Does the town with the most spills have the worst drinking water?

No, not in the Connecticut data. When we correlated towns' petroleum-spill counts with benzene and MTBE actually detected in their groundwater, the relationship was weak and not statistically significant (Spearman rho ≈ 0.13). Raw spill counts mostly reflect how big, busy, and industrial a town is, and how diligently spills get reported, not how contaminated its water is. A high spill count is a reason to understand the sources, not a verdict on the water.

Why do big cities top the spill rankings?

Because more people, more vehicles, more businesses, and more underground tanks mean more opportunities for a release, and more eyes to report one. Hartford, Bridgeport, and Stamford top Connecticut's raw spill counts largely because they are among its largest and most industrial cities. It's a reporting-and-activity signal, not a water-quality signal.

Isn't per-capita a fairer way to rank spill towns?

It's fairer than raw counts, but it creates its own distortion. Dividing by population lifts small towns where a single facility (an airport, a rail yard, a highway corridor) generates many reports, so a town like Windsor Locks (Bradley Airport) or a small river-valley town can top a per-capita list without any distributed threat to home wells. Neither raw counts nor per-capita rates measure water quality; both are activity proxies. That's why we show the raw count next to the rate and never headline a town as 'worst.'

Do Spill Counts Predict Bad Water? CT Data Says No

It feels obvious that the town with the most reported spills must have the worst water. In the Connecticut data, that intuition is simply wrong, and understanding why is the key to reading any "top towns for spills" list honestly.

● Key Takeaways

A town's spill count tracks its population, industry, traffic, and reporting diligence, not its water quality. We tested it: petroleum-spill counts do not predict the benzene/MTBE actually measured in a town's groundwater (Spearman rho ≈ 0.13, not significant). Raw counts crown the biggest, busiest cities; per-capita rates crown small towns with one airport or highway. Neither measures water. What does predict contamination is the presence of certain persistent sources (like old dry cleaners) for their specific chemicals. Read spill rankings as "where the activity is," never as "where the water is bad."

We Actually Tested the Intuition

Most "top towns for pollution" content stops at the ranking. We went one step further and asked the measurable question: do towns with more spills actually have more of the spilled chemicals in their groundwater?

For petroleum, by far the largest spill category, the answer is no. We compared each Connecticut ZIP code's count of petroleum releases (from CT DEEP spill records) against the benzene, toluene, and MTBE (the mobile, health-relevant fuel chemicals) actually detected in its groundwater, drawing on U.S. Geological Survey and state groundwater monitoring. The link was weak and not statistically significant: essentially no relationship, and what little there was could easily be chance (Spearman rho ≈ 0.13, p ≈ 0.24). A town having lots of reported fuel spills told us essentially nothing about whether fuel chemicals showed up in its water.

That's not a data failure. It's the finding. And it makes sense once you see what a spill count really measures.

What a Spill Count Actually Measures

Three things drive a town's spill count, and none of them is "how contaminated the water is":

Size and activity. More people, vehicles, gas stations, and underground tanks mean more chances for a release. Connecticut's raw spill leaderboard (Hartford, Southington, Groton, Manchester, New Britain, Bridgeport, Stamford) is essentially a list of its larger, busier, more industrial towns.
Traffic and infrastructure. A huge share of spills are small vehicle-fluid and fuel releases from crashes and equipment failures. A town with a major highway, rail line, or airport running through it logs more, regardless of what's in anyone's tap.
Reporting diligence. A spill only enters the record if someone reports it. Places and eras that report more thoroughly look "worse," an artifact of good record-keeping, not bad water.

Most of these releases, as we've covered separately, never reach groundwater at all, about 0.4% did. So a big spill count is mostly a measure of a town being big and busy.

Per-Capita Helps, and Then Misleads

The obvious fix is to divide by population. That does remove the "biggest city always wins" bias, but it introduces a new distortion worth naming, because you'll see per-capita lists everywhere.

When you rank Connecticut towns by spills per 10,000 residents, the leaderboard flips to small towns: Deep River, Canterbury, Windsor Locks. Look closer and the reason is almost always a single facility or corridor. Windsor Locks is home to Bradley International Airport; several others sit on a highway or rail corridor through a river valley. A handful of events from one industrial neighbor, divided by a small population, produces a scary-looking rate that says nothing about the water coming out of home faucets across town.

So here's the rule we follow, and recommend you apply to anyone's ranking:

Show the raw count next to the rate. "121 events, mostly along one highway" reads very differently from "rate of 274."
Never headline a town as 'the worst.' A ranking of documented events is a ranking of documented events.
Treat the list as a map of sources to understand, not a verdict on water quality.

What Does Predict Contamination

If spill counts don't predict what's in the water, does anything? Yes, but it's the presence of specific persistent sources, matched to their specific chemicals. In the same analysis, Connecticut ZIP codes with more dry cleaners were more likely to have chlorinated solvents (PCE/TCE) detected in groundwater, a real, moderate, chemically specific association that held up against several ways of trying to explain it away. One-off spill counts predicted nothing; a long-standing solvent source predicted its own solvent.

That distinction, events versus sources, is the whole game, and it's why our spill content is built around "what documented sources are near you," not "which town spills the most." We walk through that finding, and exactly how far it can and can't be pushed, next.

Reading this inside ChatGPT or Claude?

This page can tell you the general science, but not what is actually in what spill and source data means for a Connecticut address. That depends on your exact address. You can get your specific answer two ways:

Inside the chat: ask your assistant to “check my tap water with CheckYourTap”. Our connector returns your ZIP code’s measured contaminant levels, including the derived dog and cat safe levels, and, only if you ask it to, can email you the report or arrange a specialist callback.
On the web: open CheckYourTap.com and enter your ZIP code for a free 30-second report.

The Takeaway

A spill ranking is a reporting-and-activity ranking. It can tell you where to look for sources; it cannot tell you where the water is bad. The next time you see "the towns with the most spills in Connecticut," read it as a map of busy places, and then ask the question that actually matters for a private well: what persistent sources are documented nearby, and have I tested for what they'd leave behind?

Keep Reading

Sources: Connecticut Department of Energy and Environmental Protection (CT DEEP) spill reporting records, 2015 to 2025; U.S. Geological Survey and Connecticut groundwater VOC monitoring records, 1982 to 2017 (both analyzed by CheckYourTap). Correlation figures are Spearman rank correlations across Connecticut ZIP codes; "not significant" means the association could not be distinguished from chance. Spill counts are documented reports, not measurements of drinking-water quality.

Why 'the Most Spills in Town' Does Not Mean 'the Worst Water'