What is David Shor’s Theory of Change?

David Shor is the latest and greatest star among the wonks of think tanks and Twitter. The very bright data scientist has been blowing up the opinion sections of the national media the past week with three opinion articles in the New York Times, another in the Washington Post, and an entire feature in Politico. He has been making all the rounds for the past couple of years, from Vox to Bill Maher, preaching the good word of polls and reminding the left that the median voter in the United States does not share the same values as they do.

The latest run of Shor-press kicked off with his and co-contributor Simon Bazelon’s contribution to Slow Boring, the Matt Yglesias newsletter. The duo’s article was titled: “A permanent CTC expansion with a sharper means-test would protect poor kids better and be more popular.” Shor’s Blue Rose Research group has been busy conducting issue polling on hundreds of various issues, as he states here:

“Oftentimes, advocacy groups publish what amount to push polls, designed to show that every progressive issue position under the sun enjoys overwhelming support. Our polling is designed differently — it puts ideas through something resembling a real-world test to see which are genuinely compelling. That means using partisan frames: each question is structured as “Some Democrats in Congress have proposed Policy X, which would do Y and Z. Issue polling is also plagued by what’s known as acquiescence bias. This is the term for the tendency of survey respondents to just answer “yes” to any question, leading to overestimates of support for any given issue. Our polling mitigates this by providing a Democratic argument for a policy, and a Republican criticism. Our respondents are then asked “which party do you agree with more,” rather than “do you support this issue.”

The piece was undoubtedly very analytical and written by sharp minds yet concluded with policy recommendations lacking substance. The duo argued that the child tax credit should include additional means-testing to garner further voters’ support. The current child tax credit has an income threshold (for married couples) of $150,000, meaning the $3,600 per child annual tax break begins to phase out before reaching zero with a household income of $440,000.

The difference between no means-testing of the child tax credit (all parents receive $3,600 credit) and a threshold of $25,000 income is 48.9% support to 54.4% support, for a net change of 8.8 points. That puts the issue of not adding additional means-testing as unpopular for Democrats and thus likely to hurt them in electoral races. Shor’s entire philosophy is branded as “popularism,” using rigorous polling to garner the pulse of the nation so campaigning Democrats can win by running on the most popular liberal ideas. Shor doesn’t want the means testing, but that difference in polling was more than enough for him to recommend adding it. He adds a pretty poor defense that obfuscates the issue further:

“Given a choice between a more targeted benefit that is guaranteed to be around for poor families for years to come, versus a broader benefit with a significant chance of disappearing in just four years, it makes more policy sense to focus on protecting the poorest children.”

Neither the means-tested nor fully available child tax credit would be guaranteed to last forever, at most ten years. He is right that the Democrats have many limiting constraints at the moment. They have many issues to solve with assumedly even less than $3.5 trillion over ten years to spend. However, the more left-leaning critiques of means-testing are just as concerned with creating popular policy and certainly most concerned with protecting the poorest children, who are always the most at risk in a means-tested policy.

It may sound counterintuitive, but when social programs aren’t universal, you have difficulty serving the individuals who qualify for the program. This results in added administrative burden for both the state and the individual who must prove they can satisfy a test for income. From Matt Bruenig:

“The overall participation rate of the food stamp program is 85 percent and is only 75 percent for the working poor who likely have a harder time proving their eligibility to the welfare office. The participation rate of Medicaid is 94 percent for children, 80 percent for parents, and around 75 percent for childless adults. The participation rate of the Earned Income Tax Credit (and also presumably the Child Tax Credit) is 78 percent. The low participation in the EITC cuts the poverty-reducing effect of the program by around 33 percent, according to the Census Bureau, meaning that mainstream estimates of the EITC’s impact (e.g. those produced by CBPP) overstate the effectiveness of the program by at least 50 percent.”

The social democrat model would argue that the added administrative burden and lack of assistance due to income gap years due to getting fired or leaving the workforce would inevitably further cause support for the credit to decline. Whereas with a more available tax credit, you could potentially garner further support. It is important to note that even those in favor of a universal tax credit are also for taxing that money back from higher earners. Both a means-tested and a universal credit could have the same result to high earners through increased taxes, but the universal program would have the benefit of also reaching the poorest Americans. Sending a monthly $300 child tax credit to a household making $300,000 only to tax the amount back may sound counterintuitive until you factor in that the government already regularly sends payments and taxes household income. In contrast, the bureaucratic overlays required for testing income require more costly programs.

That battle, between well-designed, functional policy and policy that polls well, is ultimately the entire argument behind Shor and his more left-leaning detractors. Shor has a commendable strategy: he wants to get Democrats elected. The data paints a harrowing picture for Democrats, with the Senate and electoral college biased in favor of land instead of people. Thus, the party’s progressive wing must make concessions to help moderate Democrats win in conservative states. He believes the best bet to get blue lawmakers elected is to promote the most popular policies.

Shor is probably right on most of the above, the map looks terrible for Democrats in the short-term, and they will be disadvantaged for the foreseeable future. Unfortunately, he also has little theory of political change. Shor does not address the potential decline in popularity that could occur from a means-tested program that requires additional burdens on the poor and clawbacks, or overpayments due to the fact that the credit is based on income from the prior year (you never know the end-of-year income beforehand). Clawbacks are generally not appreciated:

“From 2003, there were substantial and very visible problems with overpayments, or more precisely visible problems in terms of the impact on claimants of having to repay overpayments. The number of overpayments, and their costs, were much higher than predicted. In the first year (2003/04), about one-third of all tax credits awards paid – nearly 1.9 million awards – were overpaid, at a cost of nearly £2 billion (House of Commons Treasury Select Committee, 2006). Media coverage reported confusion, hardship and debt as the government sought to recover overpayments.”

So, David Shor would like the Child Tax Credit to face means-testing at a threshold of $50,000, which results in a difference of polling of 48.9% to 53%. Research has often shown universal programs to be more popular than selective ones, but this meta-analysis concludes that it’s open for discussion, as the more popular policies may happen to be those that are universal, but not due to that criterium. Either way, this does represent a significant issue for Shor’s analysis because his proposal would be subject to Republicans doing away with it as well. It may be wiser to make a program that can build public trust and support, rather than a costly and over-bureaucratic policy that could reinforce notions that the government can’t do anything right.

David Shor’s entire analysis is supposed to be based on rigid polling and large amounts of data. The policy prescriptions seem pretty far afield from the recommendations on messaging that Shor became famous for on Twitter (e.g., his criticizing of “defund the police” as toxic messaging for the party). Also, it doesn’t even seem to fit his logic as his preferred Child Tax Credit prescription of permanence over universality polls is substantially worse! This Morning Consult poll has support for a permanent Child Tax Credit at 35%, much worse than the difference between the means-tested program and the universal program.

David Shor posted the following graph on Twitter that displays a mix of policies that, when talked about, either increase or decrease vote share for Democrats. Talking about policies such as Medicare pharmaceutical negotiation, adding dental and vision to Medicare, anti-usury laws, social security expansion, and taxes on the wealthy tend to increase Democratic vote share. Talking about policies such as lowering the voting age to 16, affirmative action, and police budget cuts tend to decrease voter share for Democrats.

The problem with all the “talking about” messaging electoral strategies that Shor advocates for is that they’re everlasting. It isn’t just “talking” but now enacting the highest polled policies because the electoral cycle never ends. The fear of another Trump presidency, or worse, and a long wave of red filling both chambers of the U.S. bicameral delivers a perpetual cycle where progressive Democrats will never be able to enact policies that their constituents demand, only the moderates.

It is no wonder that Shor loves Obama. He worked for his re-election campaign in 2012 and routinely praises Obama’s campaign for not talking about issues such as race and immigration. Obama was also a tremendous orator and, to many progressives, a disappointment due to his very liberal 2008 campaign where he played the hopeful antecedent of the end of the Clinton wing of the party. That disappointment contributed to a decline in the Democratic party voter identification, as Democratic party identification dropped further than Republicans during the same period while the number of independents rose.

David Shor’s star example of his messaging and platform may have won two landslide elections, outperformed polls, and garnered wide-spread support of a more diverse group of constituents, but what about the other members of the Democratic party? Under Obama, the Senate went from 60 Democratic senators to 46, and the House dropped from 257 Democratic congresspersons to 188. Democratic governors lost nine states they once led. The losses by the party were the largest under any president and led to the lowest number of Democratically held elected offices since before FDR.

What is the point of great messaging if it delivers poor policies that alienate constituents from your platform? You can point to many other potential causes of the shift away from Democratic support, but the cliff occurred during Obama’s presidency when he had a mandate and the legislative capital to enact it. He failed. There is a point in Shor’s polling where I would agree to take the current popular option over an unpopular one for the greater good of helping Democrats electoral odds, but it isn’t absolute. A minor to moderate difference in polled support between two policies must include the potential long-term consequences of enacting the more poorly designed policy.

The entire popularist prescription appears quite short-sighted, ignoring the simple fact that polling can change – once unpopular policies can become popular and vice-versa. When the Affordable Care Act was enacted in 2010, polling indicating it was generally underwater, with more individuals disapproving of the policy than approving. Despite its very moderate appeal, Obamacare was unpopular for many years until rapidly rising in popularity during the Trump presidency in the wake of the GOP’s efforts to repeal the ACA. The ACA wasn’t perfect legislation, but it reduced Americans without insurance from 16% to 8.6% by Trump’s inauguration date. By the time the GOP was able to win the requisite number of house and senate seats to pass a repeal of the ACA, they failed, at least in part due to the success of the policy.

All this should lead to a more nuanced prescription than David Shor’s strict moderation. This is unfortunate because his polling injects good data into the equation of political priority. There is a place for Shor’s analytics in the Democratic party that should lead to a more complete view of what the American people want, what can get more Democrats elected, and ultimately what can help the most people. Relying on polling alone will fail as they are too imprecise to rely on entirely. There is simply no way to craft the perfectly framed question; thus, trends over time using the same question are the only reliable use of polls. The popularist model needs to balance current polling with its prescription for the future, accounting for the potential consequences of bad policy and the rewards for substantive positive change for voters.