Food access and public policy

Guest Lecture for FSHN 455/500

Dr. Lauren Chenarides

October 14, 2026

Should New York City forgo tax revenue to attract supermarkets?

Since 2009, New York City has offered tax incentives and zoning changes to projects that develop or renovate supermarket space in designated areas.

The program’s stated goals:

• more stores selling a healthy selection of food
• located in underserved neighborhoods
• within easy reach of residents
• a stronger economic base and tax revenue for the city

Was the policy successful?

NYCEDC, FRESH program; NYC Zoning Resolution §63-00; Elbel et al. (2015).

2 minutes.

Open on the decision. Students return to this question at the end of class and give an answer.

The four stated goals come from ZR 63-00. Three of them concern getting an accessible store built; the fourth is fiscal. Health is named first and sits in the program’s name. Let students see the list without commentary now; the gap between the goals and the evidence is the subject of a later slide.

The closing question cannot be answered as asked, which is the point. Four goals, four answers. If the room is quiet, say the line: the answer depends on which goal you hold it to.

[Sources] NYCEDC, FRESH program; NYC Zoning Resolution §63-00; Elbel et al. (2015).

Goal today

Work through a few of the following questions:

• What explains limited food access?

• Is the market failing, or is the outcome simply unequal?

• Which outcome should a policy focus on?

• How can that outcome be achieved?

• What does the evidence establish?

1 minute.

Kara has introduced scarcity, supply and demand, externalities, SNAP, and the Farm Bill. Today we use those concepts to explain the case for intervention and to interpret evidence.

The one idea I want them to leave with: limited food access does not, by itself, establish a market failure. Market failure gives an efficiency rationale for intervention. Equity can motivate policy on its own. Both require asking what the instrument actually changes.

Why are there supermarkets in some neighborhoods and not others?

Households choose where to shop based on prices, available products, travel costs, and their preferences.

Retailers open stores where expected sales cover operating costs and the cost of opening. Low incomes, dispersed customers, or high property costs can keep a location unprofitable.

Even when retail arrives, it may not be a supermarket. In the U.S., more than 90% of low-income, low-access areas that received a dollar store remained without supermarket access.

Bonanno (2012); Bitler & Haider (2011); Chenarides et al. (2021).

2 minutes.

Supply and demand together, before any policy appears. A low store count could reflect costly operations, weak purchasing power, or barriers to competition, and each explanation points to a different instrument. Stores also draw customers from outside the immediate neighborhood, so the relevant shopping market has to be investigated rather than assumed.

The Chenarides result uses store-location data from 2000 to 2017 and follows qualifying census block groups after dollar-store entry; follow-up length varies with entry year. It describes persistence of limited supermarket access. It does not establish that dollar stores caused supermarkets to stay out. Two uses here: entry by full-line supermarkets is rare in these areas, and residents are already buying food somewhere, which is what any new store has to be compared against.

[Sources] Bonanno (2012); Bitler & Haider (2011); Chenarides et al. (2021).

New York City’s response: the FRESH program

FRESH (Food Retail Expansion to Support Health) lowers the cost of owning, leasing, developing, and renovating supermarket retail space. Benefits go to the operator or the developer of a qualifying project. Since 2009, 32 projects have received tax incentives and 30 have opened.

Eligible areas come from City Planning’s assessment of grocery need, built from store concentration, walkability, car access, and families with children in poverty. Tax-eligible and zoning-eligible areas are not identical.

Note: USDA classifies low-income, low-access areas by income and distance to a supermarket. The city uses its own standard.

NYCEDC, FRESH program; NYC Department of City Planning, FRESH food stores.

1.5 minutes.

The eligibility point matters because students will otherwise assume a FRESH neighborhood is a USDA food desert and import the wrong evidence base. The original boundaries came from DCP’s 2008 “Going to Market” study; the current Supermarket Needs Index uses 2018 Retail Food Stores data plus American Community Survey measures. Zoning eligibility expanded in December 2021; tax eligibility did not.

One supported store opened in Morrisania, in the Bronx, in 2011. Morrisania already had food retailers. The policy sought to change the options available, not to create the first ones.

[Sources] NYCEDC, FRESH program; NYC Department of City Planning, FRESH food stores.

What does FRESH offer and what does it require?

The city forgoes revenue it would otherwise collect and relaxes rules it would otherwise apply, in exchange for a store meeting certain requirements. No money is appropriated.

Program details:

1. Land taxes abated and building taxes stabilized for up to 25 years, sales tax waived on construction materials, mortgage recording tax reduced from 2.8% to 0.3%.
2. Zoning permits additional floor area, requires fewer parking spaces, and removes the special permit requirement in some districts.
3. A supported store must provide at least 5,000 square feet of grocery retail, at least 30% of retail space in perishables, and at least 500 square feet of fresh produce.

NYCEDC, FRESH program; NYC Comptroller (2024).

1.5 minutes.

Keep the instruments separate from the hoped-for outcomes. This is a reduction in fixed, development, and financing costs. That drives the price discussion later.

Economists still call this a subsidy. Forgone revenue transfers value to the project just as an appropriation would, and it has an opportunity cost in the services the revenue would have funded. The support is not free. What matters mechanically is that the benefit accrues to whoever bears the development and occupancy costs, which is where the price mechanism breaks down.

NYCIDA administers the tax incentives, DCP the zoning provisions. Benefits are discretionary: NYCIDA reviews financial need and whether the incentives are required for the project to proceed, which is itself an application of evaluation before implementation. These are current program features and not a claim that every provision was identical at the 2011 Morrisania opening.

[Sources] NYCEDC, FRESH program; NYC Comptroller (2024).

Broader question: Does limited access mean the market is failing?

A market failure occurs when market decisions produce an economically inefficient allocation of resources.

An allocation is inefficient when a feasible change could make someone better off while leaving everyone else at least as well off.

Three possible causes are:

1. Market power: sellers can restrict sales and raise prices when buyers have few close alternatives.

2. Externalities: a decision creates costs or benefits for others that private incentives leave out.

3. Information problems: mistaken or incomplete information leads people to choices they would change if better informed.

Bitler & Haider (2011).

1.5 minutes.

Define efficiency through a feasible Pareto improvement, then note that applied work also compares gains with losses, where making everyone better off may require compensation.

Each candidate has a test attached, which is the next slide. A market-power argument has to consider costs and product differentiation, not just prices. A diet-related externality argument has to identify costs borne by someone else, since the household’s own health benefit is normally private. An information argument requires evidence about beliefs. Groceries are rival and excludable, so public support does not make them a public good.

These are candidate mechanisms. A store closure, a high price, or a low store count does not by itself diagnose any of them.

[Sources] Bitler & Haider (2011), p. 160.

What evidence would establish a market failure?

There is no “market-failure” index.

Source	What economists estimate
Market power	Prices, costs, and purchases under stronger competition
Externality	The effect on others and the value of that effect
Information problem	How informed choices would change household welfare

Each estimate compares the observed outcome with a specified alternative.

Bitler & Haider (2011).

1.5 minutes.

Measurement comes immediately after the causes so the diagnosis stays falsifiable. Demand and cost estimates support simulations of prices, quantities, and economic surplus under alternative competition. Externality estimates require a causal effect on third parties plus a valuation. Information analysis requires a defensible informed-choice benchmark.

Explain surplus as the gains remaining after accounting for what buyers and sellers give up. High margins can partly cover fixed costs, and income affects willingness to pay, so neither high prices nor low purchases settles the question.

[Sources] Bitler & Haider (2011).

Efficiency and equity are different objectives

A market can produce an efficient outcome that society considers inequitable.

Low household purchasing power and high retail costs can produce limited options in a market that is working as the model predicts. Nothing in that outcome requires a failure.

A community can support food-access policy on equity grounds, with an explicit priority for households facing the greatest hardship.

An equity rationale does not necessarily depend on diagnosing an inefficiency.

Bonanno (2012); Bitler & Haider (2011).

2 minutes. This is the central slide of the lecture.

Bonanno explains why limited access can arise in equilibrium under different market structures. A diagnosed efficiency problem and a distributional objective are separate rationales, and both still require weighing feasible instruments, costs, and effects.

Pause here for a question to the room: which rationale do you think the city was acting on? There is no single right answer, and the program’s own materials mix both. The point is that students stop treating “market failure” as the only vocabulary available for justifying policy.

[Sources] Bonanno (2012); Bitler & Haider (2011), p. 160.

Two ways to approach a policy argument

Normative economics asks what should be done.

• Which outcomes should society prioritize?
• How should costs and benefits be distributed?

Positive economics asks what happens and why.

• How do prices affect purchases?
• How do firms respond to costs?
• What changes after a policy is introduced?

Most policy arguments mix the two.

Bitler & Haider (2011).

1.5 minutes.

This follows directly from the previous slide: equity priorities are normative, and the market-failure diagnosis is positive. Positive statements can be false and are testable. Normative reasoning involves values and can be informed by evidence without being settled by it.

Students will need this distinction twice more, on the outcome slide next and in the closing discussion. Ask them to flag which kind of claim they are making when they report out.

[Sources] Bitler & Haider (2011), efficiency and equity discussion.

Two ways to approach a policy argument

Normative economics asks what should be done.

Choosing an objective for FRESH is normative.

Positive economics asks what happens and why.

Estimating whether FRESH reaches that objective is positive.

Most policy arguments mix the two.

Bitler & Haider (2011).

What outcome “should” the policy deliver?

Goal	What would need to change?	What would we measure?
More availability	Retailers offer additional food options	Products, hours, travel time
Lower prices	Retailers lower prices for comparable items	The cost of a consistent basket
Healthier diets	Households change purchases and consumption	Dietary intake

Choosing among them is normative.

Bitler & Haider (2011); Bonanno (2012).

1.5 minutes.

Ask the class which goal they would use to judge FRESH, and tell them they will be held to that choice in the closing discussion.

Availability, affordability, and dietary intake are related outcomes with distinct mechanisms, and one policy can improve one while leaving another unchanged. A household may value reduced travel time even if its diet does not change; that is an economic benefit worth measuring, and it is not a dietary finding.

[Sources] Bitler & Haider (2011); Bonanno (2012).

How does a policy produce an outcome?

To answer this, we must look at the underlying mechanism.

A mechanism is the sequence of decisions and responses through which a policy produces an outcome.

A policy changes an incentive, a resource, or a constraint.

Households and firms respond, and those responses determine the outcome.

To define a mechanism, one must identify who changes a decision, what changes that decision, and what has to happen next.

Bitler & Haider (2011) and Bonanno (2012).

1.5 minutes.

Define the term before asking students to use it. The goal is the desired outcome, the lever is the instrument government changes, and the mechanism explains how the lever could produce the goal. A policy usually alters one step inside a longer process.

Every step is a prediction that can fail. The next two slides apply this to the three objectives from the previous slide, which is why the objective comes first. The payoff is that “the policy worked” becomes a question about which step.

[Sources] Conceptual application of Bitler & Haider (2011) and Bonanno (2012).

The availability mechanism

1. Tax incentives and zoning changes reduce the cost of developing and occupying supermarket space.
2. An operator or developer builds a store that would otherwise be unprofitable or delayed.
3. Residents gain a shopping option carrying the required food offerings.
4. Households use that option to obtain food more easily.

Steps 1 and 3 follow from the program’s terms. Steps 2 and 4 depend on decisions the city does not control: whether the support changes the investment, and whether households shop there.

Bonanno (2012); NYCEDC, FRESH program.

1.5 minutes.

First full application of mechanism, and the shortest of the three chains, which is why it comes first.

Step 2 is additionality, and it is what NYCIDA’s inducement review tries to address before approval. Step 4 depends on store size, products, hours, travel routes, and the retail environment that already exists, which connects back to the Chenarides point from the problem slide.

This is a proposed causal sequence, not a finding that FRESH produced every step.

[Sources] Bonanno (2012); program application: NYCEDC, FRESH program.

Prices and diets require more than availability

Prices. The program lowers the cost of having the store. Shelf prices are set against the cost of selling the next item, which the program does not change. Entry can still lower prices by adding competition, so a price claim needs evidence on prices at the new store and its competitors.

Diets. Steps 1 through 4, plus three more: households shop at the new store, buy differently, and eat differently. If they buy a similar basket at a closer store, access improves and intake does not.

Three objectives, three chains. Evidence on one says little about the others.

Bonanno (2012); Elbel et al. (2015).

2 minutes.

Distinguish fixed costs from the incremental cost of selling another unit. FRESH lowers the cost of having the store, so its connection to shelf prices runs only through competition. Do not overstate it: entry can change competition, and contracts and pricing strategies matter. The claim is that the channel has to be specified and compared.

For diets, ask which step the class expects to be most uncertain. Transportation, prices, cooking time, preferences, and information all enter. The policy changes the retail environment; household decisions remain the households’.

The closing line is the one students should carry into the evidence section.

[Sources] Bonanno (2012); Elbel et al. (2015).

Different research designs to evaluate policy outcomes

Design	What it compares	What it supports
Randomized experiment	Assigned treatment and control groups	A causal effect under the trial’s conditions
Difference-in-differences	Change in treated and comparison places	A causal effect if untreated trends would have matched
Modeled counterfactual	Observed outcome and a simulated alternative	An effect conditional on the model being right

The examples that follow use one of these three designs.

Bitler & Haider (2011); Allcott et al. (2019); Olsho et al. (2016).

1.5 minutes.

Mention that evaluation also happens before implementation, through modeled predictions, cost estimates, and elasticities tested for sensitivity to uncertain assumptions, and that NYCIDA’s inducement review is a version of this.

Difference-in-differences rests on an assumption about the untreated trend and the absence of other differential changes. A structural counterfactual rests on the model. A randomized design identifies the effect of the offer under the conditions of the trial and does not automatically extend beyond them.

Students should be able to label each of the four studies that follow. Name the design before accepting the finding.

[Sources] Framework: Bitler & Haider (2011); research examples: Allcott et al. (2019); Olsho et al. (2016).

Testing the diet mechanism in Morrisania

Elbel and colleagues surveyed families with children ages 3–10 in Morrisania and in Highbridge, a comparison neighborhood, before and after the FRESH-supported store opened.

The comparison asks whether the food households kept at home and children’s dietary intake changed more in Morrisania than in Highbridge over the same period.

Design: difference-in-differences. The store was not randomly placed.

Elbel et al. (2015).

1 minute.

Explain difference-in-differences in words: You want to know what a store opening did to a neighborhood, but you can’t observe that neighborhood without the store. So you find a second neighborhood that didn’t get one and use its change over the same period as a stand-in for what would have happened.

Morrisania and Highbridge: measure both before, measure both after. Morrisania’s change includes the store plus everything else going on in the Bronx that year. Highbridge’s change is everything else alone. Subtract the second from the first and what’s left is attributed to the store.

One difference is before versus after, the other is treated versus comparison. You take the difference of the differences.

The assumption is that Morrisania would have followed Highbridge’s path absent the store. If something else hit one neighborhood and not the other — a transit change, a layoff, another food program — that goes into the estimate too.

Highbridge estimates the counterfactual: the change Morrisania would have experienced without the opening. The design relies on comparable trends and on no other differential changes driving the result.

Elbel calls the first outcome “household food availability,” measured as a home food inventory. Say “food in the home” throughout. Students will otherwise hear the city’s availability goal, which is about stores, and the two sit three steps apart.

Surveys ran at baseline, about five weeks after opening, and about one year after opening; dietary recalls covered baseline and the first follow-up. Do not imply intake was observed at every wave.

Both outcomes sit at steps 3 and 4 of the diet chain, in a neighborhood that already had food retailers.

[Sources] Elbel et al. (2015).

The opening produced little measured dietary change

The study found no appreciable improvements attributable to the opening in the food households kept at home or in children’s dietary intake.

The study did not measure travel time, convenience, choice, or prices as outcomes.

Elbel et al. (2015).

1.5 minutes.

Return to the diet chain and mark where it stopped. The store opened, so the investment occurred. The household steps found little support in this setting and follow-up period.

Two inferences to block. This is a study of one store, one neighborhood, one comparison, and about a year, so it does not speak to supermarket incentives in general. And the outcomes that would capture convenience and choice went unmeasured, so the finding leaves the access rationale open.

Students who chose healthier diets as their objective should notice the problem now rather than during the discussion. Hold the broader point about stated goals until the recap slide, after Rummo.

[Sources] Elbel et al. (2015).

A later FRESH study measured a different outcome

	Elbel et al. (2015)	Rummo et al. (2022)
Stores	One, in Morrisania	Eight new or renovated
Population	Families with children 3–10	Public-school students
Outcome	Food in the home, children’s intake	Obesity risk
Follow-up	About one year	3–12 months
Design	Difference-in-differences	Difference-in-differences

Living within half a mile of a supported store was associated with about a 1-percentage-point reduction in obesity risk relative to comparison students.

The study did not observe whether households shopped at these stores or what they bought.

Rummo, Sze & Elbel (2022).

1.5 minutes.

Rummo compares students in FRESH-eligible areas before and after nearby stores opened. Both studies are observational, so other changes could affect either comparison, and neither involves random assignment.

The table is the point. These are different stores, populations, outcomes, and follow-up windows, and they are consistent with each other.

On the last line: the diet chain runs store, shop, purchase, consume, with body weight beyond it. Rummo observes the first link and the last outcome. That is a reason to keep investigating the pathway, and the estimate still stands as an association.

Second pause point if time allows: ask what would have to be measured to connect the two studies.

[Sources] Rummo, Sze & Elbel (2022).

What has not been measured

Both FRESH evaluations measure diets and body weight. The program’s own objectives sit upstream of both.

Stated goal	Evidence
A store in an underserved area	Openings counted, effects not estimated
Within easy reach of residents	Perceived access improved in Philadelphia; travel time not measured
A healthy selection of food	Store requirements, no purchase data
Economic base and tax revenue	Jobs and investment reported, not evaluated

No evidence here speaks to the access rationale, in either direction.

NYC Zoning Resolution §63-00; NYCEDC, FRESH program; NYC Comptroller (2024).

1.5 minutes.

Three of the city’s four stated purposes concern getting an accessible store built; the fourth is fiscal. Health is claimed, first in the purposes list and in the program’s name, and no step of the instrument reaches it.

So diet is a fair thing to evaluate. The city asserted the benefit. A program judged only on what its instrument mechanically controls always succeeds, and FRESH would then be judged on whether stores opened, which is close to tautological for a program that lowers the cost of opening stores.

The harder point is the gap running the other way. Researchers measured the distal health outcomes and skipped the proximate access ones. The access rationale has the shortest and most plausible mechanism, and it remains untested.

The bold line is what students need for the discussion. A student arguing for FRESH on access or equity grounds can say correctly that the relevant evidence was never collected.

The jobs and tax-base row is the only place the fourth objective appears in the lecture. Students reach for “but it created jobs” in the discussion; NYCEDC reports 1,400 jobs created and $177 million in private capital, and the Comptroller report examines job quality. Reported figures are not an evaluation.

[Sources] NYC Zoning Resolution §63-00; NYCEDC, FRESH program; NYC Comptroller (2024).

Did residents use the new store?

A Philadelphia supermarket opened in 2009 in a neighborhood classified as a food desert, supported by the Pennsylvania Fresh Food Financing Initiative.

This study asked whether households adopted the new store.Researchers surveyed residents there and in a comparison neighborhood, before and after.

Residents’ perceptions of food accessibility improved. Reported fruit and vegetable intake and BMI did not change.

Design: difference-in-differences.

Cummins, Flint & Matthews (2014).

1.5 minutes.

Design: difference-in-differences, same as Elbel and Rummo, and a different program in a different city. Follow-up was about six months.

Step 4 of the availability mechanism is the one the city does not control, and this is where it gets tested. The result splits the chain cleanly: the store registered with residents, and their reported eating did not change.

Two cautions. Perception is self-reported and so is intake. And a Health Affairs commentary noted that Philadelphia was running other food-access interventions during the study period, which may have affected the comparison neighborhood.

How much can the retail environment explain?

Allcott and colleagues asked why lower-income households purchase less healthy groceries, on average.

They examined household purchases around supermarket entry and around moves to different neighborhoods.

They used an economic model to ask how the purchasing gap would change if lower-income households faced the products and prices available to higher-income households.

Design: observed responses plus a modeled counterfactual.

Allcott et al. (2019).

1.5 minutes.

Kara’s economics lecture introduces this paper, so use the time on the research question and on the distinction between observed evidence and a simulated policy.

Entry and relocation reveal how households respond to a changed retail environment. The model combines estimated demand with product availability and prices. The outcome is the nutritional quality of grocery purchases, not food insecurity, dietary intake, or health.

This slide answers the question Elbel and Rummo leave open: why would the store-to-diet link be weak?

[Sources] Allcott et al. (2019).

Equal shopping options closed 9% of the nutritional gap

Giving lower-income households the same products and prices as higher-income households reduced the nutritional purchasing gap by about 9% in the model.

Policies that change the retail environment have limited scope to close nutritional gaps.

Household resources, information, time, and preferences explain more of the observed variation.

This gives a reason to expect a small dietary response to a single store opening.

Allcott et al. (2019).

1.5 minutes.

Explain the denominator: 9% of the measured difference in nutritional purchasing quality between income groups. This is a modeled equalization, not a 9% improvement in any household’s diet and not the effect of a particular program.

The model attributes much of the gap to differences in demand. Do not read that category as blame, and do not claim the paper separately identifies the causal contribution of each household constraint listed.

Transition: if the retail environment is a weak lever for diets, what is a stronger one?

[Sources] Allcott et al. (2019).

A differnt mechanism: change the price households face

The USDA Healthy Incentives Pilot offered randomly selected SNAP households in Hampden County, Massachusetts, a 30% rebate on eligible fruit and vegetable purchases.

Rebate → lower effective produce price → more produce purchased → more produce consumed.

Design: randomized experiment. Assignment is unrelated to households’ initial food preferences, on average.

Olsho et al. (2016).

1.5 minutes.

Same vocabulary as FRESH: goal, lever, mechanism. The chain is shorter by two steps, because the instrument acts on the price the household faces.

The pilot assigned 7,500 SNAP households to receive the incentive; dietary intake was studied in a survey sample. A rebate returns benefits after an eligible purchase, so it operates differently from a lower posted shelf price.

[Sources] Olsho et al. (2016).

The rebate increased fruit and vegetable intake

Adults in households offered the rebate consumed about 0.24 additional cup-equivalents of targeted fruits and vegetables per day, compared with the control group. The 95% confidence interval was 0.13–0.34.

A shorter chain produced a measurable change in consumption.

A randomized estimate of the offer’s effect during a pilot still leaves open long-term health, costs, and what happens at national scale.

Olsho et al. (2016).

1 minute.

The estimate compares households assigned to the incentive with those not assigned, rather than comparing self-selected users with nonusers.

Ask directly: why is the mechanism connecting this lever to produce intake more credible than the one connecting a development incentive to intake? The answer is the number of household decisions standing between the instrument and the outcome.

Also note what this instrument does not do. It adds no store, changes no travel time, and reaches only SNAP participants.

[Sources] Olsho et al. (2016).

What the five studies establish

Study	What was studied	Outcome tested	Design	Finding
Elbel (2015)	One store opening	Food in the home, child intake	Diff-in-diff	Little measured change
Cummins (2014)	One store opening	Perceived access, intake, BMI	Diff-in-diff	Perceptions improved, intake did not
Rummo (2022)	Eight store openings	Childhood obesity risk	Diff-in-diff	Associated with ~1 pp lower risk
Allcott (2019)	Equalized products, prices	Nutritional purchase quality	Modeled	Gap narrows ~9%
Olsho (2016)	30% produce rebate	Fruit and vegetable intake	Randomized	+0.24 cup-equivalents/day

1.5 minutes.

Read across by outcome, population, design, and mechanism rather than ranking the studies.

Cummins sits next to Elbel because the two are the closest thing to a replication in the table: two store openings, two cities, two programs, the same design, and consistent dietary results. Cummins adds the perception measure that Elbel lacks.

None of these findings, on its own, establishes that an intervention corrected a market failure. That requires the welfare and cost assessment on the next slide.

The untested-objectives point was made two slides ago; here, keep the focus on what the designs support.

[Sources] Elbel et al. (2015); Cummins, Flint & Matthews (2014); Rummo, Sze & Elbel (2022); Allcott et al. (2019); Olsho et al. (2016).

Would the benefits justify the costs?

The preceding slides ask whether the policy changed an outcome.

A cost-benefit analysis asks whether the gain exceeded what the city gave up.

The cost is revenue forgone plus the resources the project uses. The Comptroller puts the tax program at $3–4 million per year.

Nearby households, property owners, the supported retailer, and competing stores are affected differently. An equity rationale depends on which of them gains.

This calculation has not been done for FRESH.

Bitler & Haider (2011); NYC Comptroller (2024).

1 minute.

Avoid double-counting transfers as net social losses. A tax concession has a fiscal cost, and resources devoted to the project have opportunity costs; these are related but distinct accounting questions. Benefits can include consumer surplus, travel savings, and any identified external benefits. Competitor losses are partly transfers, though changes in market structure and access can affect total welfare.

Weighing the estimates against a community’s priorities is a normative judgment, which returns to the positive and normative slide.

[Sources] Bitler & Haider (2011); program context: NYC Comptroller (2024).

Current research: SNAP stocking standards

FRESH changes who opens a store.

SNAP and produce incentives change what a household can afford.

Another policy lever sets the terms a store must meet to accept SNAP, which is central to measuring food access.

Background

• To accept SNAP, a retailer must carry a minimum assortment of staple foods.
• A rule effective in 2018 raised both the breadth of qualifying varieties per category and the depth of stocking units required.
• Stated intent was that changes in the standard would drive healthier purchases

Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016); USDA FNS, stocking requirements.

2 minutes.

Open by placing the third lever alongside the two the lecture has covered. A household can only redeem benefits where a store is authorized, so the definition of a qualifying store is itself an access policy. It also runs opposite to FRESH: a stocking standard can remove stores from the network a supermarket incentive is trying to add to.

Intended pathway standard → retailer inventory → local availability → purchases and health Participation pathway standard → compliance cost → SNAP authorization → the authorized-store network

Two pathways, as in the FRESH case: the one the rule intends and the one the retailer’s own decision creates. The policy rationale is that earlier minimum standards allowed participation with limited nutritious variety, and that the value of authorization can be used to push neighborhood retailers to stock more.

The participation pathway is a threshold choice. Compliance costs rise by different amounts across retailers, and a store drops out when the cost of staying authorized exceeds what authorization is worth to it. Keep the notation off the slide for this audience; the sentence carries it.

Connect to the Chenarides dollar-store result from early in the lecture. The stores facing the largest compliance cost are small-format stores, which in low-access areas are often the only retail present.

[Sources] Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016); USDA FNS, stocking requirements.

The participation mechanism

Compliance costs rise by different amounts across retailers, and a store drops out when the cost of staying authorized exceeds what authorization is worth to it.

standard → compliance cost → SNAP authorization → the authorized-store network

A retailer stays authorized while the value of authorization covers the cost of meeting the standard.

Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016); USDA FNS, stocking requirements.

2 minutes.

Open by placing the third lever alongside the two the lecture has covered. A household can only redeem benefits where a store is authorized, so the definition of a qualifying store is itself an access policy. It also runs opposite to FRESH: a stocking standard can remove stores from the network a supermarket incentive is trying to add to.

Intended pathway standard → retailer inventory → local availability → purchases and health Participation pathway standard → compliance cost → SNAP authorization → the authorized-store network

Two pathways, as in the FRESH case: the one the rule intends and the one the retailer’s own decision creates. The policy rationale is that earlier minimum standards allowed participation with limited nutritious variety, and that the value of authorization can be used to push neighborhood retailers to stock more.

The participation pathway is a threshold choice. Compliance costs rise by different amounts across retailers, and a store drops out when the cost of staying authorized exceeds what authorization is worth to it. Keep the notation off the slide for this audience; the sentence carries it.

Connect to the Chenarides dollar-store result from early in the lecture. The stores facing the largest compliance cost are small-format stores, which in low-access areas are often the only retail present.

[Sources] Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016); USDA FNS, stocking requirements.

What the data can and cannot show

Using store-year authorization and location records, I ask:

1. How did the revised standards affect continued SNAP authorization?
2. Which types of retailers were more likely to lose it?
3. How did the number, type, and location of authorized stores change?
4. Were losses concentrated where more households participate in SNAP?

Store inventory is not observed. Neither are compliance costs, household purchases, or diet.

Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016).

1.5 minutes.

Question 4 is the one that matters for this course. If authorization losses are larger where SNAP participation is higher, the authorized-store network contracts most in the communities the program is meant to serve. That is a targeting problem relative to the policy’s own food-access goal, and it does not require observing a single shelf.

The last line is the lecture’s lesson applied to my own work. The intended pathway runs through inventory, and inventory is unobserved, so this study cannot speak to whether remaining stores stock more nutritious food, whether purchases changed, or whether diets changed. Same boundary that limits Rummo, acknowledged in advance rather than after the fact.

Two design problems worth naming if students ask: separating loss of authorization from ordinary store closure, and choosing the geography that defines a community.

USDA published a further final rule on May 8, 2026, with compliance beginning November 4, 2026, three weeks after this class. The same question arrives again with a new standard.

[Sources] Enhancing Retailer Standards in SNAP, 81 FR 90675 (2016); USDA FNS, stocking requirements.

Epilogue: New York plans to run its own stores

NYC announced five city-run grocery stores, one per borough, on city-owned land, by the end of the mayor’s term in 2029. The program would waive rent and property taxes and sell near wholesale cost.

Reported estimates put staples up to 25% cheaper.

The first store is scheduled to open in late 2027 in Hunts Point (Bronx) located at the former Spofford Juvenile Detention Facility site.

	FRESH	City-run stores
Objective	Access	Affordability
Who operates	A private firm	The city
Price channel	Competition only	Direct
Cost to the city	Revenue forgone	Capital plus operating losses

Fox 5 NY (2026); News 12 (2026).

2 minutes.

No store has opened, so this is evaluation before implementation, which closes the loop on the research-designs slide.

Three slides pay off here. The price mechanism: FRESH lowers the cost of having the store, so prices move only through competition, while a city store selling near wholesale changes the shelf price directly. Additionality at step 2 of the availability mechanism disappears, because the city is the builder. And the retailer participation condition from the problem slide now binds on the city: a store whose sales do not cover operating costs is a recurring expense rather than an exit.

The objective moved, so the evidence that would test it moves too. Judging this program on diets repeats the mistake the lecture has been tracing. The outcome is the cost of a consistent basket, at these stores and at their competitors.

Reported figures as of spring 2026, and outlets differ on the first location and the budget. Check current numbers before class; the capital figure and the opening date are the two most likely to have moved.

[Sources] Fox 5 NY (2026); News 12 (2026).

How does this framework compare to that of nutritional studies?

Today’s framework treats food access as a question about store location, inventory, prices, and household decisions.

If you had to choose a policy to improve food access in a specific community, which would you pick, what would you need to know about that community, and how would you evaluate its outcomes?

Fox 5 NY (2026); News 12 (2026).

10 minutes: 4 in pairs or small groups, 6 for report-out, discussion, and questions.

This returns to the question on the opening slide, now with a different instrument, which tests whether the framework transferred.

Press on specifics. On objective, affordability is the stated one; push anyone who drifts back to diets to say what the mechanism would be. For efficiency, require a named failure, and note that market power is the only one that would justify public operation rather than a transfer. Equity stands on its own.

On mechanism, the failure points are whether households shop there, whether the stores stock what people want to buy and can cook, whether prices hold once operating costs appear, and what happens to nearby private stores.

On measurement, five stores and no random assignment make this hard. Ask what the comparison is, and whether a basket-price series at these stores and their competitors beats any household survey they could field.

Raise the alternative if nobody does: a transfer leaves the choice of store with the household, and the opportunity cost of the capital is the comparison. Let them argue it.

The SNAP rule is a second example if the room stalls. It takes effect three weeks from today, and the same three questions apply.

Close on the through-line: economists diagnose the problem, specify how an instrument could change decisions, and evaluate those effects against a stated social objective. Whether to act remains a choice about what we value.

[Sources] Fox 5 NY (2026); News 12 (2026).

References

Bitler, M., & Haider, S. J. (2011). An economic view of food deserts in the United States. Journal of Policy Analysis and Management, 30(1), 153–176. DOI.

Bonanno, A. (2012). Food deserts: Demand, supply, and economic theory. Choices, 27(3), 1–4. Article.

New York City Department of City Planning. (2021). FRESH update: City Planning Commission report N 210380 ZRY. Report.

New York City Economic Development Corporation. (n.d.). Food Retail Expansion to Support Health (FRESH). Program information. Accessed October 10, 2026.

Office of the New York City Comptroller. (2024). Good jobs and the New York City FRESH program: Evaluation and recommendations. Report.

Fox 5 New York. (2026, April). Mamdani proposes city-owned grocery store in East Harlem. Report.

News 12. (2026, April 14). Mamdani announces plans for 5 city-run grocery stores. Report.

U.S. Department of Agriculture, Food and Nutrition Service. (2016). Enhancing retailer standards in the Supplemental Nutrition Assistance Program (SNAP). 81 FR 90675. Final rule.

U.S. Department of Agriculture, Food and Nutrition Service. (n.d.). SNAP retailer eligibility and stocking requirements. Program information. Accessed October 10, 2026.

The Bitler reading filename uses 2010; cite the journal publication year, 2011.

FRESH eligibility derives from DCP’s 2008 “Going to Market” study and the current Supermarket Needs Index. The four stated purposes are in ZR 63-00, reproduced in the 2021 CPC report. Program figures are from NYCEDC as of the access date.

City-store figures are spring 2026 press reporting and should be refreshed from the Mayor’s Office or EDC before the lecture.

References

Allcott, H., Diamond, R., Dubé, J.-P., Handbury, J., Rahkovsky, I., & Schnell, M. (2019). Food deserts and the causes of nutritional inequality. Quarterly Journal of Economics, 134(4), 1793–1844. DOI.

Chenarides, L., Cho, C., Nayga, R. M., Jr., & Thomsen, M. R. (2021). Dollar stores and food deserts. Applied Geography, 134, 102497. DOI.

Cummins, S., Flint, E., & Matthews, S. A. (2014). New neighborhood grocery store increased awareness of food access but did not alter dietary habits or obesity. Health Affairs, 33(2), 283–291. DOI.

Elbel, B., Moran, A., Dixon, L. B., Kiszko, K., Cantor, J., Abrams, C., & Mijanovich, T. (2015). Assessment of a government-subsidized supermarket in a high-need area on household food availability and children’s dietary intakes. Public Health Nutrition, 18(15), 2881–2890. DOI.

Olsho, L. E. W., Klerman, J. A., Wilde, P. E., & Bartlett, S. (2016). Financial incentives increase fruit and vegetable intake among Supplemental Nutrition Assistance Program participants: A randomized controlled trial of the USDA Healthy Incentives Pilot. American Journal of Clinical Nutrition, 104(2), 423–435. DOI.

Rummo, P. E., Sze, J., & Elbel, B. (2022). Association between a policy to subsidize supermarkets in underserved areas and childhood obesity risk. JAMA Pediatrics, 176(7), 646–653. DOI.

Chenarides, L., & C. Ambrozek (in progress). SNAP stocking requirements and the authorized-retailer network. Colorado State University.

Kara’s Week 8 economics lecture supplies course context; its Farm Bill figures and legislative timeline are not repeated here.

Elbel et al. evaluate the Bronx case. Cummins et al. evaluate a Philadelphia store opened under the Pennsylvania Fresh Food Financing Initiative, a separate program, and it is presented as a parallel case rather than as evidence about the Bronx opening.