6. Why towns get what they get | Commonwealth

Section summary

Size and need explain most of the variation in earmarks and competitive grants; power adds almost nothing measurable to the prediction. Size, need, capacity and local institutions predict about 56% of the cross-town variation in ten-year earmark dollars placed in the town and 58% in competitive grants. Adding leadership posts, stipends, loyalty, campaign giving and connections adds less than one point out of sample. Of the associations that hold, the largest is the Senate gavel: towns represented for more of the decade by the Senate President or the Senate Ways and Means chair received 85% more earmark money per standard deviation of exposure than the model expects (p < 0.001).

Sen. Michael Rodrigues of Westport became chair of Senate Ways and Means in February 2019. In the three budgets before, FY2017–FY2019, his First Bristol and Plymouth district received $361,667 a year in earmarks naming its towns. In the seven enacted while he held the chair, FY2020–FY2026, it received $1.08 million a year, 2.99 times as much. Statewide, earmarks naming a town rose 1.95 times between the same periods. On the second measure, which adds organizations placed in the district, its money grew 2.23 times and the state’s 2.20.

The presidency left Stanley Rosenberg’s Hampshire, Franklin and Worcester district after the FY2018 budget. In his two budgets as President it received $637,500 a year in earmarks naming its towns; in the eight that followed, $694,365 a year, 9% more, while the statewide figure rose 2.12 times.

The money is counted where it lands, whoever asked for it, on each budget’s district lines. Two districts are anecdotes, and they point in the direction of the test across all 351 towns, where gavel exposure goes with 85% more earmark money per standard deviation than the model expects.

Sources: General Appropriations Acts FY2017–FY2026 (malegislature.gov); Parcenomics analysis of earmarks by Senate district, each budget on the lines of the General Court that enacted it; Senate leadership from malegislature.gov.

The question for each town is how much of what it received can be explained by things that ought to matter, and how much by things that ought not to. The test covers the two discretionary channels, earmarks placed in the town and competitive grants, which are about 3% of the money; formula aid is set by statute and is used here as a control.

The model starts with rational factors: population, road miles, employment, income, poverty, property wealth, Chapter 70 share, a capacity proxy, the presence of a hospital or college, the number of nonprofits, county-seat and Gateway City status, and formula aid. Tested on towns left out of each fit (out of sample), these factors predict about 56% of the variation across towns in ten-year earmark dollars and 58% in competitive grants. Within the fit (in sample), size alone accounts for 53% of the earmark variation; need adds about four points and capacity two.

Then the model adds what should not matter: leadership posts held by the town’s representative and senator, stipend levels, seniority and party, documented breaks with leadership, residents’ giving to leadership committees, and ties between recipients and officials. Together they add less than one point to the prediction for towns left out of the fit. In a side-by-side run, the earmark figure moves from 56.2% to 56.2%; for competitive grants it falls, from 58.5% to 56.8%.

Formula aid per resident, counted gross, is one of the controls. Holding size and need constant, towns with more of it also received more in earmarks (34% per standard deviation, p = 0.013) and competitive grants (28%, p < 0.001), so discretionary money does not offset formula aid.

Most of the remaining 44% is project-by-project variation that no town-level factor reaches.

The power signals that hold

A few associations survived the checks we ran: the data corrections of October 18 and 19, the rerun with gross formula aid, county controls, and dropping Boston or the three metropolitan counties. They come from 516 cross-town tests, so a single result near p = 0.03 deserves caution. They are associations, not causes.

Senate gavel. Towns represented for more of the decade by the Senate President (Stanley Rosenberg to December 2017, Harriette Chandler as acting President and then President to July 2018, Karen Spilka since) or the Senate Ways and Means chair (Spilka to February 2019, Michael Rodrigues since) received more earmark money than the model expects for towns of their size and need: 85% more per standard deviation of exposure, about 1.6 of the ten budgets (p < 0.001), with each budget credited to the President and chair who held the posts when it was enacted. As a yes-or-no measure (45 towns ever represented), it is 45% per standard deviation (p = 0.02). In raw earmark dollars per resident the holders’ districts rank 7th to 30th of 54 Senate districts counting everything placed (6th to 33rd on the second measure, which the model uses). The result holds with county controls, without Boston or the three metropolitan counties, and without any one holder’s district (79% to 95%, p < 0.001); crediting the 2017–2018 presidency by General Court rather than by budget gives 90%. Within a town, the data would seem to indicate that gaining the gavel also coincides with more earmark money (p = 0.01, 40 towns with a change); before the presidency coding was corrected, that within-town result was not significant (p = 0.06). The data would seem to indicate that gavel exposure also goes with larger individual items, 11% per standard deviation (p = 0.028). When money traced to the Ways and Means chairs’ corrective amendments is taken out of the outcome, that per-item result loses significance (p = 0.079); the gavel result among towns with any earmark (p = 0.04) and the all-town result (+71%, p < 0.001) hold.
Senate assistant leaders and whips. The data would seem to indicate that towns represented by senators in these posts received more: 91% per standard deviation (p = 0.025), holding with county controls and without Boston. These posts would also seem to go with larger items, 16% per standard deviation (p = 0.008). No House post is associated with larger items.
House Republicans. Among towns that received any earmark, Republican House representation is associated with 15% less per standard deviation (p = 0.002).
Breaking with leadership. Within a town, a House member’s documented break with leadership coincides with about 90% less in later budgets (p = 0.003, 49 towns). The list of breaks is short and hand-coded.
Senate stipend and seniority, within a town. A rise in the senator’s stipend coincides with more earmark money (p = 0.027); this captures movement into posts. More seniority coincides with slightly less (p = 0.008).

Competitive grants show no power signal in the main specification: no leadership, stipend or seniority measure is significant across or within towns. Per-resident and per-mile versions show a few scattered results in both directions. Residents’ campaign giving to leadership committees shows no association with either channel.

Two cautions. House leadership exposure cannot be separated from geography: 87% of the population represented by a House gavel holder lives in Suffolk, Middlesex or Norfolk counties, and 37% in Boston. Those holders were Speakers Robert DeLeo (to December 2020) and Ronald Mariano, and Ways and Means chairs Brian Dempsey (to July 2017), Jeffrey Sánchez (July 2017 through the 2017–2018 session) and Aaron Michlewitz (since 2019). And the largest items mostly have no filed amendment behind them. Only 24 of the 61 clauses of $950,000 or more placed in a town (second measure) match a member’s amendment. They originate in Ways and Means text or in conference, where the public record does not say who asked.