Poverty before the war
Poverty fell before LBJ declared war on it. Does that mean it would've kept falling without the war?
For a couple of weeks, I’d been struggling to know where to start in writing here about the trajectory of poverty in the US over the 20th and early 21st centuries. Then, conveniently, economists Richard Burkhauser (Cornell) and Kevin Corinth (AEI) released a paper on exactly that:
We compare trends in absolute poverty before (1939–1963) and after (1963–2023) the War on Poverty was declared. Our primary methodological contribution is to create a post-tax post-transfer income measure using the 1940, 1950 and 1960 Decennial Censuses through imputations of taxes and transfers as well as certain forms of market income including perquisites (Collins and Wanamaker 2022), consistent with the full income measures developed by Burkhauser et al. (2024) for subsequent years. From 1939–1963, poverty fell by 29 percentage points, with even larger declines for Black people and all children. While absolute poverty continued to fall following the War on Poverty’s declaration, the pace was no faster, even when evaluating the trends relative to a consistent initial poverty rate. Furthermore, the pre-1964 decline in poverty among working age adults and children was achieved almost completely through increases in market income, during which time only 2–3 percent of working age adults were dependent on the government for at least half of their income, compared to dependency rates of 7–15 percent from 1972–2023. In contrast to progress on absolute poverty, reductions in relative poverty were more modest from 1939–1963 and even less so since then.
(Paywall-free version available here.) Burkhauser and Corinth have been working on poverty metrics for some time, and the general takeaway from their work on this topic is that absolute poverty has fallen dramatically over the past several decades. There’s pretty broad consensus on that point: AEI’s Scott Winship on the right, the duo of Bruce Meyer (UChicago) and James X. Sullivan (Notre Dame) (who I’d call roughly center-right), and Jane Waldfogel and the team at the Columbia Center on Poverty and Social Policy (broadly center-left) have been documenting this trend for years.
In the past, papers in this literature have shown poverty plumetting largely because benefit programs have become more generous. That’s interesting, but also kind of an inconvenient position for a conservative like Burkhauser or Corinth (both alums of the the Council of Economic Advisors in Trump I) to take. It allows a center-left person like me to waltz along with a Sickos.jpg interpretation of the finding: yes indeed, the safety net has effectively reduced poverty! It is in fact very good.
The latest Burkhauser/Corinth joint stands out because it is suggesting that while the raft of post-1960 safety net programs did reduce poverty, they may not have been necessary to achieve that end: observe that poverty fell just as fast under their measure between 1939 and 1963, before the War on Poverty, as it did from 1963 to 2023, and it fell overwhelmingly in the prior period because of growth in market incomes, not in government assistance. They’re rather explicit that this is the intended point of the paper:
What is less clear is whether the War on Poverty and the modern social safety net it created were necessary to reduce poverty … The difficulty in measuring the causal impact of the War on Poverty motivates our extending poverty trends prior to when it was officially declared. While it is not possible to infer the causal effect of the War on Poverty from the pre-trend since numerous other policies and macroeconomic conditions changed at the same time, some authors have pointed to potential progress prior to 1964—especially among Black people—as evidence that reducing poverty may not have required the implementation of the modern social safety net.
When evaluating this claim, a few points occurred to me right away:
It’s strange to compare a 24 year period (1939 to 1963) to a 60 year period (1963 to 2023).
1939 was just before World War II launched the US into its period of fastest economic and wage growth in its history. By starting there and not in, say, 1929, you’re including the war and postwar booms but not the Great Depression. Market incomes were not doing great in the 1930s.
1939 is also just before Ida May Fuller of Ludlow, VT received the first-ever Social Security check on January 31, 1940 (okay, so her name didn’t occur to me right away, but the Social Security point seemed obvious).
Burkhauser and Corinth are smart guys and they have answers for some of these. Their definition of market income explicitly excludes Social Security benefits, so the market income-driven reduction in senior poverty from 1939 to 1963 they observed can’t be driven by that program rolling out. They also explicitly compare the periods 1939-1963 and 1963-1987, both 24 years, to get around point 1 above:
In Table 3, we compare progress in reducing poverty in the 24-year period 1939–1963 to progress in the 24-year period 1963–1987, in which the initial poverty rate for both time periods is 19.5 percent. Relative to a 19.5 percent poverty rate baseline in 1939, the poverty rate fell to 5.8 percent 24 years later in 1963. Relative to a 19.5 percent poverty rate baseline in 1963, the poverty rate fell to 7.8 percent 24 years later in 1987. Thus, poverty fell faster in the 24 years before the War on Poverty was declared (by 13.6 percentage points) than the 24 years afterward (by 11.6 percentage points), when assessed using a consistent initial poverty rate.
Emphasis mine. But they have a telling footnote here too: “When including health insurance as a resource, the poverty rate falls slightly faster in the 24 years after the War on Poverty was declared (by 14.5 percentage points) than the 24 years before (by 14.0 percentage points), as reported in Appendix Table A4.” (Emphasis mine again). I tend to think we should include health insurance in these measures, so “poverty fell faster before LBJ declared the war” doesn’t seem like a persuasive claim to me.
The choice to start in 1939 has a relatively anodyne explanation. Burkhauser and Corinth construct the data prior to 1963, when official US government poverty measures start, using the decennial censuses of 1940, 1950, and 1960, which had questions about income in the previous year. Thus, the data beginning in 1939. Why not include the 1920 or 1930 decennial censuses? Well, because they didn’t ask people about their incomes. 1939 is the first year you can get data for in this way.
Fair enough! Getting data on individual incomes in the 1920s and 1930s is tough; for much of that period, income tax was restricted to the very wealthy, so there’s not a lot of administrative data to look at either if you’re concerned with poverty.
That said, this comment of their kind of gives the game away in my view:
…our results cannot establish whether poverty would have fallen after 1963 even if the War on Poverty had not been declared. Growth in market incomes could still have stagnated from the late 1960s through the early 1990s due to slowed economic growth when compared to the stronger post-war growth of the 1940s and 1950s.
Well … yeah. 1939 to 1963 was a pretty golden period for the US economy. I asked my buddy Claude to overlay trends in real GNP per capita from 1939 to 1963 with those in the ensuing two 24-year periods (1963 to 1987; 1987 to 2011). I used GNP because the GDP series starts in 1947.
One of these periods is not like the other ones:
(The weird collapse about 5 years in the blue line is the combination of wartime production ending, leading real GNP to fall substantially, and returning servicemembers and the early Baby Boom boosting population.)
Per capita income grew wildly, wildly faster from 1939 to 1963 than it did in the next 24 year increments. The average annual rate of per capita income growth was 3.16% from 1939 to 1963; 2.43% from 1963 to 1987; and 1.61% from 1987 to 2011.
What that implies to me is that the poverty reductions of 1939 to 1963, based as they were in rising market incomes, were not repeatable in later periods. The US had a historically anomalous war-based economic surge, followed by a period of extremely fast growth. The French don’t call 1945 to 1975 les trentes glorieuses for nothing: the US, Europe, and Japan were growing like gangbusters. That could not be sustained, and relying on this level of income growth persisting would not have been a viable poverty reduction strategy.
So I continue to think the safety net programs rolled out post-1963 were both beneficial in terms of poverty (which Burkhauser and Corinth document) and probably necessary for continued improvements. It was a different macroeconomic climate, and for poverty reduction to continue at its previous pace required a change in how income growth was distributed within the economy. We accomplished that change in distribution through antipoverty programs.