2024
D+39.2marginKamala HarrisDem
554,471
68.0%
Donald TrumpRep
235,118
28.9%
All other candidates
25,243
3.1%
814,832 ballots cast · 49.0 per 100 residents (2024 population estimate)
Residents1,663,929+2.4% since 2020 (+39,468 people)
Middlesex County
Massachusetts·FIPS 25017·county nonfunctioning·1,669,979 residents (2025 estimate)
2024 result
D+39.2
814,832 ballots cast
What happened: the county's margin each cycle, against the national result.
Lean vs. nation
D+42.0
How this county sits relative to the country in 2024, with the national environment removed.
Trend since 2000
+8.6 pts
Net movement of that lean across 7 cycles — toward Democrats.
Elasticity
-0.05
Counter-trend: moves against the national tide.
Bellwether error
33.6
Average gap between this county's margin and the nation's.
4 consecutive cycles moving toward Democrats relative to the nation.
Votes by candidate
7 cyclesCertified vote counts for each party's nominee, the combined total for everyone else on the ballot, the number of ballots cast, and how many people lived here at the time.
2020
D+44.9marginJoe BidenDem
617,196
71.0%
Donald TrumpRep
226,956
26.1%
All other candidates
25,195
2.9%
869,347 ballots cast · 53.5 per 100 residents (2020 population estimate)
Residents1,624,461+1.9% since 2016 (+29,797 people)
Measured across a change of Census vintage. The two series share no common year, so how far apart they sit cannot be measured — this change carries an unknown amount of revision on top of four years of real change.
2016
D+38.3marginHillary ClintonDem
520,360
66.4%
Donald TrumpRep
219,793
28.0%
All other candidates
43,881
5.6%
784,034 ballots cast · 49.2 per 100 residents (2016 population estimate)
Residents1,594,664+3.4% since 2012 (+51,797 people)
2012
D+27.1marginBarack ObamaDem
471,804
62.6%
Mitt RomneyRep
267,321
35.4%
All other candidates
15,045
2.0%
754,170 ballots cast · 48.9 per 100 residents (2012 population estimate)
Residents1,542,867+4.5% since 2008 (+65,818 people)
Measured across a change of Census vintage. Where the two vintages overlap, 96% of counties agree within 0.5%, so read this figure to about half a point and no finer.
2008
D+30.1marginBarack ObamaDem
464,484
64.0%
John McCainRep
245,769
33.9%
All other candidates
15,778
2.2%
726,031 ballots cast · 49.2 per 100 residents (2008 population estimate)
Residents1,477,049+1.2% since 2004 (+17,817 people)
2004
D+29.5marginJohn KerryDem
440,862
64.0%
George W. BushRep
237,815
34.5%
All other candidates
10,283
1.5%
688,960 ballots cast · 47.2 per 100 residents (2004 population estimate)
Residents1,459,232−0.6% since 2000 (−8,016 people)
2000
D+31.2marginAl GoreDem
404,043
61.5%
George W. BushRep
198,914
30.3%
All other candidates
54,091
8.2%
657,048 ballots cast · 44.8 per 100 residents (2000 population estimate)
Residents1,467,248
“All other candidates” is every remaining line on the presidential ballot combined — minor parties, independents and write-ins where the state reports them. Shares are of all ballots cast, so the three rows sum to 100%.
Ballots cast per 100 residents is not a turnout rate. It measures votes against every resident — children and non-citizens included — not against eligible or registered voters, which is what turnout means and which this site does not have a county-level figure for. It is shown only where an election year and a population estimate share the same year and describe the same territory; where the Census figure is stated on boundaries the vote was not cast on, the card says so and the ratio is withheld.
A change in residents is not a change in voters. The percentage beside each year is the change in resident population since the previous presidential election — everyone who lives here, including children and non-citizens. It is not turnout, not registration, and not the same quantity as the change in ballots cast. Population is measured on 1 July of the election year, four months before the vote, which is the closest an annual series gets to election day.
Population figures on this page come from 3 Census series: PEP intercensal county estimates, 2000–2010; PEP vintage 2019 county estimates, 2010–2019; PEP vintage 2025 county estimates, 2020–2025. The 2012 and 2020 figures are the first in a new series, so the change shown for those years is measured across a change of Census method as well as across four years. Each of those cards says so. Two figures drawn from different vintages are not two readings of one series.
Every cycle
| Year | Margin | Lean | Swing | Rel. swing | Votes | 3rd party |
|---|---|---|---|---|---|---|
| 2024 | D+39.2 | D+42.0 | -5.8 | +0.3 | 814,832 | 3.1% |
| 2020 | D+44.9 | D+41.7 | +5.6 | +3.3 | 869,347 | 2.9% |
| 2016 | D+38.3 | D+38.4 | +12.9 | +14.7 | 784,034 | 5.6% |
| 2012 | D+27.1 | D+23.7 | -3.1 | +0.2 | 754,170 | 2.0% |
| 2008 | D+30.1 | D+23.4 | +0.9 | -9.0 | 726,031 | 2.2% |
| 2004 | D+29.5 | D+32.4 | -4.1 | -1.0 | 688,960 | 1.5% |
| 2000 | D+31.2 | D+33.4 | — | — | 657,048 | 8.2% |
Margin is over all votes cast. Lean is the two-party margin minus the nation's — the county's own position with the national environment removed. Relative swing is the change in that lean. “Votes” is the number of ballots cast in the presidential contest, not a share of eligible voters.
Counties on a similar trajectory
Sources
Presidential returns from the MIT Election Data and Science Lab (County Presidential Returns 2000–2024, CC0). Geography and FIPS codes from the US Census Bureau (public domain). Where a state reports both a total and per-mode vote counts, the reported total is used.
Resident population from the US Census Bureau, Population Estimates Program (Vintage 2025), measured as of July 1, 2025. Resident population — all residents, including children and non-citizens. Not an electorate, and not comparable to a vote total as a turnout rate. Conclusions drawn from it are ours, not the Bureau's.
Population at each election is measured on 1 July of the election year and comes from the US Census Bureau's Population Estimates Program (public domain). The 2000–2024 span is covered by several separately estimated vintages of that programme rather than by one continuous series — the ones used on this page are named under the cycle cards, and every change measured across a boundary between them is marked there. Notes about territory changing hands come from this site's own change ledger and are only as complete as it is.