Tables for Research Railroads and the Rise of the Factory - Evidence for the United States, 1850‐1870

Table 1: Sample Statistics Panel A: Distribution of Establishments by Region Linked Manufacturing‐Transportation Database, 1850‐ 1870 State Number of Establishments Percent of Total Sample (Establishments) Northeast 6,123 71.2% Midwest 1,874 21.8 South 600 7.0 Total 8,597 100.0 Source: see text, Atack and Bateman (1999) and Atack, Haines, and Margo (2008). Sample is a repeated cross‐section of manufacturing establishments from 1850‐70 manuscript censuses of manufacturing linked to county level data on transportation infrastructure. Counties included are those that existed in the census years 1850‐70 and which had no or very minimal changes in county boundaries over the period, and which were not in the far West. To be included in the sample, establishments had to have positive values of employment, capital invested, raw materials, value of output, and value added (value of outputs – value of raw materials) > $500.00 and total employment <=1,000, and be located in one of the sample counties (N = 368 counties). Panel B: Sample Establishment Means, Factory Status and Railroad Access 1850 1850 1860 1860 1870 1870 Weight Equal Workers Equal Workers Equal Workers Factory = 1 0.124 0.675 0.142 0.727 0.145 0.772 Rail Access = 1 0.781 0.883 0.925 0.948 0.978 0.992 N (establishments) 2,874 31,446 2,897 37,602 2,826 40,018 Unit of observation is the manufacturing establishment. Factory = 1 if number of employees ≥ 16. Rail Access = 1 if railroad passed through county boundary as of year shown. Equal: each establishment counts equally. Workers: establishments are weighted by reported employment. In 1850 and 1860, reported employment is the sum of male and female employees; in 1870 reported employment is the sum of children, female, and adult male employees. Table 2: Differences‐in‐Differences Regression Estimates of the Effect of Rail Access on Factory Status Column 4, Add State and Water Weighted by Transportation Employment Year Interactions 1850, 1860 0.021 0.015 0.046* 0.092** (0.018) (0.018) (0.021) (0.053) 1860, 1870 0.020 0.024 0.039 (0.027) (0.025) (0.027) 1850, 1870 0.023 0.017 0.035 (0.020) (0.019) (0.021) 1850, 1860, 0.022 0.019 0.038* 0.087* 1870 (0.015) (0.014) (0.015) (0.041) Source: see notes to Table 1. Figures shown are coefficients of rail access dummy variable (=1 if railroad Sample County and Year Fixed Effects Add Establishment Urban Status and Industry line passes through county boundary, 0 otherwise) on factory status (=1 if number of workers ≥ 16). Standard errors, shown in parentheses, are clustered by county. Industry: dummy variables for 3‐digit SIC codes. Urban = 1 if establishment located in incorporated village, town, or city of population 2,500 or more. State: dummy variables for state; West Virginia observations are recoded as Virginia for consistent classification over time. Great Lakes = 1 if county bordered on one of the Great Lakes. Canal = 1 if a canal passed through the county as of 1850. River = 1 if a navigable river passed through the county boundary as of 1850. Ocean =1 if county bordered on ocean as of 1850. All regressions include year dummies. In columns 4 and 5, state and water transportation dummies interacted with row‐ appropriate time dummies are included. Standard errors are clustered by county. *: significant at the 5 percent level. **: significant at 10 percent level. Table 3: Instrument Variables Estimation of the Effects of Rail Access on Factory Status in 1850 Sample First Stage 2SLS Congressional Survey IV Midwest 0.472* (0.092) 0.076 (0.064) 0.541 Port IV Northeast, South 0.248* (0.072) 0.158 (0.124) 0.334 Significance Level, 2SLS – DID (β = 0.038) Number of 427 2,435 establishments Congressional Survey IV = 1 if county lies on a straight line connecting beginning and endpoints (counties) of a congressional railroad survey; see text and Atack, Haines, and Margo (2008). Port IV = 1 if county lies on a straight line between urban area of population 2,500 or more in 1820 and nearest major port; see Atack, Haines, and Margo (2008). The congressional survey IV estimation is restricted to sample counties located in the Midwest. The Port IV estimation is restricted to sample counties located in the Northeast and South. First Stage: coefficient of rail access in 1850 on instrument. 2SLS: coefficient of factory status on predicted rail access using first stage regression. Column #1 (Midwest) specification includes dummies for urban status, 3‐digit SIC industry code, canal, river, Great Lakes, and state. Column #2 (Northeast, South) specification includes the same dummies plus a dummy for ocean counties. Significance level: significance level of test of the difference between the 2SLS coefficient and DID coefficient (β = 0.038) from Table 2, row 5, column 4. Table 4: Explaining the Change in Factory Status, 1850‐1870: The Role of Rail Access Weights: Equal Workers Actual Change in 0.197 0.109 Rail Access, 1850‐ 70 Predicted Change 0.007 0.009 in Factory Status, 1850‐70 Actual Change in 0.021 0.097 Factory Status Percent Explained 33.3% 9.3% (1 – Row 2/Row 3) x 100% Change in Rail Access: see Panel B, Table 1. Predicted Change in Factory Status: Actual Change in Rail Access x regression coefficient, column 4 or 5, row 5, Table 2. Actual Change in Factory Status: see Panel B, Table 1. Equal: establishments weighted equally. Workers: establishments weighted by number of employees.