and equity
• Political viability
  • enough support
  • not too much opposition

Projecting outcomes

One the problem space is identified, the problem is defined, criteria are determined, and alternatives invented, the analyst must project outcomes. An outcome projection describes the outcome in terms of the criteria, including both the general direction and estimated magnitude of each value.

Extrapolation comprises three parts: secular trends, cyclical fluctuations (with seasonal variations as a special case), and irregular movement. In other words, what would happen if everything proceeded normally, what kinds of predictable periodic changes can be identified and filtered out, and what sorts of surprises may happen?

Extrapolations depend on models. Joel's note: Models 101: Models are simplified representations of the aspects of reality that are relevant to a specific problem. The path of a baseball in flight can be modeled fairly accurately with a simple Newtonian equations plus a fudge factor for air resistance. A whiffle ball is strongly affected by air and so the model for the baseball won't work as well for the whiffle ball; you'll need a different model. Everything you think you think in the context of a model, whether you realize that or not. Nobody uses formal Newtonian models to catch baseballs; but we still do have some model working in our brain based on our past observations of how baseballs behave. This can be proved by painting a whiffle ball to look like a baseball and then throwing it to someone expecting a baseball. They'll use the wrong model and fail to catch the ball. But if they know it's a whiffle ball, they'll adapt their working model, change their position, and make the catch. You should always be aware of what model you are using, its limitations, and alternative models you might need to use. End of Models 101.

Side note: the 51-49 principle (perhaps it could better be labeled "the 51-49 fallacy"?): Pushed by antagonistic group behavior, we treat our own 51% confidence as 100% confidence.

Judgmental forecasting: AKA expert knowledge forecast. Based on past experience and training, make a qualitative guess about the outcome.

The internet has a few catalogs of forecasting methods.

Feasibility

Is a proposed alternative politically possible?

Readings

Bardach Part 1, pp 25-47. Part II, pp 61-88

Patton, C & D. Sawicki (1986). Basic Methods of Policy Analysis and Planning. Englewood Cliffs: Prentice-Hall. Chapter 5, "Establishing Evaluative Criteria", Pages 139-175

Guess, G. M., and P. G. Farnham (2000), Cases in Public Policy Analysis. Chapter 4, Forecasting Policy Options, pp. 135-207, Georgetown University Press.

Policy Formulation and Policy Design

What is the scope of possible alternatives? Can the analyst propose new options or only evaluate already identified options? Status quo is usually an option.

Recipes for getting new solutions from existing solutions:

• Magnify
• minify
• substitute
• combine
• re-arrange

Generic policy options for correcting market failures (from Weimer):

• If there could be a working market but there isn't ...
  • Is the market too restrained?
    • Is it overly regulated, granting some actors monopoly rents and hence causing inefficiency? deregulate, but look out for windfall losses and gains during the transition. Examples: US deregulation of airlines, telephones (etc etc etc).
    • Is something illegal when it need not be? legalize. Example: drugs
    • Does the government own companies in an industry that isn't actually subject to market failures that can be solved by nationalized companies? Privatize.
  • Is there potential for a market but none exists yet?
    • Would assigning property rights create a healthy market? Then assign property rights. Example: water markets. Watch out, once again, for windfall losers and winners. Also, the new market may be very thin.
    • Could a new good be created with legal backing? Example: pollution permits.
  • Could a market be simulated?
    • An auction can expose non-market situations (natural monopolies, public goods) to market forces. Instead of competition within the market, which is hard or impossible, promote competition for access to the market.
• Is market failure endemic, or is something else more important than efficiency?
  • supply-side taxes
    • output taxes
    • tariffs
  • supply-side subsidies
    • matching grants
    • tax expenditures (business deductions and credits)
  • demand-side taxes
    • commodity taxes
    • user fees
  • demand-side subsidies
    • in-kind subsidies
    • vouchers
    • tax expenditures (personal deductions and credits)
• Rules
  • Framework rules
    • civil laws
    • criminal laws
  • Regulations
    • price regulation
    • quantity regulation
    • direct information
    • indirect information
• direct supply
  • government provision
  • independent agencies
    • government corporations
    • special districts
  • Contracting
    • direct contracting
    • indirect contracting

Risk-averse behavior

Scenario: a flu outbreak expected to kill 600 people, and two different programs to fight the outbreak. The class was split in two and each half presented with a choice. Half the students see two choices: option A saves 200 people, option B gives a 1/3 chance of saving 600 and a 2/3 chance of saving none. That group split between the choices. The second half had option C, dooming 400, and option D, a 1/3 chance of nobody dying and a 2/3 chance of 600 dying, and went unanimously for option D. Of course A and C are the same program, and B and D are the same program, and both programs have the same statistical outcome.

This was presented as an example of the power of framing, in that the choice between two options changed dramatically based on the wording. But I reasoned differently: this is a PR problem. (Since there's no statistical difference, neither option is obviously better from the information at hand.) The public isn't going to do policy analysis of a hypothetical 600 losses from flu. The only number that will matter to the press is the actual number of dead. From that perspective, 400 and 600 are the same number, and the real choice is:

• Option 1: certainty of failure
• Option 2: 1/3 chance of not failure, 2/3 chance of failure

and that choice is easy.

Readings

Correcting Market and Government Failures: Generic Policies, Policy Analysis: Concepts and Practice Chapter 9, Weimer and Vining. 1999.

Reading notes

Economies don't have purpose. They just happen. Just as wearing a striped shirt in front of a television will cause a pattern to appear; collecting a number of independent actors who can exchange things of value with one another will cause an economy to happen. There's no purpose for the moire pattern on TV; there's no purpose for an economy. It just is. Since material wealth (broadly defined to include drinking water, health care, etc) is the primary source of happiness, a good economy is better than a bad economy. A good economy is one with allocative efficiency: the most possible output for a given input. Of course that's not the only way to define "good" for an economy; other contenders include Pareto efficiency, equality, and fairness, but in purely economic terms the best economy is the most efficient one. The challenge for public managers is to balance the various definitions of "good" in a way that more or less reflects the preferences of the public, and having defined aggregate good, to do what they can to help achieve it. Last weeks' readings pointed out the danger of skipping that first part, the definition of the problem. This week's first reading assumes the problem is simply one of efficiency:

Boadway, R. & D. Wildasin (1984). Public Sector Economics, Chapter 3, "Market Failures"

A economic fundamentalist justification of the public sector's existence as a cure for (some) market failures.

• p 58: "... the jointness in use of public intermediate goods prevents the market from extracting a price from firms unless they can be excluded." Ack. No. Jointness of use does not prevent the market from extracting a price; only lack of excludability does. Jointness of use, more commonly (but no more usefully) called rivalry, is separate. Consider a golf course. It is, up to a point, non-rival. If I'm playing the tenth hole, I could care less whether or not somebody is playing the first hole. Within a certain range, let's say between one and a hundred golfers at a time, adding an extra golfer has no cost. That's great for the club because it provides them with a large "inventory" of time slots to sell. But it neither enables nor prevents them from "extracting a price". What enables the golf club to extract a price from golfers is the fence they put up around the course to keep the public out. That's excludability.
• p 73 to 77: charts of US government expenditure. Not adjusted for inflation, so mostly worthless; they do include percent of GDP, though. The total expenditure table does include state and local, which is nice. Still, I think this is pretty skippable.
• p 77: "Veterans Services", as it usually is, is mistakenly broken out into a separate category instead of included within "National Defense". If you send a million people to war and promise to take care of them, surely the lifelong cost of that care is part of the cost of war. (And if you manage that obligation as a cost to be cut, obscenities like this will be inevitable.

Kleiman, M., and S. Teles (2006). "Market and Non-Market Failures," in Moran, M., M. Rein and R. Goodin (eds) Oxford Handbook of Public Policy, Oxford University Press, pp. 624-650

Same topic, two decades later. Let's see if anything has improved.

• p 625: "Under certain highly restrictive assumptions, the market can be shown to be a a Pareto optimum." Those are better qualifiers for "assumption" than the last article's "It has been shown by Friedman and Savange (1948) that under certain reasonable assumptions individuals will behave [so as to maximize their expected utility]". Does that reflect 20 years of progress or are these authors just more rational (hah hah) to begin with?
• p 625: "The economic analysis underlying the doctrine of market failure assumes an individual capable of maximizing expected subjective utility ... That assumption is obviously false for children and the insane. But it is also false for many decisions made by ordinarily competent people about, for example, time management, saving, financial risk taking, diet, exercise, and the use of psychoactive chemicals. ... Thus the scope of suboptimal performance in voluntary individual choice and spontaneous organization is substantially larger than orthodox welfare-economics approaches suggest." Yay! Much better.
• p 637: causes of government failure: Inadequate Penetrative Capacity, i.e., not enough detailed information; Inadequate voluntary cooperation; institutional overhead; voter attention and inattention; path dependence of political decision-making; competition for technical expertise; weak administrative culture.
• p 647: "Public policy, institutional analysis, and political philosophy do not deal with three distinct subject matters; rather, they are three different attempts to deal with the problem of how human beings ought to govern themselves."

Much better.

Class notes

Class lecture starts with the material from class 2 reading.

What is public policy analysis?

Problem conditions, policy problems. Traffic congestion is a problem condition. The policy problem may be one or more of:

• gas price is too low
• import tariff is too low
• public transportation is inadequately used
• car culture dominates
• Insufficient support for bicycles

How do we know there's a problem? People complain.

Traffic congestion problem turns out to be a market failure because of externalities. Joel's Q: Is this a commons question? The common good is empty freeways, which get over-consumed? I think that's the same problem and same solution from a different analytical angle.

Some tools:

Boundary Analysis

Saturation sampling: sample everything. E.g., talk to everybody.

Sample and interview.

Boundary estimation

Joel's note: I bet card-sorting exercises (from usability testing) would be applicable. You would interview a sample of stakeholders to get things to put on the cards (in the prison example, phrases like "prison" and "sentencing guidelines" and "crack cocaine" and "building prisons"), and then give people stacks of cards and ask them to sort them into two piles, a pile of issues that are relevant to the problem and a pile of issues that aren't relevant. One challenge would be to get enough issues that are probably out of the boundary to be sure the boundary is big enough and well-defined. Hmm, looks like somebody's done something similar, though not exactly boundary analysis.

Classification Analysis

Hierarchy Analysis

Possible causes, plausible causes, actionable causes.

Multiple Perspective Analysis

Readings

Quade, E. S. (1982) Analysis for Public Decisions, Chapter 1, "The Need for Analysis"

Quade, E. S. (1982) Analysis for Public Decisions, Chapter 20, "Politics, Ethics, and Guidance from Analysis"

Executive summary: public policy analysis is thick with politics.

Weimer, D ... Policy Analysis

Howlett, M & M. Ramesh (2003) Studying Public Policy ...

Bardach Part I, pp 1-10

Dunn, William N, 2007. Public Policy Analysis: An Introductory .... Chapter 3, pp 71-120, 4th ed. Prentice Hall

• Policy problems are "unrealized needs, values, or opportunities for improvement."
• The biggest part of public policy solutions is discovering and defining problems, aka "structuring" problems.
• policy problems arise out of problem situations that are sensed by analysts and other stakeholders.
• policy problems are interrelated
• policy solutions go obsolete even though the problem may remain
• Errors of the third type, solving the wrong problem. Errors of the first type are false positives, or, in bad English, "rejecting the null hypothesis when it is true". Errors of the second type are false negatives, or, "accepting the null hypothesis when it is false".
• p 87. Descriptive and normative models. Finally, the right language choice. I don't quite follow the example of using the formula for compound interest is a normative model. I guess you could (and I guess the author does) consider the formula itself descriptive, but the act of using the formula normative since it pushes the analyst towards assuming that having the most money is the most important thing.
• p 92. Another sloppy example, in which drawing a graph (er, "simple symbolic model expressed in the form of a linear equation") relating rainfall and reservoir depth is alleged to lead an analyst to confusing correlation with causation, or of getting the causal arrow backwards. I understand that it's useful to remember that drawing a graph is an act that implies a certain model, and I understand that correlation should not be confused with causation, but I don't follow the path (to hell) between the two.
• p 99: rules for good classification systems
  1. Substantive relevance
  2. All things to be classified must fit
  3. Categories must be mutually exclusive
  4. Consistency (which follows from rules 2 and 3)

Class notes

"eightfold path to policy analysis": define the problem, assemble some evidence, construct the alternatives, select the criteria, project the outcomes, decide!, tell your story. Looks more or less like the scientific method. Researching a paper last year for States Markets, I came across this paper by Philip Gorski (that URL is specific to NUS's JSTOR access, but I bet you can hack it for your own institution), which argues that the Hypothetical-Deductive Model (the fancy label for the scientific method as more or less practiced in the 20th century) doesn't work well for social science, and a different model should be used. He argues for ... well, I am skimming his conclusion and it's so jargon-heavy that it's hard to summarize what he's arguing for. I think he's arguing that we can't use the pure HDM in social science, but we can and should still use the principle of falsifiablity.

Anyway, back in class discussion, we are still working through basic models of decision-making. A Chinese classmate offers this example of a limited decision-making model (my paraphrase): "When the Chinese government planned the Three Gorges Dam, they were concerned with energy and the economy. Then some experts said there would be environmental problems. But when the officials were planning the dam, they did not have that information." My own mental model of how that decision was made was that environmentalist voices were probably actively excluded; that officials were aware there was some environmental impact but simply didn't care until public (and foreign) pressure about the environment made it into a political issue.

We are still talking about the "incremental model". This overlaps closely with the PMI seminar I went to the other night, which was talking about how to handle risk. Nobody in class (to the extent that I've been listening, which is a smaller extent than it should be as I fiddle with JSTOR and Gorski and the lot) is talking about risk or how this model addresses risk.

Other models of decision making:

• Two-stage mixed scanning: first stage is groping around to set up the problem; second stage is rational decision model.
• garbage can model: all the inputs, context, etc, get thrown into the bin together and the decision/solution is the result of everything banging into each other. Or, as the original authors put it, "one can view a choice opportunity as a garbage can into which various kinds of problems and solutions are dumped by participants as they are generated. The mix of garbage in a single can depends on the mix of cans available, on the labels attached to the alternative cans, on what garbage is currently being produced, and on the speed with which garbage is collected and removed from the scene." The different things that go into the can are problems, solutions, participants, and choice opportunities. This was then somehow implemented in a computer simulation.

  A later article points out that "although the model's verbal description assumes the presence of decision makers, the FORTRAN code retains just problems, choices, and solutions as elements of the decision-making process." Class consensus: what's the point? I think that it's confusing to talk about this model without noting its roots in computer analysis of decision making. I guess I would summarize it thusly; the garbage can model is that people, problems, solutions, and opportunities all go zooming by more or less randomly. That's simply how it works in the real world, like it or not. So as a policy analyst you should focus on what you can control: coming up with solutions to problems that are current and that are "in play". And keeping history in context: there may be good solutions on the shelf, or good historical lessons about bad solutions; and your solution may in turn sit on the shelf.

• elitist model: the elitists get what they want.

Meanwhile, please enjoy this amusing picture.

Class discussion has worked back to "common sense" and "gut feeling" as decision-making processes. There are three interesting lines of thought here, in my opinion:

1. Research into how brains make decisions, for example the tendency to decide first and rationalize/explain second.
2. Antonio Damasio's books about the role of emotion in making decisions. Unfortunately the one I tried to read kept putting me to sleep, or perhaps I shouldn't have tried to read it at bedtime, so I never got past the second chapter. But it was very interesting.
3. Stephen Colbert's research into Truthiness. Somebody should get some research grants. Actually, if the show and audience ended up funding serious research in cognitive science, that would be totally awesome.

We seem to still be recapitulating the reading. It would be nice if we could start from the reading instead of ending there. For example, somebody just said it's important to be objective, but one of the readings, perhaps the Dunn, pointed out that analysts can only avoid intentional bias, not unintentional bias (and we know from much research that unintentional bias is incredibly powerful, from constraining the imagination of the possible to confirmation bias etc). While the tour through the methodology is interesting, we haven't caught up to the essential, unsolved problems in the field and in its foundations:

1. principle-agent problem. For practical reasons, the public has to delegate power to decision-makers. How can those elected or self-appointed decision-makers be trusted?
2. preference discovery and aggregation. how can we figure out what we collectively want?

Apropos to that, see this thread, about the new US Republican Party slogan, "The change you deserve." Highlights from the comments include, "Democracy is the theory that the common people know what they want, and deserve to get it good and hard," and "Mercy, High Ones. Not justice, please, not justice. We would all be fools to pray for justice," "You know, I used to think it was awful that life was so unfair. Then I thought, wouldn't it be much worse if life were fair, and all the terrible things that happen to us come because we actually deserve them? So, now I take great comfort in the general hostility and unfairness of the universe." (that one's more poignant because the actor being addressed died of a stroke aged 44) The slogan is already being used to sell an antidepressant, Effexor, which bears a warning that it may promote suicidal tendencies.

It's now noon; there is an interesting seminar in the same room at 12:15, and of course class at 2 pm (I was going to go home to walk the dog, and come back for the evening version of the 2 pm class, but it's raining and I think I'll just stick it out (Kona seems to be fine skipping the noon walk, not that I make a regular habit out of it, but it stresses me out to no end). I guess I'll try to get lunch after the talk and take it with me to 2 pm class. Grump grump grump.

Meanwhile we continue to recapitulate the reading; we're talking about the importance of defining the problem correctly before taking action to solve it. This is a huge point in the reading, and even that may understate the issue. As they say, "ἐπὶ δηλήσει δὲ καὶ ἀδικίῃ εἴρξειν", "never do harm to anyone", or less accurately but even more pointedly, "first, do no harm". I think that can be extended to, "zeroethly, stop doing harm."

The second core class for the special short semester, two three-hour classes per week for six weeks. Six classes on Policy Analysis, four on cost-benefit analysis, three on program evaluation.

One of the other students asked me how to blog at the end of last semester (that is to say, last Friday) and said he experimented with all of the big blogging sites over the weekend, so I hope to share a link to his new blog this week.