Social Security’s long-term shortfall is likely significantly worse than projected. In 2025, the Social Security Trustees estimated that the program faces a 75-year funding shortfall equal to $27 trillion in present-value terms. But those projections rely on unusually optimistic assumptions about future US fertility rates. The Trustees are likely understating Social Security’s insolvency problem by assuming Americans will start having far more children than current trends suggest.

If fertility instead follows projections from the Congressional Budget Office (CBO) or the US Census Bureau, Social Security’s 75-year deficit rises to at least $30 trillion. In other words, the Trustees and the Social Security Administration (SSA) may be understating the size of Social Security’s long-term financing gap by at least $3 trillion in present value terms.

Examining Social Security’s Fertility Rate Assumptions

Figure 1 compares the Trustees’ total fertility rate projections with those of other peer organizations. The Trustees assume that the current fertility rate of about 1.6 children per woman will increase to 1.9 by the early 2040s. But the Trustees are, by far, the most optimistic about future fertility. By 2100, the SSA projects a total fertility rate roughly 30 percent higher than both the Census Bureau and the CBO, who project that fertility is likely to continue to decline. 

The SSA gives three reasons for its optimistic fertility forecasts. 

First, the SSA cites American exceptionalism, arguing that the United States has a distinctive “cultural and economic climate” that will prevent the kind of prolonged fertility decline experienced in many other OECD countries.

Andrew Biggs at the American Enterprise Institute interprets this claim as referring to how religious Americans are when compared to their OECD counterparts. But Biggs also points out that young Americans are becoming less religious. 

One way to test whether the US exhibits uniquely high fertility patterns relative to its OECD peers is to compare actual US fertility rates against predicted OECD patterns. As Figure 2 shows, the United States converged to OECD fertility patterns after the 2007 Recession. 

This convergence has also been documented in the economics literature. In a 2022 report, economists Melissa Kearney and Phillip B. Levine argue that “the US fertility rate is belatedly converging toward other high-income countries,” while Erasmo Papagni also documents the strong statistical convergence of US fertility rates with other high-income, low-fertility countries. The SSA’s view that the United States will remain demographically distinct from its OECD peers no longer reflects current fertility trends.

Second, the SSA cites survey data that the number of children that Americans state they wish to have remains above 2.0. The Trustees take this to be evidence that fertility will likely increase toward that level. 

The SSA’s claim, however, goes against decades of demographic research that finds an increasing divergence between intended fertility and actual fertility. As Figure 3 shows, intended fertility rates do not closely follow actual fertility rates. 

To more rigorously test whether one can project actual fertility rates from intended fertility rates, we also conduct a Granger causality test for up to three lags (code here). These tests indicate that neither series (at least in aggregate) is useful for forecasting the other. As such, the fact that Americans say they want to have more than two children does not necessarily mean that Americans will have more than two children. 

This is not uncommon. Even in low-fertility Europe, women typically state that they desire a significantly larger family than they end up having. The Trustees should not rely on unrealized fertility intentions to project actual fertility. 

Finally, the Trustees suggest that falling fertility rates may simply reflect delayed births rather than fewer births overall, assuming that postponed births will eventually show up at older ages, thus stabilizing or increasing future fertility. 

The Trustees’ optimism is not supported by historical trends. Table 1 shows the change in birth rates by age group between 2007 and 2023. Older cohorts are indeed having slightly more children, and younger women are having fewer. But the gains among older cohorts are not nearly large enough to compensate for falling fertility among younger cohorts. 

We estimate that, from 2007 to 2023, for every additional birth gained at older ages, twelve births were lost at younger ages. So, for delayed births to stabilize or increase fertility, older cohorts should have at least twelve times more children. This may be technically possible, but historical evidence gives us little reason to consider it likely. It is also unrealistic if we consider that many other countries have experienced a delay in fertility without any corresponding increase in fertility rates.  As such, the Trustees cannot reasonably claim that delayed fertility will stabilize or increase total fertility rates in the future. 

Consequences of the Trustees’ Optimistic Projection

Figure 4 plots the SSA’s projections of total population against peer forecasters. The Trustees project a far larger population compared to the CBO or the Census Bureau, by as much as 25 percent above the next largest estimate.

Figure 5 plots the SSA’s projections of the dependency ratio against peer forecasters. Again, the SSA produces low forecasts for the dependency ratio, with its 2100 forecasts a fifth lower than the next-nearest projection. 

Optimistic population assumptions translate into optimistic financial projections. In a pay-as-you-go system like Social Security, higher assumed fertility means more future workers paying taxes to finance benefits and thus a smaller projected shortfall. To estimate the fiscal impact of the Trustees’ fertility assumptions, we rerun the Trustees’ intermediate long-range projections using CBO and Census fertility projections instead (code here). As Figure 6 shows, adopting more realistic total fertility rate assumptions can lead to much more pessimistic projections of Social Security’s 75-year balance. 

The Trustees project that Social Security’s 75-year shortfall is $27 trillion in present value terms. However, with more realistic total fertility rate assumptions, Social Security’s 75-year shortfall increases to $30 trillion (with Census projections) or $31 trillion (with CBO projections) in present value terms. 

Note that the Census’s and CBO’s projections of future fertility may also be optimistic. Both the Census’s and CBO’s estimates assume that long-term fertility rates will plateau at around 1.53 to 1.56. Yet fertility has been declining in the United States for the last two decades. And there seems to be little reason to believe that this decline will stop anytime soon. 

Congress Should Demand Answers

The Trustees’ fertility assumptions are not a technical footnote. Because Social Security is financed on a pay-as-you-go basis, fertility projections directly shape estimates of the program’s long-run solvency. If the Trustees’ fertility assumptions are too optimistic, then Social Security’s financing gap is significantly larger than official projections suggest. 

As Congress debates Social Security reform, it should be wary of relying on demographic assumptions that are substantially more optimistic than those used by other government forecasters. Policymakers risk basing reform decisions on projections that overestimate the future size of the workforce and, in turn, the program’s future revenue base. At a minimum, lawmakers should explore why the Trustees’ fertility projections appear to be such outliers, and the Social Security Advisory Board should reconvene the Technical Panel on Assumptions and Methods to better assess the fertility rates likely to occur in the US in the future.

The authors thank Andrew Biggs for reviewing and editing this piece. 

State-run, federally funded welfare programs lose billions of dollars to improper payments every year. Examples abound. Medicaid paid more than $200 million to deceased enrollees in one year; 2.8 million people were enrolled in two or more Medicaid/​Obamacare exchange plans in 2024; and half a million Supplemental Nutrition Assistance Program (SNAP, formerly known as food stamps) recipients were reportedly receiving benefits twice or more under the same name.

Medicaid and SNAP alone combined for nearly $50 billion in payment errors in fiscal year (FY) 2025. Most of that money is never recouped.

The Treasury’s Do Not Pay (DNP) system, a data-sharing tool that screens payments for eligibility in real-time, is designed to solve these problems by catching erroneous payments before the money goes out the door. But the system is underutilized and underequipped, and agencies are not required to respond when it flags potential errors. Congress should address these shortcomings by strengthening and expanding the use of DNP. More importantly, Congress should pair it with structural reforms that give states a stronger reason to use the system properly.

Preventing Improper Payments Before They Happen

The Congressional Research Service (CRS) has identified data-access issues as one of the root causes of payment errors—agencies either cannot access the information they need to verify eligibility or fail to use the data already available to them.

Additionally, rather than catching errors up front, agencies frequently default to a difficult, expensive, and largely ineffective “pay and chase” approach. They issue the payment first, find the error later, and try to claw it back after the fact. It rarely works. Of the $175 billion lost to overpayments in FY 2023, only 15 percent was successfully recouped.

The DNP system is meant to stop this cycle by providing agencies with access to data sources to cross-check and verify applicant eligibility at the point of issuance.

The system has a proven track record. In a three-year pilot that gave DNP access to the Social Security Administration’s (SSA) Death Master File, the system identified or prevented $113.5 million in improper payments at a cost of just $4.6 million—a 23-to‑1 return on investment in the first year alone. After such remarkable success, Congress permanently authorized SSA to share its full death data with DNP.

In FY 2025, DNP reportedly helped agencies prevent, detect, and recover $11.7 billion in potential improper payments.

House Oversight Bills Strengthen DNP

Despite its proven value, DNP has several gaps that limit its effectiveness. A House Oversight hearing on fraud in state-run, federally funded welfare programs highlighted the need to expand and improve the system and give it the data, authority, and reach that it currently lacks. In April, the committee advanced several bills aimed at doing exactly that.

• DNP lacks access to important databases. The system screens well for deceased enrollees, but only $1.5 billion in FY 2024 overpayments were death-related. Far larger drivers of overpayments—income and financial data ($80.2 billion), employment status ($68.9 billion), and identity ($31.6 billion)—are not fully accounted for in DNP’s current databases. The Pre-Payment Fraud Prevention and Treasury Data Access Act (H.R. 8463) from Reps. James Comer (R‑KY) and Jodey Arrington (R‑TX) would help address these gaps by expanding DNP to include the National Directory of New Hires, SSN/​name matching, and tax return income data for agencies to use during prepayment eligibility screening.
• DNP can flag a potentially improper payment, but agencies are under no obligation to act on it. As Rep. Comer points out, Treasury “is not fully empowered to stop” potentially improper payments flagged by DNP, leaving agencies free to disburse funds even when the tool identifies a risk. To address this, Reps. Comer and Arrington’s Stopping Fraudulent Payments Act (H.R. 8464) would require agencies that wish to proceed with a payment flagged by DNP to first take corrective action and verify the payment before certification.
• DNP is woefully underused. Only 4 percent of eligible programs across the federal government fully utilize DNP. Many states lack access to DNP and instead rely on outdated and fragmented data systems with significant gaps in eligibility verification. The Timely and Accurate Benefits Act (H.R. 1755) from Rep. William Timmons (R‑SC) would help close this gap by expanding DNP’s availability to state agencies and requiring them to use DNP or other real-time data-matching tools for fraud prevention and eligibility determinations.
• Agencies aren’t using the data they already have. CRS found that 73 percent of data access–related overpayments between fiscal years 2021 and 2024—$556.6 billion—were human errors resulting from agencies failing to use the data that already existed. Ensuring administrators know how to use the tools available to them is one potential solution, and the Federal Fraud Prevention Workforce Training Act (H.R. 8428) from Reps. Glenn Grothman (R‑WI) and Raja Krishnamoorthi (D‑IL) would help by establishing a training program on fraud risk management and the use of tools such as DNP for federal and state agencies administering federal programs.

Together, these bills would give DNP better data, strengthen enforcement when it flags a potential error, and expand its availability to more states administering federal programs.

Better Tools Still Need Better Standards

Even with these improvements, DNP would still have significant limitations that the aforementioned bills do not address. CRS cautioned that the system sometimes produces false positives and contains inaccurate data, undermining its effectiveness for some agencies. Moreover, even if every federal agency were using the DNP system as intended, it cannot account for the different laws and regulations governing eligibility across programs, which require case-by-case determinations that no centralized database check can substitute for.

Congress should continue to expand DNP to give agencies more information during eligibility screening, such as access to the Systematic Alien Verification for Entitlements service to verify applicants’ citizenship status. Legislators should also ensure the Treasury continues to improve the accuracy of data already in the system and establish clear standards for what triggers a payment delay. H.R. 8464 triggers if a payment has an “elevated risk of fraud” based on a “fraud-risk indicator” but doesn’t specify what either term means in practice. Stricter screening is warranted, but poorly defined criteria risk creating administrative backlogs and unnecessary red tape that delay payments to eligible recipients.

Accountability Starts With Incentives

More fundamentally, DNP is a tool that is only useful if agencies have the right incentives to use it. As the Government Accountability Office’s Seto Bagdoyan testified, states can have intricate fraud risk management systems, but without a “culture of accountability,” those systems will not deliver results. Many of the 90-plus federal welfare programs are state-administered but funded largely, and in some cases almost entirely, by federal taxpayers. States face few fiscal consequences for tolerating waste, fraud, and abuse when the costs fall on federal coffers rather than state budgets. Ideally states choosing to run welfare programs would both fund them and answer directly to the constituents paying for them.

Congress should strengthen and expand use of the DNP system to ensure taxpayer dollars reach the people they are intended for. But relying on DNP is not enough. Fraud prevention works best when the governments administering benefits are also responsible for the costs of errors.

The federal government collects and publishes extensive data that divide Americans by race. These data have been used to target racial groups and execute policies that discriminate among individual Americans, giving or denying benefits or preferences, and providing unequal treatment to them based on their race or ethnicity.

Erosion of American liberty through racial profiling was severe in the early days of the republic with slavery, but it was slowly mitigated with the abolition of slavery, the elimination of Jim Crow laws, and the passage of the 1964 Civil Rights Act. At the same time, civil society has become far less intolerant across racial lines. Today, 20 percent of newly married couples cross racial lines in selecting a partner for the most intimate human relationship, with the interracial proportion reaching almost half for some races. 

Unfortunately, the government has pushed racial profiling to differentiate and mistreat based on race while claiming to pursue diversity. A disturbing recent example of this perversity is the 2024 Office of Management and Budget (OMB) revision of “Statistical Policy Directive No. 15: Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity” (SPD 15), which directs federal agencies to expand data by race. 

SPD 15 may seem to be merely statistical, but the harsh restraint on liberty is revealed by its requirements: “These standards must be used by all Federal agencies for civil rights and other compliance reporting from the public and private sectors and all levels of government.”

On day one of his second term, President Trump issued a promising executive order, “Ending Radical and Wasteful Government DEI Programs and Preferencing.” The next day, another order set more specific requirements to stop “demeaning and immoral race- and sex-based preferences under the guise of so-called ‘diversity, equity, and inclusion.’” It required all agency heads to develop “a plan of specific steps or measures to deter DEI programs” within 120 days. 

Because race-based data collection and publication are enablers of those noxious initiatives in both government and the private sector, one would expect an immediate reversal of the 2024 expansion of race-based data and hope for a complete elimination of those unconstitutional and immoral data. The specific requirements for race-based data collection exist only through the administrative OMB SPD 15, so plans to reverse and eliminate them would be a straightforward administrative action. 

Unfortunately, a year after the plans required by the executive order were due, there is no evidence of any plans to stop the expansion of race-based data. In fact, the “2030 Census Operational Plan” dated July 2025, six months after the executive orders, lists as its top demographic improvement: “Increasing the level of race and ethnicity detail for the estimates.”

On May 5, 2026, the official Census Bureau website reported, “The Census Bureau is preparing to implement OMB’s updated 2024 race and ethnicity standards in the 2027 American Community Survey and the 2030 Census.” Furthermore, as recently as April 29, 2026, the Census was using public funds to conduct free webinars to train people in the application of race data, as announced in the following email.

The link in this email has since been disabled, with an “Error 404: Page Not Found.” One cannot be certain whether this is merely sloppy housekeeping or a conscious effort to hide the activity after the conflict of this webinar with stated policy in the executive orders was pointed out.

Another webinar was announced to train organizations to use race data to qualify for grants. As with the previous webinar, the link to the relevant page has been disabled.

Fortunately, I had previously retrieved the material used in this webinar before the site was disabled. Here is page 14 from that material, which shows race and ethnicity as the third item in a list of eleven data elements available from the Census to help people win grants.

These announcements and webinars illustrate how the Census Bureau is still moving full speed to promote racial profiling to support disparate impact and DEI objectives. Strong executive action is required to stop and reverse this refusal to change.

Timing is critical. Any substantive changes to the 2030 Census must be submitted to Congress by April 1, 2027. The current plans implement more extensive racial profiling under the 2024 SPD 15, which, without quick intervention, would be included in that congressional submission. Stopping this tragedy requires only administrative changes, but they must go through the Administrative Procedures Act steps, which require significant time.

Although OMB and Census have not moved to fix SPD 15 and its implementation, bureaucrats are keeping track of the formalities. OMB has issued two revisions to the date by which implementation plans are required from agencies—first from September 29, 2025, to March 28, 2026, and then to March 28, 2027. When will leadership move to correct the substance?

During the public comment period before the 2024 expansion of SPD 15, the case against the expansion was well documented, including a detailed analysis of how it could harm newly designated races. Then, after the revised SPD 15 was promulgated, further analysis detailed its flaws and the actions needed to end it.

In addition to the flaws outlined above, these previous analyses have shown the following:

1. By their own admission, OMB and Census created races that are merely subjective responses from Census surveys, not scientific facts.
2. Even these subjective responses are corrupted by Census edits that change the actual responses from residents when Census bureaucrats don’t like the answer, for example, forcing Brazilians into the residual “Some Other Race and/​or Ethnicity” category despite the fact that 67 percent of them identify as Latino.
3. The racial coding specifies 6,676 detailed racial categories. The racial distinctions are so extensive that the detailed classes within the American Indian and Alaska Native category have an average of only 678 people in each class.
4. The definitions of races are inherently wrong. They are of the form: “[A race]. Individuals with origins in any of the original peoples of [a geographic area], including, for example, [some races].” With the possible exceptions of a very few remote tribes, we have no record and no idea of who the original people were for any piece of land on earth. For at least 200,000 years, Homo sapiens have moved across the globe, removing previous inhabitants from the land by expulsion, murder, enslavement, or absorption. One example illustrates the broad failures in the directive. OMB defines: “White. Individuals with origins in any of the original peoples of Europe, including, for example, English…” 
   1. The first example is English, which is absurd. The word itself comes from the Angles, who, with the Saxons and Jutes, conquered the island of Britannia in the fifth century, following the withdrawal of Rome after its 500-year occupation of the Celtic Britons, who, in turn, were one of the Celtic tribes that had migrated there from Iberia, which is now Spain. So, are the “original” people of the island Britannia some conquerors from Espana and hence Hispanics? Or are the originals the Anatolian soldiers in the Roman army who left their genetic imprints on inhabitants of “English” villages even to this day? Certainly, they are not the English derivates of the Angles who didn’t invade before 450 BCE. How about the Vikings or Norsemen from the ninth century or the Norman French from 1066? Not to mention the Paleolithic hunter-gatherers who were driven from the island by the last ice age.
   2. Census also labels Turks as White, which by definition means they are the original people of Europe. But only 15 percent of Turkey’s population and 3 percent of its landmass are on the western (European) shore of the Bosporus. The 97 percent of Turkey that lies in Asia has a greater landmass than any European country. Reference books, the UN, and the Turkish government itself identify Turkey as primarily lying in Anatolia, Asia Minor, or Western Asia. Furthermore, Turks are not even close to being “original” people. They conquered and settled in eastern Anatolia only in the late eleventh century, coming from a part of Central Asia, an area now constituting Turkmenistan. For more than 300 years, they systematically annihilated, exiled, enslaved, or absorbed Greeks, Armenians, Kurds, and dozens of other ethnicities in Anatolia until they finally completed their conquest with the fall of Constantinople in 1453.
   3. And what about Spaniards? This definition makes them European and White. But OMB has snatched them from all other Europeans and made them “Hispanic or Latino.” Spanish is a European language of the Romance family. Creating a separate race of “Hispanic or Latino” makes no more sense than creating a separate race for “Anglophones” from Great Britain, the US, Canada, Australia, New Zealand, Singapore, the Falkland Islands, and some others. 
5. Collecting data by race is also illegal, violating 13 US Code § 221(c): “Notwithstanding any other provision of this title, no person shall be compelled to disclose information relative to his religious beliefs or to membership in a religious body.” Responses to racial questions on the decennial census are compulsory, and some choices of race require disclosure of religious beliefs. For example, Serbs, Croats, and Bosniaks are each identified and coded differently in the Census racial classification. Yet, before the ninth century, the ancestors of all three were one undifferentiated people. They still speak the same language, and even native speakers typically cannot distinguish the different “native” tongues of their interlocutor from the broader population. A compulsory race question to someone from one of these ethnicities is equivalent to asking them to declare their religious affiliation or heritage, and is a clear violation of the law.

These examples are only illustrative of fundamental flaws in the OMB/​Census scheme of race classification. Combined with the constitutional and ethical failures, they constitute an overwhelming case to stop the madness and eliminate racial profiling in our statistics now. 

Starting the administrative process immediately to correct SPD 15 can halt racial profiling and disparate impact analysis for the 2030 decennial census, as well as for the American Community Survey, the Current Population Survey, and other statistical systems that engage in racial profiling. Ideally, this administrative initiative should be followed by legislation to eliminate all government collection and publication of racial profiling data so that a later administration cannot easily reverse the progress. Even without legislative follow-up, this change can immediately eliminate the harm of racial profiling and disparate impact analysis at least for the present.

AI in Health Care: A Policy Framework for Innovation, Liability, and Patient Autonomy—Part 8

The Senate Judiciary Committee advanced Senator Josh Hawley’s Guidelines for User Age-verification and Responsible Dialogue (GUARD) Act. The bill would require every American to verify their age before using a generative AI chatbot and would bar anyone under eighteen from using a “companion” chatbot at all. In the room during the markup were the parents of children who died by suicide after conversations with AI products. Their grief is unimaginable, and their motives are beyond reproach. But, concerningly, such a policy might quietly cost rather than save lives.

The strongest claim animating this bill is the belief that restricting minors’ access to AI chatbots will prevent suicide. On the available evidence, that claim is closer to a hypothesis than a finding—and a hypothesis that runs against several decades of data on how young people die. 

According to the Centers for Disease Control and Prevention, the American suicide rate began climbing around the year 2000—before ChatGPT, smartphones, or social media even existed. It accelerated through the 2010s, then, contrary to popular narrative, plateaued and modestly declined after 2018—even as generative AI moved from research labs into the pockets of nearly every teenager in the country. If chatbots were a meaningful driver of adolescent suicide, the curves should have moved together. They have not, and, importantly, suicide rates among young Americans remain the lowest among any age group. 

While any loss of a young life to suicide is a tragedy, whatever is killing young Americans predates the technology that lawmakers now propose to ban them from using. 

What the GUARD Act’s sponsors do not seriously consider is the other side of the ledger. There are cases where AI could help Americans of all ages when it comes to mental health. Roughly half of Americans with a diagnosable mental health condition never seek professional help; stigma, cost, and fear of involuntary intervention keep them silent. For some of them—especially adolescents in households where therapy is unaffordable, unavailable, or unsafe to disclose—a chatbot is their most reliable form of emotional support. 

In a survey of over 1,000 adolescents and young adults, 13 percent had used a chatbot for mental health support, and more than 90 percent of those found it helpful. In another study of over 1,000 users of Replika, a popular AI chatbot, 30 reported without solicitation that their artificial companion saved them from suicide. 

We do not know how many lives generative AI has saved by improving access to mental health care. But for every incidence of AI psychosis or suicide, there may be dozens of unobserved positive outcomes. Policy that presumes only the worst outcomes also prevents the best.

The consequences of the proposal could also dissuade investment or chill speech that would make better options available. Faced with $100,000 per-violation penalties, providers will not invest in better suicide-detection models and instead likely remove any content that could be related to such a topic, thus limiting resources to crisis hotlines for those who are struggling. It would also limit the availability of information for those seeking to understand a deeply traumatic event or help a friend who may be struggling. Clinicians have known for decades that abrupt treatment referrals without first building rapport can deepen shame and shut down disclosure. The best science suggests suicide-prevention frameworks place trust-building before resource provision precisely because the order matters. A regulatory regime that punishes providers for nuance will produce less of it.

Beyond being bad policy, such laws are almost certainly unconstitutional. The underlying policy is not based on a compelling government interest nor is it narrowly tailored. It impacts the speech rights and anonymity of all users of online tools, not just minors, on the basis of justifications that are far from accepted. The compliance regime is broad enough to capture homework helpers, customer-service chatbots, and search engines that produce conversational responses, placing a “papers, please” approach to a broad and growing swath of the internet. To enforce it, every American adult would have to upload a government ID or submit to biometric scanning to ask a question, complete a customer service interaction, or practice a foreign language.

More measured and better policy responses are available if policymakers want to support parents and teens who may encounter difficulties with AI chatbots or generative AI. That includes training and providing appropriate resources for law enforcement to go after the bad actors who abuse technology to create or solicit sexual content from minors. Investment in AI literacy, of the kind Idaho recently codified into its public schools, equips young people to use these tools the way they will inevitably need to use them as adults and can include information on what to do if they encounter problems. 

Far from being a problem, liability shields modeled on Section 230, paired with safe-harbor incentives for providers that invest in better mental-health detection, would reward the kind of careful development the current bill punishes. None of those would deliver the cathartic clarity of a ban, but all of them are more likely to save lives. Importantly, they also empower parents and other trusted adults, not policymakers, to be the ones who determine what makes sense when it comes to kids and teens’ AI use.

The bill’s sponsors are not acting in bad faith. The cases motivating them are real, and the impulse to protect the vulnerable is one of the more honorable features of our political instincts. But the pattern is familiar from earlier moral panics over comic books, rock music, and video games. Each was sincerely felt. Each rested on weak social science amplified by strong public emotion. Each produced a policy that aged poorly.

The GUARD Act asks us to trade a measurable loss of liberty and privacy for an unmeasured, and possibly negative, impact on safety. The forces that drive people toward suicide—isolation, family conflict, untreated illness, loss of meaning—operate on timescales and through mechanisms that no technology policy will address. To pretend otherwise is to offer grieving families a consolation that policy cannot honestly deliver while quietly closing a door through which other young people, less visible to us, are still walking toward help. 

To read other parts of this blog series, go here.

Congress is reportedly weighing whether to use reconciliation—a special budget process that allows certain legislative changes to be enacted with a simple majority vote in the Senate—to provide $200 billion for the war in Iran, plus additional money for immigration enforcement. Neither issue merits new spending. But if legislators nevertheless choose to use reconciliation, they must make the package fiscally responsible. 

Ideally, Congress should aim to reduce deficits to 3 percent of GDP or lower, a commonsense target to achieve debt sustainability. Short of that, Congress should, at a minimum, achieve $600 billion in net deficit reduction. That’s the amount the final 2025 reconciliation bill exceeded the House-passed maximum allowable non-interest borrowing. Anything less falls short of the bare minimum amid a mounting fiscal crisis.

A $600 Billion Deficit Reduction Target Should Be the Bare Minimum

The federal government routinely runs annual deficits of $2 trillion. Publicly held federal debt is projected to exceed the nation’s entire annual economic output this year. By 2030, debt is projected to surpass the World War II debt-to-GDP record high. By 2056, the debt will reach 175 percent of GDP, even under optimistic projections with no new wars, pandemics, or economic recessions.

The main culprits of America’s persistent and growing deficits are old-age entitlement programs, chiefly Medicare and Social Security. These two programs, along with interest costs and other mandatory spending programs, dominate the federal budget, growing on autopilot without regular congressional approval or review. Absent reform, entitlements and interest costs will consume all federal revenue by 2036.

Additional deficit spending is inexcusable given this trajectory, and simply aiming for deficit neutrality is insufficient. Consider that the $200 billion being currently floated for the Iran war would add roughly $87 billion in additional interest costs between FY2026 and FY2036. That $287 billion price tag does not account for the hidden health care costs for veterans of the war, additional immigration enforcement spending Congress is talking about adding, or the enormous constellation of deal-sweetening spending provisions that might be tacked on to secure votes in an election year with tight vote margins in both chambers. Reconciliation 2.0 could easily spiral into another budget buster unless legislators set a firm deficit-reduction target early in the process.

Last year, Republicans used reconciliation to pass the One Big Beautiful Bill Act (OBBBA), a tax-and-spending bill that added $3.4 trillion to 10-year primary deficits. Originally, the House budget resolution—the fiscal framework that guides reconciliation—allowed for a maximum 10-year primary deficit increase of $2.8 trillion. The final law exceeded this threshold by $600 billion.

Accordingly, legislators should demand that any reconciliation 2.0 effort include a binding requirement for a net deficit reduction of at least $600 billion—ideally aiming much higher, with the objective of reducing deficits to 3 percent of GDP or less. Only deficit reduction of this magnitude can arrest the growth of America’s national debt. 

Cost-Saving Measures Abound

There is no shortage of options to secure at least $600 billion in net deficit reduction. We catalog a range of options here, including reforms to Medicare, Medicaid, Obamacare, SNAP, and the tax code. Reforms to Social Security are also needed, but changes to the government pension program are not eligible for reconciliation. A variety of deficit-reducing budget options are included in the PDF here.

As an illustrative example, Congress could eliminate a loophole used to avoid SNAP eligibility standards (Broad-Based Categorical Eligibility), end Medicaid financing scams, and eliminate Medicare Advantage subsidies linked to waste. Those three reforms would save upwards of $1.1 trillion, more than offsetting a $200 billion spending increase and surpassing the $600 billion deficit-reduction target.

The offset exercise itself—rather than any particular combination of reforms—is the main point of requiring at least $600 billion in deficit reduction early on. Congress needed to start offsetting every new piece of legislation yesterday. With interest costs already exceeding what the United States spends on defense, taxpayers cannot afford existing deficits, let alone for Congress to increase them.

Fix the Process

Beyond offsets, Congress should also address the recurring abuse of reconciliation to widen deficits and bypass appropriations. Reconciliation was originally intended to facilitate deficit reduction, increase revenues, and cut mandatory spending on programs like Medicare and food stamps.

Over the last decade, Congress has increasingly used it for opposite ends. Under Biden, Democrats used reconciliation for new stimulus checks with the American Rescue Plan, as well as green energy subsidies and IRS funding through the Inflation Reduction Act. Under Trump, Republicans added nearly $375 billion in new spending via the OBBBA, primarily for defense and immigration. Now, a second reconciliation bill is again being pitched as an opportunity to spend more.

This has insidious consequences. The appropriations process is supposed to be Congress’s annual opportunity to scrutinize spending levels and consider tradeoffs. That’s particularly true if discretionary spending caps are in place. 

By contrast, using reconciliation for new spending lowers the bar for enactment, encouraging lawmakers to prioritize speed and political wins over prudent design. And when discretionary programs are shifted over to the mandatory side of the ledger, reconciliation erodes fiscal controls and the checks and balances embedded in the appropriations process.

The long-term consequences will be larger deficits and more partisan policies. Future Congresses and administrations will face political pressure to use reconciliation for whatever the day’s political priority is. And when they do, fiscal responsibility will suffer.

One option Congress should consider is to restore the Conrad Rule, a simple Senate rule that prohibited reconciliation from adding to deficits. Better yet, Congress could adopt a “Super” Conrad Rule, requiring any new deficit spending to be offset on a 2:1 basis—two dollars in savings for every new dollar spent. Congress routinely underestimates the true fiscal costs of new tax and spending measures. A 2:1 offset ratio would deter the kind of excessive logrolling that ballooned IRA and OBBBA’s deficit impact and help ensure reconciliation is used to improve the fiscal outlook rather than worsen it.

Pay for It, or Don’t Spend

New spending for the Iran war and immigration enforcement without offsets would be fiscally irresponsible and institutionally corrosive. It would accelerate the growth of the deficit and continue a now-bipartisan abuse of using reconciliation as a blank check. 

Reconciliation 2.0 must follow through on the promises made during the OBBBA process and meaningfully reduce the deficit. At a minimum, legislators should demand at least $600 billion in net deficit reduction and commit to using reconciliation as a tool for fiscal discipline.

The message is simple: pay for what you want to spend, or don’t spend at all.

The federal government spent roughly $100 billion on the Supplemental Nutrition Assistance Program (SNAP) in fiscal year 2024, but at least $10.5 billion of that was lost to improper payments, according to the Government Accountability Office.

To improve integrity and reduce payment errors in SNAP, the One Big Beautiful Bill Act (OBBBA) requires states with improper payment rates above 6 percent to pay between 5 and 15 percent of SNAP’s benefits starting in fiscal year 2028.

The ink on the bill is barely dry. Yet states are wasting no time exploiting a last-minute carveout to quickly undermine what the reform intended.

The “Alaska Carveout” Loophole

To address Senate concerns, Republicans included a provision—nicknamed the “Alaska carveout”—that delays the cost-sharing requirement for states with exceptionally high improper payment rates by up to two additional years. Specifically, states with error rates above 13.34 percent in fiscal year 2025 won’t face cost-sharing requirements until fiscal year 2029. Likewise, those exceeding the threshold in fiscal year 2026 get an exemption until fiscal year 2030. The delay can only be used once, based on either fiscal year 2025 or 2026 data, but not both.

In total, ten SNAP jurisdictions would qualify for this temporary reprieve based on fiscal year 2024 data: Alaska, the District of Columbia, Florida, Georgia, Maryland, Massachusetts, New Jersey, New Mexico, New York, and Oregon (Figure 1).

As Reason reporter Eric Boehm points out, this provision inadvertently rewards the worst-performing states while penalizing those that improve their oversight. For example, Senator Brian Schatz (D‑HI) highlighted this perverse incentive, noting that Hawaii reduced its error rate from 20.94 percent in 2023 to 6.68 percent in 2024, only to face penalties for doing so. Had Hawaii kept its error rate high, it could have avoided paying more than $30 million in potential cost-sharing requirements.

New Mexico offers another urgent example. The state reported a 14.61 percent error rate for payments in 2024 and had been making steady progress toward reducing it to below 10 percent by 2026. But, as Patrick Lohmann reported for Source NM, officials have abruptly halted that effort, with the director of the state’s Income Support Division admitting to doing a “balancing act”—reducing New Mexico’s SNAP error rate, but not before the end of fiscal year 2026. The immediate financial stakes explain why: dropping below 13.34 percent could trigger up to $153 million in cost-sharing requirements. It’s now cheaper for the state to maintain high error rates.

Officials in other states have been even more brazen in their misconduct. As Brooke Conrad reported for Fox Baltimore, a whistleblower claimed that two senior officials in Maryland were planning to deliberately leave correctable errors uncorrected to keep the state’s payment error rate elevated so it could “get the state off the hook for $240 million in new federal penalties.” After disclosing this “unlawful and unethical conduct” to the Office of the Inspector General, the whistleblower was fired.

These loopholes have broader consequences: OBBBA’s requirements also risk penalizing states with transparent quality control processes while rewarding those who game the system by underreporting improper payments.

Most states already have a track record of doing this. In 2014, the USDA identified quality control issues in 42 out of 53 state SNAP agencies. States had weakened their QC procedures by hiring private third-party consultants to exploit the subjectivity in the review process to artificially lower their reported improper payment rates and avoid financial sanctions.

Another tactic involves exploiting discretionary “no good cause” waivers, which allow states to exempt up to 8 percent of able-bodied adults from SNAP work requirements for any reason. As the Foundation for Government Accountability has highlighted, states can use these waivers retroactively to hide improper payments. From 2016 to 2020, roughly 100,000 discretionary waivers were issued retroactively across just five states.

How Congress Can Finish the Job

Looking at next steps, Congress had the right intentions in giving states some skin in the game to reduce improper payments in SNAP.

Nonetheless, with billions of dollars in penalties at stake, states have strong incentives to exploit loopholes rather than fix the underlying problems.

Congress must act immediately to build on OBBBA’s reforms and curb welfare fraud by eliminating the “Alaska carveout” and other loopholes that states exploit, such as “no good cause” waivers.

Requiring states to pick up part of the tab for SNAP benefits when they can’t keep improper payments in check is a reasonable step toward improving the program’s integrity. No state should be exempt from that arrangement.

Ultimately, however, the best way to fix SNAP’s moral hazard problems that underlie many of SNAP’s integrity failures is to eliminate federal funding for the program and leave nutrition assistance up to the states.

The best way to stop waste, fraud, and abuse in SNAP is to make states pay for it.

Spending on farm subsidies is rising after a series of costly expansions in recent years. The subsidies are not an emergency safety net for poor farm families but rather permanent welfare for high-earning businesses. With Congress now considering more farm spending, let’s examine the largest farm subsidy program—crop insurance.

The government often calls crop insurance “market based,” but that cannot be true because the program costs taxpayers billions of dollars a year. Also, it is run by a 413-person bureaucracy—the Risk Management Agency—that imposes a 941-page rulebook on the top-down system. The government works with 12 insurance companies that sell policies covering just about every farm contingency. The government subsidizes the companies and pays most of the insurance premiums.

The Congressional Budget Office (CBO) has published new estimates for crop insurance spending, summarized in the illustration. In 2026, crop insurance premiums are expected to be $20.4 billion, of which farmers will pay $7.8 billion and the government will pay $12.6 billion. 

Farm insurance policies are projected to pay out $17.4 billion this year in indemnities. Thus, farmers are expected to receive a net payment of $9.6 billion ($17.4 billion – $7.8 billion). The CBO expects similarly large net payments to farmers in coming years, so these are subsidies, not insurance. Even with the generous benefits, some farmers defraud the system for extra gains.

In 2026, the government is expected to pay the insurance companies $5.1 billion, including $2.4 billion for administrative costs and $2.7 billion for underwriting gains. The Government Accountability Office (GAO) found that the 12 crop insurance firms earn profits far above market rates. It seems very inefficient that to give farmers a subsidy of $9.6 billion, taxpayers also have to pay $5.1 billion to corporate middlemen.

All in all, taxpayers are expected to pay $14.7 billion in 2026 for the federal crop insurance program, with rising amounts in the years ahead, according to the CBO. Republicans expanded the program in their Big Beautiful Bill last year, and there may be further tweaks in the farm bill moving through Congress.

The crop insurance program is like the government giving you $900 a year for your $1,500 car insurance premium, all while paying billions of dollars to GEICO, State Farm, and other insurance firms to boost their profits. The government calls its crop insurance program “actuarially sound,” but it relies on huge subsidies from the actuarially unsound federal budget.

Here’s one more ridiculous thing about the program: There are no income limits on crop insurance subsidies. As a result, the subsidies flow heavily to the top end. One American Enterprise Institute analysis found that the largest 10 percent of farms receive 56 percent of all crop insurance subsidies.

GAO has reported that 1,341 crop insurance recipients have annual incomes of more than $900,000, and that even some billionaires receive the subsidies. We don’t know which billionaires because the program’s specific payments are secret. That combination of secrecy and high-income benefits has made crop insurance the favorite subsidy machine of the congressional farm committees.

In sum, here is the current situation with crop insurance and farm subsidies:

• Your earnings are taxed to pay subsidies to wealthy farm businesses and big insurance companies, of which half are headquartered abroad such as Switzerland-based Chubb Limited. 
• The government uses Orwellian language for crop insurance such as “market based” and “actuarially sound” to pretend that it is being responsible. 
• Despite huge federal budget deficits, Republicans continue to increase farm subsidies, all while GOP farm subsidy supporters claim to be fiscal hawks. 
• Republicans are currently moving another major farm bill through the House and may also add $15 billion of “emergency” farm spending on top.
• Republicans demand personal responsibility from low-income welfare recipients but do not apply the same approach to high-income farm welfare recipients.
• Farmers should save in good years to cover their losses in bad years. Farmers should make their own “safety nets,” as do millions of businesses in unsubsidized industries. 

Farm subsidies are a relatively small part of the federal overspending problem, but they exemplify the reckless approach that Congress is taking toward rising federal debt. If Americans elect a more fiscally conservative House and Senate this fall, crop insurance should be high on the list of wasteful programs to cut.