1. Hannon, “Poverty in the Antebellum Northeast.”
2. Hannon, “Poor Relief Policy.”
3. He concludes that this bias is small. According to the Annual Reports, however, temporary relief accounted for 43 percent of all relief expenditures and almost 73 percent of all recipients in New York in 1850 (see Hannon, “Poor Relief,” Table 5). If, as Lebergott and local evidence suggest, few temporary recipients were likely to be on the rolls in June, the majority of New York's recipients are omitted from Lebergott's denominator, but expenditures on them (over 40 percent of total expenditures) are included in his numerator. Given the available data, there are two ways to avoid this bias. One could subtract temporary relief costs from total relief costs and divide the remainder by the June I count of recipients, yielding an estimate of expenditures per full-year recipient of $36 in 1850. Alternatively, one can accept what local relief officials reported as the cost of supporting a pauper for a year in the poorhouse as I have done in Table 5, yielding an 1850 estimate of $34. The fact that these two procedures yield very similar estimates reinforces my inclination to accept the estimates of contemporary relief officials.
4. Lebergott's estimates are derived from the U. S. Census Office, A Compendium of the Ninth Census of the United States, 1872, pp. 533–34, which contains data on the “cost of annual support,” the “number of persons receiving support on June 1,” and the “number of persons supported during the year,” for each state in 1850, 1860, and 1870. The census figures on annual costs and on the number receiving support on June 1 are consistent with the data published in the Annual Reports on Statistics of the Poor, although the latter provides the number in the poorhouse on December I rather than the number receiving support on June I. The Annual Reports, however, give a much larger figure than does the census for the number receiving relief during the year, a discrepancy that is difficult to explain but which Walker in his notes to the census tables suggests might be attributable to a misinterpretation of the census question. Neither Lebergott nor I used this figure to estimate expenditures per full-year recipient or the benefit-earnings ratio, so the discrepancy in the published data cannot explain the large difference between our estimates. Lebergott's estimate of expenditures per full-year recipient is simply annual expenditures divided by the number receiving support on June I. In effect, he uses the number receiving relief on a single day as a proxy for the number of full-year recipients, which he justifies as follows (p. 61): “On a June date the figure would include the hard core of the poor regularly supported by the state.” Lebergott admits (p. 61) that his estimate may be biased upward because expenditures on partial and temporary relief are included in the numerator, whereas few temporary recipients are likely to be included in the June 1 count of recipients used in the denominator.
5. Estimated as a weighted average of what New York City's expenditures per recipient would have been under the specified assumption and actual expenditures per recipient in the rest of the state, where the weights are the fractions of total recipients in each location.