Revolvers’ Dollars: methodology and cautions

by and

In our Revolvers’ Dollars series, we built a panel dataset covering all active firm lobbyists between 1998 and 2012. This dataset contains 64,536 lobbyist-year observations, with 14,306 unique lobbyists each active for an average of 4.51 years. The sections below describe the methodology we used to build and analyze our dataset, and offer some cautions on the limitations of these data and methods.

Estimating salary: using revenue as a proxy

While we would ideally like to know lobbyists’ salaries directly, this information is not publicly disclosed. We can, however, estimate the revenue that individual lobbyists are associated with. Since generally lobbyists are compensated based on the business they generate, we can assume that these revenue numbers are a decent proxy for salary. Indeed, in a perfectly competitive market for lobbyists, their compensation would approach the amount of revenue they bring in. Because of the limits to this approach, we confine ourselves to looking at lobbyists who work in lobbying firms (“contract” lobbyists), excluding “in-house” lobbyists who exclusively represent one organization, following the lead of Bertrand, Bombardini and Trebbi, and Vidal, Draca and Fons-Rosen.

Revolvers' Dollars Methodology Excerpt Image

We estimate revenues for individual lobbyists using the weighted measure described by Vidal, Draca and Fons-Rosen using the Center for Responsive Politics’ bulk Lobbying Disclosure Act (LDA) data. For each LDA report we allocate a share of the total lobbying revenue to each lobbyist weighted inversely by the total number of lobbyists named in each section of the report. For example, an individual lobbyist listed alongside nine other lobbyists on a quarterly activity report with a value of $100,000 would be assigned $10,000 of revenue from that report. We repeat this allocation for all reports on which every lobbyist is listed for each year from 1998 to 2012, generating a revenue estimate for all years in which the lobbyist was active. To restrict our estimation to “active lobbyists,” we further limited our sample to lobbyists who were listed on contracts with non-zero values in at least two quarters in a given year. All values have been inflation adjusted to 2012 dollars.

Certainly, we don’t know how much of any given lobbyists’ total time these lobbying reports represent, and we don’t have access to the compensation structures inside lobbying firms. Still, we believe these numbers provide the best estimates available for describing the patterns of individual lobbyist revenue, and thus a reasonable proxy for their compensation.

It is important to note, however, that it is quite possible that many of these lobbyists (especially former members of Congress) actually earn substantially more in salary than they generate in revenue. They are likely to engage in a wider array of activities that do not meet the fairly narrow definition of lobbying under the Lobbying Disclosure Act, and thus remain unreported. It is also possible that many of the lobbyists who do not have government experience are career lawyers, who do some lobbying, but it does not account for the majority of their time.

As Sunlight Foundation Academic Fellow Tim LaPira explained recently on our blog, the number of government affairs professionals in Washington whose work would be counted under a more inclusive definition of lobbying is probably double what gets reported on lobbying disclosure forms, so we are obviously missing a good deal of activity.

Identifying lobbyists with previous government experience

Our method for identifying lobbyists with previous government experience is limited by our reliance on lobbyists’ self-reported disclosures. Lobbyists are required to report previous government experience in the “covered position” field on their Lobbying Disclosure Act forms. This is a free text field, and their self-disclosures are far from 100 percent reliable. It is not uncommon for lobbyists to misinterpret the question and list their current position (e.g. Senior Partner, CEO, Director of Government Affairs), rather than listing a former government position. This field suffers from situational LDA external validity problems, but the reliability is consistent and unbiased over time. However, they are the best publicly-available data that we have.

To clean this field we used a best-efforts string matching script to exclude titles that were unlikely to be legitimate government positions. The set of terms used to identify these positions can be found here. When positions were ambiguous (e.g. policy associate, with no mention of a government office), we erred in the direction of classifying them as non-revolvers. This approach is imperfect, but performed well on a set of hand-tagged titles. Any lobbyist who reported a title on any disclosure report in our dataset that our classifier identified as a government position was classified as a lobbyist with previous government experience.

Identifying congressional office alumni

We also used the “covered position” field on the LDA forms to identify lobbyists who worked in specific congressional offices. As a result, our identification suffers from the same limitations as our more general identification of revolvers. We looked for mentions of the last names of all previous members of Congress, in the titles that lobbyists listed. It is commonplace for lobbyists who worked in congressional offices to list the name of the representative or senator for whom they worked, but no doubt there are cases where lobbyists list a title but not the office. Our identification strategy, unfortunately, would misclassify such cases.

Additionally, as members of Congress are often listed by last name only (e.g. Rep. Smith), we were unable to disambiguate between lobbyists who worked for different members of Congress with the same last name. Because of this we chose to only present data about the top offices, where we could hand-verify that lobbyists were correctly classified. We restricted our sample to congressional offices for which we identified at least 10 unique lobbyists active at some point between 1998 and 2012. We excluded John, White, Johnson and Davis and Washington, because of the high degree of misspecification and difficulty with disambiguation as well as Frank and Dodd, which were mostly mentioned in reference to the recent financial regulation. Revenue from lobbyists who mentioned more than one member of Congress was counted towards the totals for all members they listed.

As always, our data are only as reliable as the forms from whence they come. However, despite these limitations, we are confident that the numbers we’ve produced offer the best estimates available using public data, given the limitations we’ve described above.

You can download the panel dataset of active firm lobbyists here.