Nominal Capital-Labor Ratios in Manufacturing, 1850 and 1880

Table 1: Nominal Capital-Labor Ratios in Manufacturing, 1850 and 1880 1-5 Workers No 6-15 16-100 100+ 1-5 Workers Yes 6-15 16-100 100+ Adjustment for No No No Yes Yes Yes Entrepreneurial Labor Input? Adjustment for No No No No Yes Yes Yes Yes Working Capital 1850 Sample 5.77 5.70 5.75 5.74 5.88 6.08 6.18 6.08 Mean, Ln (K/L) 1880 Sample 6.17 6.05 6.26 6.03 6.20 6.41 6.65 6.47 Mean ∆ (1880-1850) 0.40 0.35 0.51 0.28 0.32 0.33 0.47 0.39 ∆ (1880-1850), -0.05 0.11 -0.12 0.01 0.15* 0.07 relative to 1-5 (0.08) (0.07) (0.07) (0.06) (0.05) (0.05) workers Regression -0.09 0.21* 0.28* -0.10 0.16* 0.37* Adjusted, ∆ (0.06) (0.05) (0.07) (0.04) (0.04) (0.05) (1880-1850), relative to 1-5 workers 1850, Share of 0.214 0.164 0.343 0.279 0.241 0.177 0.309 0.273 Employment in: 1880, Share of 0.141 0.139 0.335 0.385 0.158 0.152 0.321 0.369 Employment in: ∆ (Share of -0.073 -0.025 -0.008 0.106 -0.083 -0.025 0.012 0.096 Employment) Source: 1850 and 1880 Atack-Bateman national manufacturing samples; see Atack and Bateman (1999). Adjustment for entrepreneurial labor input and working capital: see text and Atack, Bateman, and Margo (2005, p. 587, footnote 7). Establishments are included in the sample if they reported positive employment (males + females >0 in 1850 and children + adult females + adult males > 0 in 1880), capital invested, outputs produced, raw materials, and value added (value of output – value of raw materials). We also deleted observations whose estimated rate of return on capital invested in either census year fell outside the 1st through 99th percentiles of the distribution of such returns (see Atack, Bateman, and Margo (2005)) as well as observations in miscellaneous manufacturing (SIC = 999) and gas works and distribution (SIC = 492). These assumptions assure compatibility with the samples analyzed in Atack, Bateman, and Margo (2005). We also exclude establishments reporting more than 1,000 workers (only a handful of establishments fall into this group). Regression adjusted changes from 1850 to 1880: the reported coefficient in each column is the coefficient on the interaction between size class of establishment (e.g. 6-15 workers, 16-100, 100+) and dummy variable for year = 1880; the regression also includes a dummy variable for year = 1880, integer values of the total number of workers hired, dummy variables for urban status (establishment located in a city or town of population 2,500 or larger), state, and 3-digit SIC industry code; 1850 and 1880 samples are pooled to estimate the regressions. Standard errors are shown in parentheses; * = significant at the 5 percent level. There are 4,905 establishments in the 1850 sample and 7,175 establishments in the 1880 sample. Table 2: Regressions of Percent Female (1850-60) or Percent Female and Child (1870-80): U.S. Manufacturing Establishments, 1850-1880 Panel A: 1850-1860 Sample Pooled Pooled 1850 Dependent Variable Percent Female Percent Female Steam Power = 1 0.012* (0.005) {0.014*} 0.003 (0.005) {0.007} 0.014 (0.010) {0.010} -0.030* (0.005) {-0.028*} -0.005 (0.004) {-0.005} 0.018 (0.013) {0.011} 0.043* (0.001) {0.042*} 0.052 [0.231] Percent Female, County Fixed Effects -0.041* (0.009) Pooled, Steam Powered Percent Percent Female, Female County Fixed Effects -0.018* (0.007) -0.003 (0.008) -0.005 (0.007) 0.040 (0.022) -0.017 (0.020) 0.051* (0.002) 0.034* (0.002) Water Power = 1 Ln (Capital/Value Added) x 10-1 Ln (# of workers) 1860 -0.011 (0.033) {-0.009} 0.023* (0.003) {0.022*} 0.034 [0.102] Pooled, Water Powered Percent Female Pooled, NonPowered Percent Female 0.027 (0.018) {0.028} 0.040* (0.002) {0.041*} 0.036 [0.390] 0.021 (0.019) {0.019} 0.049* (0.002) {0.047*} 0.062 [0.206] Mean value of 0.052 0.055 0.048 dependent [0.231] [0.219] [0.243] variable Mean value, # 9.41 9.41 9.04 9.78 18.59 9.13 7.87 of workers Adjusted R0.492 0.546 0.513 0.510 0.539 0.660 0.529 Square # of 10,122 10,122 5,039 1,144 2,646 6,332 establishments Source: 1850 and 1860 samples of manufacturing establishments, Atack and Bateman (1999). Pooled Regressions include dummies for urban status, 3-digit (SIC) industry code, year (1860), state, and state x year. Coefficients in {} are for regressions with county fixed effects rather than state fixed effects. Nonpowered establishments include observations for which the power source is not reported. Numbers in [] are mean of the dependent variable when establishments are weighted by reported employment. *: significant at 5 percent level. Panel B: 1870-1880 Sample Pooled Pooled 1870 1880 Dependent Variable Percent Female and Child Percent Female and Child Percent Female and Child, County Fixed Effects -0.023* (0.010) Percent Female and Child, County Fixed Effects -0.029* (0.008) -0.017 (0.013) -0.034* (0.010) -0.044 (0.030) -0.041* (0.023) 0.045* (0.003) 0.038* (0.002) Steam Power = 0.021* 1 (0.005) {0.024*} Water Power = -0.008 1 (0.007) {-0.010} Ln -0.017 (Capital/Value (0.016) Added) x 10-1 {-0.028} Ln (# of workers) -0.029* (0.005) {-0.026*} -0.025* (0.0070 {-0.027*} -0.029 (0.016) {-0.043*} 0.041* (0.017) Pooled, Steam Powered Percent Female and Child Pooled, Water Powered Percent Female and Child Pooled, NonPowered Percent Female and Child -0.006 (0.030) {-0.031} 0.028* (0.003) {0.028*} 0.081 [0.236] 0.015 (0.024) {0.026} 0.026* (0.030) {0.023*} 0.033 (0.200) -0.048* (0.022) {-0.052*} 0.046* (0.002) {0.046*} 0.085 [0.330] Mean of 0.077 0.077 0.077 0.077 Dependent [0.270] [0.270] [0.239] [0.286] Variable Mean of # of 12.36 12.36 12.89 12.08 27.51 10.07 7.47 workers Adjusted R0.347 0.383 0.360 0.323 0.361 0.548 0.375 Square Number of 11,084 11,084 3,885 7,199 2,323 1,464 7,208 Establishments Source: 1870 and 1880 samples of manufacturing establishments, Atack and Bateman (1999). Regressions include dummies for urban status, 3-digit (SIC) industry code, year (1880), state, and state x year. Coefficients in {} are for regressions with county fixed effects rather than state fixed effects. 1880 observations are re-weighted to correct for under-reporting of special agent establishments; see Atack, Bateman, and Margo (2004). Non-powered establishments include observations with un-reported power source. Numbers in [] are the mean of the dependent variable when establishments are weighted by reported employment. Table 3: Regressions of Reporting of Skilled Daily Wages and of Estimated Percent Unskilled: Manufacturing Establishments in 1880 Panel A: Report of Skilled Daily Wage: 1880 Manufacturing Establishments Ln (# of workers) Ln (Capital/Workers) Steam Power = 1 Water Power = 1 Artisan Wage Reported Only Artisan Wage Reported Only -0.075* (0.010) -0.052* (0.010) 0.013* (0.005) -0.165* (0.016) -0.083* (0.021) 0.245 Both Artisan and Unskilled Wage Reported 0.089* (0.011) Both Artisan and Unskilled Wage Reported 0.061* (0.011) -0.002 (0.006) 0.201* (0.017) 0.045* (0.023) 0.667 Sample Mean of 0.245 0.667 Dependent Variable Adjusted R-square 0.113 0.126 0.095 0.113 Source: 1880 Atack-Bateman manufacturing sample. Number of establishments is 7,119. Establishments are re-weighted to correct for under-sampling of special agent establishments; see text and Atack, Bateman, and Margo (2004). Panel B: Percent Unskilled Dependent Variable Percent Unskilled Steam Power = 0.130* 1 (0.019) {0.145*} Water Power = 0.076* 1 (0.027) {0.070*} Ln 0.015* (Capital/Value (0.006) Added) {0.016*} Ln (# of workers) Percent Unskilled Percent Unskilled 0.005 (0.021) {0.016} 0.032 (0.027) {0.030} 0.014* (0.006) (0.013*} 0.090* (0.006) {0.091*} 0.005 (0.021) {0.016} 0.033 (0.027) {0.030) 0.014* (0.006) {0.013} 0.090* (0.006) {0.091} Percent Unskilled, Steam Powered Establishments Percent Unskilled, Water Powered Establishments Percent Unskilled, Non-Powered Establishments 0.006 (0.014) {0.012} 0.077* (0.012) {0.085*} 0.006 (0.019) {0.043} 0.035 (0.029) {-0.01} 0.017* (0.007) {0.015} 0.096* (0.008) {0.096*} Controls for No No Yes Yes Yes Yes Operating Times? Adjusted R0.147 0.188 0.188 0.168 0.038 0.123 Square {-0.026} Sample Mean, 0.428 0.428 0.428 0.609 0.606 0.349 Dependent [0.685] [0.685] [0.685] [0.745] [0.640] {0.391} Variable Mean Value, # 11.30 11.30 11.30 30.3 3.90 6.41 of Workers Number of 4,428 4,428 4,428 796 506 3,084 Establishments Source: 1880 Atack-Bateman sample of manufacturing establishments (Atack and Bateman 1999). See Appendix A for construction of dependent variable. Establishments are re-weighted to correct for under-reporting of special agent industries; see Atack, Bateman, and Margo (2004). All regressions include dummy variables for urban status, 3-digit industry (SIC) code, and state; coefficients in { } are from regressions with county fixed effects instead of state fixed effects. Mean values shown in brackets re-weight establishments by reported employment (adjusted for under-reporting of special agent industries). Table 4: Occupation Distributions: U.S. Labor Force, Age 10 and over: 1850-1910 Panel A: Manufacturing Industries 1850 1860 1870 1880 1900 1910 White Collar 3.1% 3.2% 4.8% 4.7% 6.8% 11.9% Prof-Tech-Manager 3.0 3.1 4.2 4.0 5.2 5.6 Clerical-Sales 0.1 0.1 0.6 0.7 1.6 6.3 Skilled Blue Collar 39.4 38.5 31.8 29.2 28.7 22.8 Middle Skill 2 39.5 38.6 32.4 29.9 30.3 29.1 (Skilled Blue Collar + Clerical/Sales) Operative/Unskilled 57.5 58.3 63.4 67.8 64.5 65.4 Source: 1850 to 1910 Census IPUMS; see Appendix B. Manager: includes Proprietors and Officials. Panel B: Aggregate Economy 1850 White Collar 6.9% Professional-Technical 2.3 Manager 3.1 Clerical/Sales 1.5 Skilled Blue Collar 11.6 Operative/Unskilled/Service 28.7 Agriculture 52.7 Operator/Supervisory Farm Laborer 23.9 28.8 1860 8.3% 2.6 3.6 2.1 11.2 30.1 50.5 1870 10.6% 2.9 4.4 3.3 10.7 32.4 46.4 1880 11.6% 3.4 4.3 3.9 9.1 37.7 41.6 1900 17.1% 4.3 5.7 7.2 11.0 36.4 35.3 1910 19.7% 4.6 5.6 9.5 11.9 37.9 30.5 23.2 27.3 24.8 21.6 24.8 16.8 20.0 15.5 16.6 13.9 1860 8.3% 6.2 1870 10.6% 7.3 1880 11.6% 7.7 1900 17.1% 10.0 1910 19.7% 10.2 34.3 35.5 33.9 31.1 28.5 36.2 38.8 37.8 38.3 38.0 57.4 54.0 54.5 51.9 51.8 Source: 1850 to 1910 Census IPUMS; see Appendix B. Panel C: Skill Groups, Aggregate Economy 1850 6.9% 5.4 High Skill 1 (White Collar) High Skill 2 (Prof/Tech/Man) Middle Skill 1 (Blue Collar 35.6 + Agricultural Operator/Supervisory) Middle Skill 2 (#1 + 37.1 Clerical/Sales) % Low Skill (Oper/ 57.5 Unsk/Serv/Farm Lab) Source: Computed from Panel B. Table 5: Wages of Common Labor, Skilled Artisans, and White Collar Workers: Aggregate Time Series, 1820-1880 by Five-Year Periods and by Decades Panel A: Nominal Wage Series Common Labor Day $0.71 0.70 0.73 0.82 0.79 0.81 0.90 1.01 1.47 1.40 1.11 0.70 0.78 0.80 0.95 1.47 1.26 0.0135 (0.0011) Skilled Artisan Day $1.31 1.38 1.45 1.56 1.40 1.41 1.57 1.83 2.47 2.64 2.29 1.36 1.51 1.40 1.70 2.47 2.46 0.0127 (0.0010) Pay Period 1821-25 1826-30 1831-35 1836-40 1841-45 1846-50 1851-55 1856-60 1866-70 1871-75 1876-80 1821-30 1831-40 1841-50 1851-60 1861-70 1871-80 Coefficient of Linear Time Trend, Log Wage Regression Source: see Appendix C. Standard error of coefficient reported in parentheses. White Collar Month $33.62 35.12 35.23 43.75 42.43 44.70 51.00 52.29 71.52 74.38 72.69 34.45 39.49 43.56 51.65 71.52 73.53 0.0159 (0.0006) Panel B: Wage Ratios Artisan/Common Labor Clerk/Common Labor Clerk/Artisan 1821-30 1.94 1.89 0.97 1831-40 1.94 1.95 1.01 1841-50 1.75 2.09 1.19 1851-60 1.79 2.09 1.17 1861-70 1.68 1.87 1.11 1871-80 1.95 2.24 1.15 Source: computed from Panel A. Wages ratios for clerks assume 26 days per month (daily wage = monthly wage/26); see Margo (2000). Table 6: Occupation Distributions (in Percent): U.S. Civilian Employment, Age 16 and Over, 1920 to 2010 Panel A: Aggregate Economy Detailed Occupations White Collar Professional-Technical Manager Clerical/Sales Skilled Blue Collar (Craft) Operative/Laborer/Service Operative/Laborer Service Occupations Agricultural Occupations Farmer/Supervisory Farm Laborer Skill Groups High Skill (Prof/Tech/Manager) Middle Skill (Clerical/Sales/Farmer/Craft) Low Skill (Operative/Laborer/Farm Laborer/Service) 1920 25.5 5.6 6.7 13.1 14.1 35.5 27.3 8.2 24.9 16.3 8.6 1930 30.8 6.9 7.8 16.1 12.7 34.2 24.3 9.9 22.4 13.7 8.6 1940 31.9 7.7 8.0 16.2 11.6 39.1 27.4 11.6 17.4 11 6.4 1950 37.5 8.9 9.0 19.6 14 36.8 26.5 10.3 11.7 7.7 4.1 1960 43.3 11.9 8.9 22.6 14.3 36 24.4 11.6 6.3 4.1 2.2 1970 48.5 15.5 7.9 25.1 13.6 34.8 22.1 12.6 3.1 1.9 1.3 1980 53.8 17.5 10.4 25.9 12 31.9 19.2 12.9 2.2 1.3 0.9 1990 58.8 20.4 12.8 25.5 10.5 29.2 15.9 13.2 1.6 0.9 0.7 2000 61.8 23.4 14.2 24.2 9.8 27.1 14.1 13.0 1.2 0.6 0.6 2010 62.5 25.1 14.3 23.1 8.1 28.3 12.6 15.7 1.1 0.4 0.7 12.4 43.6 14.7 42.5 15.7 38.9 17.9 41.3 20.7 41.0 23.4 40.5 27.8 39.3 33.3 36.9 37.6 34.6 39.4 31.6 44.1 42.8 45.4 40.8 38.3 36.0 32.9 29.9 27.7 29.0 Panel B: Manufacturing Industries Occupational Groups White Collar High Skill (Prof/Tech/Manager) Clerical/Sales Skilled Blue Collar (Craft) Middle Skill (Clerical/Sales/Craft) Low Skill (Operatives/Laborers/Service) 1920 14.8 6.1 8.7 24.8 33.5 60.4 1930 19.4 8.2 11.2 23.0 34.2 57.6 1940 21.5 7.9 13.6 18.9 32.5 59.5 1950 23.5 9.7 13.9 19.6 33.4 56.9 1960 28.4 13.0 15.4 20.1 35.5 51.4 1970 30.5 15.3 15.2 19.3 34.5 50.2 1980 33.5 18.1 15.5 19.3 34.8 47.2 1990 39.3 23.9 15.4 19.0 34.4 41.7 2000 41.5 27.6 13.9 18.0 31.9 40.5 Sources: 1920 to 2000 Census of Population IPUMS and 2010 American Community Survey IPUMS. Notes: Occupation shares are reported in percent. Occupations are classified into occupational groups based on 1950 occupation codes using the consistent coding of occupations in all years into 1950 codes (the OCC1950 variable) in the IPUMS. For 1930 to 2010 the samples include all individuals age 16 or older who were employed in the civilian workforce during the reference week for the Census or American Community Survey (EMPSTAT = 1 excluding those in the armed forces) and who reported a valid occupation (OCC1950 from 0 to 970). Employed individuals (those with EMPSTAT = 1) are excluded from the sample as being members of the armed forces if they are categorized in the armed forces by the detail EMPSTAT codes (EMPSTAT = 13 in 1930 to 1950 and EMPSTAT = 14 or 15 in 1960 to 2010) or they list a military occupation (OCC1950 = 595). For 1920 the sample includes all individuals age 16 or older in the civilian labor force (LABFORCE = 2) excluding those listing military occupations (OCC50=595). The 1940 Census occupation codes do not allow one to separate accountants (a professional occupation) from bookkeepers and cashiers (a clerical occupation) with all three groups being in occupation code 210. We allocate individuals in 1940 occupation code 210 in the 1940 Census IPUMS into professionals (27.7 percent) and clerical occupations (72.3 percent) using the share of accountants among accountants, bookkeepers, and cashiers in the 1950 Census IPUMS sample. 2010 45.6 31.7 13.9 15.8 29.7 38.6 Appendix A: Construction of Percent Unskilled: 1880 Manufacturing Sample The 1880 census reports sufficient data with which to compute an estimate of the average daily wage of workers at the establishment level (see Atack, Bateman, and Margo 2004). The census also inquired about the average daily of unskilled labor and the average daily wage of “mechanics”. We use these data to construct a proxy for the percent skilled. First, for firms that reported both types of wages, we use the following equation to estimate, u, the percent skilled w = wuu + ws(1-u) Here, w is the average daily wage at the establishment level, u is the share unskilled, and s is the share of skilled blue collar. We compute the estimate of u with this equation, retaining only those observations such that u is non-negative. For firms that report the skilled wage but not the unskilled wage, we set s = 1. For firms that report the unskilled wage but not the skilled, we set s = 0. Our estimate of u is clearly biased because our procedure assumes, in effect, that all employees were production workers, which is obviously incorrect. It is straightforward to show that our estimate of u will be biased downwards. Let m be the share of non-production workers, and wm the average daily wage of non-production workers. Then the correct estimate of u is given by the equation u = (w – (wm – ws)m – ws)/(wu – ws) Whereas our estimate is u* = (w – ws)/(wu – ws) Both the numerator and denominator of the right hand side of the expression for u or u* are negative, but the numerator of the right hand side of u* will be biased towards zero, hence our estimate of u* will be smaller than the correct estimate, u. However, this bias is likely to be very small, because m was still relatively small for most establishments in 1880. The bias will be increasing in absolute value as establishment size increases, so the coefficients on size in the regressions in Panel B of Table 4 are biased downwards. The downward bias reinforces the substance of our argument, namely that the percent of operatives/unskilled was increasing in establishment size. Appendix B: Construction of Occupation Distributions In Table 4 we present estimates of the occupation distribution in manufacturing and in the overall economy from 1850 to 1910. Our estimates pertain to individuals ages 10 and over. The estimates are based on detailed adjustments to occupation distributions that are derived from the various IPUMS samples over the same period. IPUMS samples exist for every census year between 1850 and 1910, except 1890. The IPUMS occupation and industry data pertain to free persons aged 16 and over in the labor force; persons are considered to be in the labor force if they reported a gainful occupation. B1. Estimation of Occupation Distribution in Manufacturing The 1910 census was the first to ask individuals about their industry of employment as well as occupation. While there have been several attempts by scholars to produce estimates of the share of the labor force in manufacturing for the pre-1910 census years, to the best of our knowledge there have been no attempts to provide estimates of the occupation distribution within manufacturing. Our estimates make use of the imputed IPUMS variable IND1950. The IPUMS staff created this variable after observing that in the census manuscripts individuals very frequently provided information recorded in the occupation column that readily identify the industry of employment. Nineteenth century census officials were long aware of this characteristic of the data collection; indeed, this was a primary reason why the census created the separate question in 1910. The IPUMS staff has devised a set of protocols to determine industry of employment based on the information contained in the census manuscripts (IPUMS-USA, undated). Importantly, these protocols were designed to provide a series that could link up to 1910, not to answer the question addressed in this paper. While we do not believe that the imputed industry classifications are accurate enough to produce reliable estimates of the occupation distribution at, say, the 3-digit SIC code, we do believe the classification is accurate enough to broadly distinguish the manufacturing and non-manufacturing labor force. A small caveat to our estimates is that we make no adjustment for slave labor in manufacturing for 1850 or 1860 because we believe the available data to make such an adjustment is too sparse. Slaves were certainly employed in ante-bellum manufacturing; the best known example is the tobacco industry in Richmond, Virginia (Goldin 1976). Excluding slave biases downwards the share operative/unskilled in 1850 and 1860; however, the numbers are too small in the aggregate to noticeably alter the levels of the estimates, much less the trends that we observe. As in our estimates for the overall occupation distribution (see below) we begin with samples of individuals drawn from the IPUMS with LABFORCE = 2. We then retain all observations for which IND1950 indicates a manufacturing industry. Using the variable OCC1950, we classify the manufacturing labor force into white collar, skilled blue-collar, and operative/unskilled. This constitutes our preliminary distribution. We make a series of adjustments to the preliminary distribution. In summary form, the adjustments are: Re-allocation of Shoemakers and Operatives in Boots and Shoe Industry: see the discussion below. Adjustment for Female Labor, 1850: the 1850 census reported occupations for males only. We assume that the ratio of males to the total manufacturing labor force, age 16 and over by occupation group (white collar, skilled blue collar, operative/unskilled) was the same in 1850 as in 1860. Applying these ratios to the IPUMS totals by occupation group in 1850, we produce estimates of the occupation distribution in manufacturing for both genders, age 16 and over. Adjustment for Free Child Labor, Ages 10-15: for 1870-1910, we use the published census of manufacturing to compute the ratio of child workers (ages 10-15) to the sum of adult females and males. We multiply this ratio by the number of individuals in our preliminary occupation distribution; this generates an estimate of the number of child workers to be added to the total. We assume that all child workers are employed as operatives. For 1850 and 1860, we assume that the ratio of child workers to the sum of adult males and females was the same as in 1870. B2. Estimation of Overall Occupation Distribution Our estimates of the overall occupation distribution do not rely on the imputed IPUMS variable IND1950, but do make use of the IPUMS variable OCC1950, which classifies occupations . We use this variable to divide up the labor force into seven occupation groups. The first five are non-farm occupations and the remaining two are farm: Professional/Technical, Managerial/Official/Proprietor, Clerical/Sales, Skilled Blue Collar, Operative/Unskilled/Service, Farm Operator/Supervisor, and Farm Laborer. The first three occupations are the traditional subdivisions of the white collar labor force. In Panel C of table 4, we present calculations that define the “unskilled” share of the labor force to be the sum of the shares in operative/unskilled/service and farm laborer. For each IPUMS census year we begin by extracting a sample of individuals, selecting those for whom the variable LABFORCE =2 (in the labor force). All such individuals in the IPUMS are age 16 or over, and all report a gainful occupation. Using the variable OCC1950, we create a preliminary occupation distribution for the seven occupation groups listed above. We then make a series of modifications to the preliminary distribution. In summary form, these adjustments are: Adjustment for Free Female Labor Force, Age 16 and over, in 1850: the 1850 census only reported occupations for free males. We make an initial imputation of the free female labor force by occupation group by assuming the ratio “Total Free Labor Force, Age 16 and over/Male Free Labor Force, Age 16 and over” was the same in 1850 and in 1860. We then adjust the occupation group totals for females so that the overall gender composition matches Weiss’s (2006, tables Ba1-10, Ba40-49 ) estimates for 1850. Adjustment for laborers not elsewhere classified, living on farms, ages 16 and over: in the preliminary distribution, persons for whom OCC1950 = 970 (laborers, not elsewhere classified) are allocated to the operative/unskilled/service group, regardless of their residential location. However, it is widely believed by economic historians that such individuals who lived on farms were almost certainly farm laborers rather than non-farm. Therefore, the number of individuals who report OCC1950 = 970 and also FARM = 2 is subtracted from the operative/unskilled group and added to the farm laborer group. Adjustment for Child Labor, free labor force: using Weiss (2006), we compute ratios by gender of the labor force age 10-15 to age 16 and over. We apply these ratios to the IPUMS sample totals, thereby generating estimates of children workers whose occupations were not reported in the labor force. Next, for each IPUMS census year, we extract a sample of individuals, ages 10-15, computing the proportion, by gender, who were living on farms (FARM = 2). We multiply our estimated number of child workers by the share living on farms (by gender), and call the result the number of child workers employed in agriculture. We assume all such workers were farm laborers. The remainder is assigned to the operative/unskilled/service group. Adjustment for Shoemakers and Operatives in Shoe Factories, 1850-1870: according to the 1850 and 1860 occupational classifications in the IPUMS, virtually all persons who were employed in the boot and shoe industry were classified as skilled artisans (shoemakers, or OCC1950 = 488) rather than operatives in shoe factories. However, according to the AtackBateman-Weiss samples from the manufacturing census, fully two-thirds of all labor employed in boots and shoes worked in establishments of sixteen or more workers – the commonly used definition of the “factory” – as early as 1850. In the 1870 IPUMS sample, the proportion of operatives in boot and shoes is much closer to two thirds than in 1850 or 1860. Consistent with the general trend towards factories, the proportion employed in boot and shoe establishments of 16 or more workers increased between 1850 and 1880, but the trend is a gentle one, not abrupt. We believe that in 1850 and 1860 the census had difficulty in distinguishing shoemakers in artisan shops from those working as operatives in establishments; the 1870 and 1880 censuses, however, contain instructions to enumerators to pay close attention to such distinctions in manufacturing. For 1850 to 1870, therefore, we assume that 65 percent of workers age 16 and over in the boot and shoe industry were operatives, and 35 percent were skilled artisans. We adjust the preliminary occupation distribution in these years to reflect this assumption. For 1880, we assume that the occupation classification for persons in boots and shoes are correct, as indicated by OCC1950. Adjustment for Slave Occupations, 1850 and 1860: We use Weiss (2006) estimates to calculate, by gender, the ratio, “Slave Labor Force/Free Labor Force”. We apply these ratios to the adjustment number of free workers, by gender (i.e. the adjustments described above). This generates an estimate number of slave workers, to be included in our occupation distributions. We use Olson (1992, p. 139, table 8.1) to allocate our estimated number of slave workers to the occupation groups. For men this results in allocations to the skilled artisan, operative/unskilled/service, farm operator/supervisor, and farm laborer groups; for women, the allocations are to operative/unskilled/service and farm laborer. It is likely that this procedure underestimates the number of skilled artisans among male slaves because the Olson table is based on probate and plantation records and, as such, probably under-weights slaves in urban areas; however, any bias is likely to be very small, because the share of slaves in urban areas was less than 5 percent in the decade before the Civil War (Goldin 1976). Appendix C: Construction of Wage Series, 1866-1880 We provide annual estimates of nominal wages for three occupations: common labor, artisans, and white collar workers (clerks). The data source is the Reports of Persons and Articles Hired (Record Group 92, National Archives) used extensively by Margo (2000) in his construction of analogous wage series for the antebellum period. The data pertain to civilians hired at US army installations, and cover the period 1866-1880. The data were extracted (by hand) from monthly payrolls at the forts which have survived and have been deposited at the National Archives. There are approximately 17,000 wage observations available in the data set (a wage observation refers to a person month, for example, a carpenter hired at St. Louis for one or more days during the month of October in 1879). Like those for the antebellum period, the data appear to reflect labor market conditions in the labor market surrounding the fort. Wage series are produced from hedonic regressions. In these regressions, the dependent variable is the logarithm of the daily wage (if a worker is hired on a monthly basis, the daily wage is computed by dividing by an assumed 26 days of labor per month). Control variables include occupation dummies, month of the year, place of hire (or state in which the fort is located), other characteristics of the worker if known (e.g. race), and year. 33 For common labor and artisans the regressions are estimated separately by region. For white collar workers the number of observations is insufficient to estimate regional series and instead a national regression is estimated with extensive controls for location. From the year dummies we produce annual indices of nominal wages. These annual indices are then benchmarked to wage estimates from other sources (see Margo 2002, 2004, and the text of the current paper). For common labor the benchmarking is extensive (multiple years between 1866 and 1880) and the year by year values of the indices are adjusted to reflect the benchmarking. The data for common labor has previously been analyzed, and annual wage series for census regions (Northeast, Midwest, South Atlantic, and South Central) produced in Margo (2002, 2004). Except for California during the Gold Rush period (see Margo 2000) no series have been produced for the western United States. For this paper we estimated annual series of daily wages for artisans by region and an annual series for white collar workers (see above regarding the white collar regression). For common labor and artisans we aggregated the regional series using region-specific weights. To compute these weights we drew extracts from the 1870 and 1880 IPUMS samples and estimated regional shares of occupations for 1870 and 1880. We linearly interpolated the shares between 1870 and 1880. For 1866-69 we used the 1870 weights. 33 In some cases the year dummies refer to groups of years rather than single years. In such cases we linearly interpolate to produce the annual series.