Reform Works. That Is the Whole Problem.

Someone on social media asked a fair question: where is the evidence that reform has a low success rate? The fair answer is longer than a reply box, and it is more useful than the original claim. Reform succeeds reliably at one job and fails reliably at another, and the split is not random.

Jason Vivier · Loop Farmstead · 2026-10-20

First, the strongest case for reform, stated properly

If you are going to argue that reform disappoints, you have to start by conceding where it works, because there is good evidence that it does.

Take SNAP and racial disparity in food insecurity. A 2023 cross-sectional study using the Survey of Income and Program Participation, 26,215 participating households, found that food insecurity was significantly higher for households that were entirely Black (prevalence ratio 1.18, 95% CI 1.04 to 1.33) and multiracial (PR 1.25, 95% CI 1.06 to 1.46) than for entirely White households. Then the authors split the sample by program participation, and the pattern changed:

Among households that did not participate in SNAP, those that were entirely Black (PR 1.52) or multiracial (PR 1.42) were more likely to be food insecure. The disparity was not present among households that did participate.

That is reform doing its job. A federal program, administered by a bureaucracy, funded by appropriation, closed a measurable racial gap in hunger. If you want a source that reform works, that is a good one, and it is not an outlier. The observational literature on SNAP, WIC and school meals consistently finds reductions in food insecurity and improvements in child outcomes.

So the claim that reform never works is false, and anyone making it should be corrected. Now here is the part that makes the picture interesting.

Then the randomized trials that found nothing

The USDA funded demonstration projects to test whether raising SNAP benefits reduced child food insecurity. Two of them ran randomized controlled trials. Both came back null on the primary outcome.

Kentucky, Ticket to Healthy Food (17 counties, 2,202 households at baseline, 1,639 at follow-up):

Child food insecurity was 37.1% in the treatment group and 35.2% in the control group, P = 0.812. Very low food security among children, P = 0.204. Adult food insecurity, P = 0.654.

Nevada, Healthy Hunger-Free Kids (Las Vegas, 3,088 households at baseline, 2,074 at follow-up):

Child food insecurity was 31.2% in the treatment group and 30.6% in the control group, P = 0.620.

Raising benefits did not move the headline measure in either trial. And both trials found something else that matters: household food spending rose by $20 a month in Kentucky (P = 0.030) and $23 a month in Nevada (P < 0.001), and in Kentucky more households reported their benefits lasting at least three weeks, 65% against 56% (P = 0.009).

The Nevada authors were honest about the tension. Their conclusion says the null "runs counter to prior research showing that SNAP and similar forms of food assistance have reduced food insecurity," and notes the project ran during a period of substantial economic growth.

Both findings are true, and the reconciliation is the useful part

This is where a serious reading beats both slogans.

Food insecurity is measured as a threshold. A household either crosses it or it does not. An extra $20 a month is real money and it demonstrably changed what people bought, and it was not enough to move a household from one side of a binary line to the other. That is a dose finding, not a verdict on transfers. Two trials of a modest benefit increase, run in a growing economy, found the increase too small to flip the measure.

So the correct statement is narrower and stronger than either side's version:

• Large, sustained transfers close gaps. The disparity study shows it.
• Small, marginal increases do not clear a threshold. The trials show it.
• Neither result tells you anything about whether the structure of the food system can be reformed, because neither trial attempted that.

Now the question actually being asked

The person asking for a source was not asking whether a benefit increase moves a survey threshold. They were asking whether the system changes when people want it to. That question has its own literature, and the answer is the one you remember.

Gilens and Page, Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens, Perspectives on Politics, 2014. They tested four theories of American politics against each other inside a single statistical model, using measures of the key variables across 1,779 policy issues. From the abstract, verbatim:

Multivariate analysis indicates that economic elites and organized groups representing business interests have substantial independent impacts on U.S. government policy, while average citizens and mass-based interest groups have little or no independent influence.

They go on to say the results support Economic-Elite Domination and Biased Pluralism, and do not support Majoritarian Electoral Democracy or Majoritarian Pluralism. That paper has been cited about 2,700 times. The finding is not fringe and it is not new.

Read the direction of that sentence carefully, because it is often misreported. It does not say public opinion never matters. It says that once you account for what economic elites and organized business groups want, the independent effect of average citizens on policy is near zero. Public opinion and elite opinion usually move together, and when they do, policy follows and everyone claims credit. The test is what happens when they diverge. That is where the coefficient goes to nothing.

The mechanism, which requires no villains

This is usually presented as a conspiracy. It is not one, and understanding that makes the conclusion stronger rather than weaker. Four pieces, all old, all boring, all sufficient.

Collective action arithmetic. Mancur Olson, The Logic of Collective Action, 1965. A benefit spread across millions of people is worth a few dollars each. A cost concentrated on a few firms is worth billions each. The concentrated side organizes because it pays to organize. The diffuse side does not, because for any one person the expected return on political effort is smaller than the effort. This is arithmetic, not indifference.

Regulatory capture. George Stigler, 1971. The industry being regulated has the expertise, the lobbyists, the revolving door and the sustained attention. The public has a news cycle. The rules get written by whoever is still in the room.

The bias of the agenda. E. E. Schattschneider, The Semisovereign People, 1960. Some questions never reach a vote. The most effective use of power is keeping an option off the table, because a proposal that is never scored cannot be defeated on its merits.

Non-decision-making. Bachrach and Baratz, 1962. Power is not only winning the decision. It is limiting the scope of what gets decided.

Put those four together and you have an explanation for why the reform that passes is systematically the reform that does not threaten the arrangement. Nobody has to be bribed. The filter is in the structure.

What this looks like in food and agriculture specifically

The distinction between the two kinds of reform is clearest in the one domain where you can watch it for fifty years.

The ameliorative layer works. SNAP, WIC, school meals, food banks, and the charitable feeding network. These move money and food toward people who need it, they are measurably effective at reducing hunger and closing gaps, and they have been expanded and defended repeatedly. When people organize around hunger, they get hunger programs.

The structural layer does not move. Land ownership, consolidation across the input and processing sectors, the farmer's share of the consumer food dollar, and dependence on purchased inputs. Decades of attempted antitrust enforcement in agriculture, payment caps, and Packers and Stockyards action have not reversed consolidation. The number of US farms stood at 2,128,982 in 2002 and the long trend is down, not up. The local and direct market sector has grown enormously in absolute terms and remains a small single-digit share of food sold.

That combination is the finding. A system can become steadily better at softening its own outcomes while the structure producing those outcomes is untouched. The hunger programs are not a failure. They are the part that was permitted, and the part that was permitted is the part that does not require anyone to give up anything structural.

The conclusion, stated plainly

Reform succeeds to the degree that it does not threaten existing power, and fails to the degree that it does. That single sentence covers the disparity study, the two null trials, the 1,779 policy issues, and the fifty years of food policy.

So the original claim on social media was half right and stated badly. Reform does not have a low success rate. It has a high success rate at amelioration and a near-zero rate at transformation, and the reason is not that reformers are stupid or corrupt. It is that transformation requires the concentrated side to lose, and the concentrated side is the side that is still in the room when the bill is drafted.

The practical consequence is the part worth acting on. If the reform you can actually get is the reform that does not change the structure, then waiting for structural reform to arrive before you do anything is a way of never doing anything. Take the ameliorative programs, because they are real and they help. And build the household capability in parallel, because that is the part you control and the part no committee has to approve.

Produce more than you consume. Buy less of what you can make. Keep the skills in the house rather than in a supply chain. That work does not require a majority, a lobbyist, or a hearing, which is exactly why it is available and exactly why nobody is stopping it.

Sources

1. Gilens M, Page BI. Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens. Perspectives on Politics. 2014;12(3):564-581. doi:10.1017/S1537592714001595. 1,779 policy issues.
2. Testa A, Jackson DB. Supplemental Nutrition Assistance Program Access and Racial Disparities in Food Insecurity. JAMA Network Open. 2023. PMID 37358853.
3. Kentucky Ticket to Healthy Food randomized controlled trial. Journal of the Academy of Nutrition and Dietetics. 2021. PMID 33342530.
4. Nevada Healthy, Hunger-Free Kids randomized controlled trial. Journal of the Academy of Nutrition and Dietetics. 2021. PMID 33342522.
5. Olson M. The Logic of Collective Action: Public Goods and the Theory of Groups. Harvard University Press, 1965.
6. Stigler GJ. The Theory of Economic Regulation. Bell Journal of Economics and Management Science. 1971;2(1):3-21.
7. Schattschneider EE. The Semisovereign People: A Realist's View of Democracy in America. Holt, Rinehart and Winston, 1960.
8. Bachrach P, Baratz MS. Two Faces of Power. American Political Science Review. 1962;56(4):947-952.

What this piece does not claim. It does not claim that reform never helps, because the disparity study shows it helping. It does not claim that public opinion is irrelevant, because opinion and elite preference usually agree and policy follows. It does not claim that anyone is bribed. It claims that the independent influence of average citizens on policy is near zero once elite preference is accounted for, that this is measured across 1,779 issues, and that the food system shows the same split between the programs that pass and the structures that never move.