Tips on Writing a Policy Memo

The memo should be no longer than necessary. Don't leave important things out but don't write a paragraph when a sentence or two will do.

Bad: "Some people will react to the increase in the gas tax by taking taxis, buses or other public transportation since those forms of transportation are now somewhat less expensive than using a private automobile. However, some people may not be willing or able to make such a change: they might live far from public transporation or might have medical conditions that made it necessary for them to drive. These people will continue driving but they would generally be likely to take fewer trips than before the tax."

Problem: Far too many words for the basic points being made.

Better: "A higher gas tax would reduce the amount of driving by people who can easily use other forms of transportation. People who can't switch would continue to drive, although they would probably drive fewer miles than before."

Don't ask people to take your results on faith. Briefly explain the key mechanisms and make clear how your analysis links to underlying data, and be sure to indicate the original source of the data.

Bad: "The $7,500 tax credit will increase the number of electric vehicles sold by 75,000 per year."

Problem: The reader has to take your word for it with no supporting evidence: it's not clear what underlying data you used to reach the conclusion.

Better: "Survey X of vehicle manufactures (or whatever the source was) has shown that electric vehicle sales rise by 10,000 for every $1,000 reduction in price. As a result, the $7,500 credit could be expected to raise sales by 75,000 vehicles."

Although it's important to explain your results briefly, avoid going step by step through the calculations. The details can be put in a separate report to be circulated to people interested in the technical issues.

Bad: "To calculate the effect of the proposed $5 tax on each new tire sold, it is first necessary to convert the tax to an equivalent percentage change in the price. The average price of a tire is $100 so the tax would raise the price by 5%. Next, the percentage change in the price is multiplied by the price elasticity of demand for the good. The elasticity shows the percentage change in the quantity demanded for a percentage change in the good's price. Multiplying it by the price change thus shows the percentage change in the amount of the good that will be demanded. The demand elasticity for tires has been estimated to be in the range of -0.2 to -0.4. Thus, the 5% increase in tire prices would reduce tire sales by 1% to 2%."

Problem: The reader needs to know how the answer is linked to the input data but does not want to wade through a blow by blow description of the calculation.

Better: "The tax would raise tire prices by 5%. Census data from 2010-2015 shows that tire sales fall by 2-4% for a 10% increase in price, so in this case it is likely that tire sales would fall by 1% to 2%."

The only time you need to put in much detail about the underlying calculations is when there are multiple competing methods that give conflicting results. For example, suppose you needed to report the average number of cigarettes consumed by women between the ages of 19 and 25. It would be important for the reader to know whether the data was obtained by surveying women about their smoking habits or by observing the sales of cigarettes. People tend to underreport their consumption of cigarettes so survey data would be biased downward.

It's very important to figure out who would be helped and who would be hurt by a proposal. After all, the point of public policy is to solve public problems and thereby make people (at least some people) better off. No policy analysis is really complete until the winners and losers have been identified.

In addition, knowing who gains and loses can be very helpful in anticipating how the political process will play out. If compensation for the losers can be built into the policy, which is always potentially possible for efficient policies, it can help reduce political obstacles.

Bad: "Eliminating rent control will benefit an average tenant by $75."

Problem: Too little detail; it sounds like everyone gains when really some tenants lose.

Better: "Eliminating rent control will cause rents to rise by $600 for 500 tenants currently in rent controlled apartments. However, it will also bring 300 new apartments on the market. The average value of each new apartment to its tenant will be $1,200 above the amount the tenant pays in rent."

Look over your preliminary results and try to anticipate what questions a reader would be most likely to ask, or what additional calculations she might want to do to understand your results fully.

Bad: "The policy reduces the average wages for entry level clerical workers from $6.00 per hour to $5.50 per hour."

Problem: Virtually every reader will mentally convert this to a percentage to gauge how important it is.

Better: "The policy reduces the average wages for entry level clerical workers from $6.00 to $5.50, a decline of 8.3%."

Avoiding jargon will make it easier for your results to be understood by a wide range of readers.

Bad: "The income elasticity of medical care is 1.5 so a 20% increase in average household income would increase the consumption of medical care by 30%."

Problem: Incomprehensible to anyone not thoroughly trained in economics.

Better: "Analysis by X of historical data on medical expenditures shows that a 1% increase in household income leads to a 1.5% increase in expenditure on medical care. Since the proposed policy would raise household income by 20%, it should raise the demand for medical care by 30%.

A table can often be worth a thousand words. It's a much faster way to present a set of numbers than to include them in the text and it can be a lot easier for a reader to understand. That's especially true when you need to show results for different demographic groups.

Make sure the rows and columns of the table are labeled clearly and avoid including unimportant data.

Bad: "The benefit of the food stamp program to eligible households having two children would be $5,000. These households have an average income of $15,000 so the benefit would be equivalent to a 33% increase in their income. Eligible households with three children would receive $6,000 in benefits and have an average income of $17,000 (benefits are 35% of income). Households with four or more children have an average income of $18,000 and would receive $6,500 (36% of income). Households with one child would receive $4,000 and households with no children would receive $3,000. The average incomes of these two groups is $14,000 and $18,000 respectively."

Problem: TL;DR. Lots of information but it's tedious to read and it's hard for the reader to compare across groups. People may skip it.

Better: "Table 1 shows the food stamp benefits and average income for eligible households with different numbers of children. Also shown is the benefit as a percent of the household's income."

Table 1: Benefits received by households with different numbers of children

You'll usually know more about the policy problem and the analytical methods you use than the reader of the memo. That makes it easy to inadvertently write the memo as though the reader were ignorant or stupid. The problem is subtle because it usually happens as a result of good intentions on your part: trying to be as clear as possible. However, you can end up "wonksplaining": tiresomely explaining things that the reader already knows or can easily figure out.

Bad: "The policy will lower the price faced by consumers.  When a good's price declines, consumers generally buy more of the good.  Therefore, we would expect more units of the good to be sold."

Problem: It's both basic economics and common sense.

Better: "The policy will lower the price faced by consumers and thus increase total sales of the good."

Wherever possible, the memo should include all the facts a policy maker would need to reach her own conclusions and should not emphasize your personal opinion.

Bad: "US car manufacturers would gain by $1 billion per year if fuel economy standards were relaxed while US consumers would not be hurt significantly."

Problem: The reader doesn't know what you consider significant.

Better: "US car manufacturers would save $1 billion per year in costs if federal fuel economy standards were relaxed. The net effect on consumers is much smaller: the reduction in vehicle prices would save a typical household $200 but the drop in fuel economy would increase gas expenditure by $220."

Please note: these figures were made up as an illustration and are not based on real facts about US automobile industry.

Finally, focus on whether the policy is a good means for achieving its stated or implicit purpose, not whether the purpose itself is good or bad. For example, suppose Policy A intended to help group X is grossly inefficient. The memo could argue that Policy A is a bad way to help group X (i.e., that there are other policies that could achieve the same goal with less waste). However, it should avoid arguing that helping group X is good or bad per se. That is, take it as given that group X is to be helped and evaluate whether or not the policy is a good way of doing so.

Problem: Helping group X is the purpose of the policy so this is a statement about the goals of the policy, not about its means. It's not addressing whether or not Policy A is the best way to help group X.

Better: Policy A helps group X but is much more expensive than delivering the same benefits via Policy B.