Smarter charging infrastructure can produce big welfare gains for EV drivers—but won’t substantially increase EV adoption in the short run

Electric vehicles (EVs) are becoming a larger share of the global auto market, growing from less than 1% of new vehicle sales in 2016 to nearly 20% worldwide by 2024. Yet charging infrastructure has not kept pace. A single Level 3 (DC fast) charging station can cost as much as $200,000 to install, and a persistent fear of running out of charge—"range anxiety”—remains a significant deterrent for prospective EV buyers. Congress allocated $5 billion for fast charging infrastructure under the 2021 Bipartisan Infrastructure Law (BIL), but implementation has been painfully slow.

Because access to EV charging infrastructure is critical for encouraging adoption, a core problem many states face is how to maximize their investment by knowing where to put new chargers. Using detailed data from Connecticut on vehicle registrations, travel patterns, and charging station locations, new research constructs a model of how EV drivers plan trips and where they stop to charge, then identifies charging locations that would maximize total driver welfare.

What we learned

• In Connecticut, current charger deployment is misallocated. Connecticut's existing fast chargers are concentrated along the shoreline and the I-91 corridor, even though many inland areas face much more charging congestion. Regions like Naugatuck Valley South and the New Britain area have total wait times across all charging stalls exceeding 18,000 hours per day, while Stamford and Greenwich are at zero. The problem is not just the number of chargers, but where they are.

• Optimal placement would generate enormous welfare gains at low cost. Adding just 100 L3 chargers in the locations the model identifies as highest value would reduce aggregate waiting time by more than half and generate $205 million in annual consumer welfare — at an annualized installation cost of only $1 million. This 205-to-1 benefit-cost ratio reflects how poorly the spatial match is between current supply and demand.
• Charging infrastructure does relatively little to grow EV market share. Despite the large welfare gain for existing EV drivers, the modeled expansion of the charging network barely moves EV market share — from 1.868% to 1.882%. The welfare gains are almost entirely on the intensive margin: improving the experience for people who already own EVs, not converting gasoline-car buyers. For context, this is a short-run change, and over the longer run, adding charging infrastructure and reducing “range anxiety” would increase sales of EVs.
• For influencing consumer behavior on EVs, availability of Level 3 chargers is critical. L2 chargers use AC power and are standard for home and workplace use, while L3 chargers use DC power and allow drivers to add hundreds of miles of driving time in less than an hour. When estimating how charging infrastructure affects vehicle purchasing decisions, the research finds much larger effects from L3 charger density. The ability to recharge quickly is what shapes consumer behavior at the margin.

Policy Takeaways

This analysis shows that states can generate big benefits for EV drivers by being smarter about what chargers they install and where they place them. However, optimizing charger placement and density won’t by itself supercharge adoption of EVs, and government needs to explore other policy instruments if EV adoption is the primary goal.

• To help EV drivers, use data-driven optimization to choose where to locate chargers — don't just build where it's politically easy or where demand is already high. The welfare gains to current EV drivers identified in this paper depend on placing chargers in the right locations, not simply building more. In Connecticut, existing deployment has favored wealthy coastal areas already well-served by charging. Federal and state agencies administering funds allocated towards charging infrastructure should require applicants to demonstrate data-driven site selection, ideally using spatial models of the road network and travel demand.
• Don't expect charging infrastructure alone to dramatically accelerate EV adoption. This research finds that even a large increase in L3 charger availability moves EV market share by barely a rounding error. While over time, there would be an effect of better charging infrastructure, policymakers counting on public charging buildouts to drive mass EV adoption should temper those expectations. Purchase incentives, vehicle cost reductions, and home charging access are likely more powerful levers for expanding the EV market than public fast chargers.
• If the goal is to influence vehicle purchase decisions, prioritize Level 3 chargers over Level 2 chargers. The demand estimates show the effect of L3 charger density on EV purchase decisions is substantially larger than for L2 chargers. L3 chargers are much more expensive and are more difficult to find suitable locations for, though, so this expense would have to be weighed against the added benefits.
• Fix the implementation bottlenecks that produced only 7 deployed chargers in two years. The BIL allocated $5 billion for EV charging, yet permitting, site requirements, utility interconnection delays, and federal Buy America provisions combined to produce near-zero deployment in the program's first two years. Streamlining these processes is a prerequisite to realizing the welfare gains this paper identifies.
• Target underserved inland areas where charging congestion is worst and marginal welfare gains are highest. The model identifies that the greatest unmet need is in inland communities, not already-served urban and shoreline corridors. Equity and efficiency point in the same direction here: directing chargers to high-congestion, lower-income areas away from the coast would both reduce the worst wait times and produce the largest gains in consumer welfare per dollar spent.

Data and Methodology

The authors combine four primary data sources, all covering Connecticut:

• Vehicle registrations at the zip-code-by-month level (2018–2022), including every model, trim, and drive type
• Cell phone movement data from Advan Neighborhood Patterns (2022), used to construct an origin-destination matrix of trips by Census Block Group
• The location and specifications of every charging station in the state through 2024, from the National Renewable Energy Laboratory
• Gasoline prices from the U.S. Energy Information Administration

Data are aggregated to Connecticut's 25 Public Use Microdata Areas (PUMAs). EV driver routing decisions are modeled using the Electric Vehicle Routing Problem (EVRP), solved as a Mixed Integer Linear Program (MILP) using the Gurobi solver. Optimal charger placement is found using a greedy algorithm that iteratively assigns the next charger to the location yielding the greatest marginal welfare improvement. Vehicle demand is estimated using a random-coefficients logit model (Berry, Levinsohn, and Pakes 1995), estimated separately for each of Connecticut's eight counties, with standard BLP instruments for price.

Related Work

• Allen, Treb and Costas Arkolakis, “Trade and the Topography of the Spatial Economy,” The Quarterly Journal of Economics, 2014, 129 (3), 1085–1140.
• Allen, Treb and Costas Arkolakis, “The welfare effects of transportation infrastructure improvements,” The Review of Economic Studies, 2022, 89 (6), 2911–2957.
• Desaulniers, Guy, Fausto Errico, Stefan Irnich, and Michael Schneider, “Exact algorithms for electric vehicle routing problems with time windows,” Operations Research, 2016, 64 (6), 1388–1405.
• Springel, Katalin, “Network externality and subsidy structure in two-sided markets: Evidence from electric vehicle incentives,” American Economic Journal: Economic Policy, 2021, 13 (4), 393–432.