Lecture3 - Industrialization Regression Multiple Regression...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Industrialization Regression Multiple Regression Rise of Factory System The Early Factory System American Economic History University of California, Berkeley Department of Economics September 2, 2010 Econ 113 (UC Berkeley) Lecture 3 9/2/2010 0 / 34 Industrialization Regression Multiple Regression Rise of Factory System Today's Agenda: Industrialization, 19th century U.S. economic expansion: occurred in both manufacturing and agriculture Why did the Northeast industrialize before the South? Women and children were less productive in Northern agriculture They helped expand Northern manufacturing firms and industrialize Was the factory system more efficient than artisanal shops? Growth Accounting provides a useful theoretical tool Instead of imposing , we will estimate with data using regression Econ 113 (UC Berkeley) Lecture 3 9/2/2010 1 / 34 Industrialization Regression Multiple Regression Rise of Factory System Women and children helped Northeast industrialize Econ 113 (UC Berkeley) Lecture 3 9/2/2010 2 / 34 Industrialization Regression Multiple Regression Rise of Factory System Factor endowments differed in North and South South: conducive to growing sugar, cotton, and tobacco Required little skill or supervision Men and women were almost equally productive Children were especially productive in cotton picking North: conducive to grains wheat, hay, and dairy Women and children were less productive than men Wages: measures productivity Relative female wage indicates their relative productivity Econ 113 (UC Berkeley) Lecture 3 9/2/2010 3 / 34 Industrialization Regression Multiple Regression Rise of Factory System Relative female and child wages greater in South Source: Goldin and Sokoloff (1984). Econ 113 (UC Berkeley) Lecture 3 9/2/2010 4 / 34 Industrialization Regression Multiple Regression Rise of Factory System Industrialization and urbanization in 1850 Econ 113 (UC Berkeley) Lecture 3 9/2/2010 5 / 34 Industrialization Regression Multiple Regression Rise of Factory System Rise of manufacturing in North: 1820-1850 Source: Goldin and Sokoloff (1982). (also, next 2 slides) Econ 113 (UC Berkeley) Lecture 3 9/2/2010 6 / 34 Industrialization Regression Multiple Regression Rise of Factory System More women and children in manufacturing: peaking 1830s Econ 113 (UC Berkeley) Lecture 3 9/2/2010 7 / 34 Industrialization Regression Multiple Regression Rise of Factory System Relative female wages increase in Northern manufacturing Econ 113 (UC Berkeley) Lecture 3 9/2/2010 8 / 34 Industrialization Regression Multiple Regression Rise of Factory System Summary: Women and children helped North industrialize Women and children were less productive in grains than cotton and sugar Relative female wage higher in Southern agriculture North industrialized from 1820 to 1850, and perhaps earlier Most of manufacturing was sufficiently low-skilled Women and children increasingly entered manufacturing Relative female wage in manufacturing increased Econ 113 (UC Berkeley) Lecture 3 9/2/2010 9 / 34 Industrialization Regression Multiple Regression Rise of Factory System Regression: The empirical economist's best friend Throughout the course, we will ask how much "X" relates to (and possibly affects) "Y" Example: Did manufacturing firm size ("X") lead to efficiency gains ("how much") in output ("Y")? Solution: 1. Collect data on "X" and "Y" 2. Estimate a regression of "Y" on "X" (think best fit line) Econ 113 (UC Berkeley) Lecture 3 9/2/2010 10 / 34 Industrialization Regression Multiple Regression Rise of Factory System Regression examples: measures how much two things relate Suppose we want to estimate how related two things are: whether college graduates earn more than high school graduates whether the stock market changes with the weather in New York whether height is related to caloric intake With the relevant data, run the regression: a regression of earnings on years of schooling a regression of the Dow Jones Index on the average daily temperature a regression of heights on average daily calories Econ 113 (UC Berkeley) Lecture 3 9/2/2010 11 / 34 Industrialization Regression Multiple Regression Rise of Factory System Graphical Example of Regression Econ 113 (UC Berkeley) Lecture 3 9/2/2010 12 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Calorie Consumption and Height Cross-country regression circa 1800: average adult male in each country Dependent Variable: Height (mm) Coefficient t-statistic Calories Constant 0.5* (0.2) 2.5 5 400* (80) 0.5 is the coefficient on calories How large the relationship is between calories and height Positive coefficient means more calories taller 2.5 is the t-statistic on calories' coefficient Indicates how precise the relationship is Econ 113 (UC Berkeley) Lecture 3 9/2/2010 13 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Interpreting Coefficients Dependent Variable: Height (mm) Coefficient t-statistic Calories Constant 0.5* (0.2) 2.5 5 400* (80) Suppose calories in the U.S. was 2,200. What is expected average height? 0.5(2200) + 400 = 1500 mm = 59 inches Econ 113 (UC Berkeley) Lecture 3 9/2/2010 14 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Comparing Observations Dependent Variable: Height (mm) Coefficient t-statistic Calories Constant 0.5* (0.2) 2.5 5 400* (80) If the estimates were also valid for 1850 and the average U.S. adult male took in 100 more calories, by how much did the average height increase? 0.5(2300) + 400 = 1550 1550 - 1500 = 50 A faster way: 0.5(100) = 50 because height = 0.5calories Econ 113 (UC Berkeley) Lecture 3 9/2/2010 15 / 34 Industrialization Regression Multiple Regression Rise of Factory System Precision: Interpreting Standard Errors Precision: how reliable the regression estimate is Suppose the actual U.S. average height in 1800 was 54 inches Is 0.5 a good coefficient for estimating this relationship? Suppose we were talking about GDP. An error of 5 would be very good Standard error: measures error in our estimate for all observations Not always reported, but typically denoted by parentheses or brackets In this example, the standard error is 0.2 on calories Is 0.2 too big? Difficult to interpret... Econ 113 (UC Berkeley) Lecture 3 9/2/2010 16 / 34 Industrialization Regression Multiple Regression Rise of Factory System t-statistic is easier to intercept for precision t-statistic: normalization for whether the error is too big Approximate formula: t-statistic = Coe f f icient StandardError Rule of thumb is |t| 2 indicates precise estimate 95% of the time, the coefficient is not just as likely to be 0 as it is to be the number estimated Statistically significant: well estimated regression coefficient Often denoted by an asterisk 2.5 from the example denotes statistical significance Econ 113 (UC Berkeley) Lecture 3 9/2/2010 17 / 34 Industrialization Regression Multiple Regression Rise of Factory System Economic and Statistical Significance Economic significance: what does the coefficient mean In our TFP formula, an of 0.9 has a very different meaning than 0.02 Statistically significant as long as t-statistic is greater than 2 A relationship can be economically significant but not statistically significant 95% is just a rule of thumb after all Econ 113 (UC Berkeley) Lecture 3 9/2/2010 18 / 34 Industrialization Regression Multiple Regression Rise of Factory System Multiple Regression: includes controls We might want to control for additional factors in our regressions Return to the example of whether education affects wages Surely, work experience might affect wages so we would control for it Controls help the estimate be robust to alternative hypotheses Estimates from most robust regression are often most trusted The controls take on a ceteris paribus interpretation Given two people with the same work experience, how does education affect earnings? Econ 113 (UC Berkeley) Lecture 3 9/2/2010 19 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Multiple Regression Dependent Variable: Height (mm) Coefficient t-statistic Calories Former British Colony Mean Temperature (C) Constant 0.45* (0.2) 2.25 20 -1 10 200* (10) -0.5 (0.5) 300* (30) Significant at 5%. Mean Temperature: neither economically nor statistically significant Negligible impact on height given its magnitude Absolute value of its t-statistic is less than 2 Econ 113 (UC Berkeley) Lecture 3 9/2/2010 20 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Dummy Variable Dependent Variable: Height (mm) Coefficient t-statistic Calories Former British Colony Mean Temperature (C) Constant 0.45* (0.2) 2.25 20 -1 10 200* (10) -0.5 (0.5) 300* (30) Significant at 5%. Dummy variable: takes on the value 0 or 1, as in British colony variable If 0 (e.g., Brazil) , constant remains the same 300 If 1 (e.g., U.S.), coefficient is added to constant 500 Econ 113 (UC Berkeley) Lecture 3 9/2/2010 21 / 34 Industrialization Regression Multiple Regression Rise of Factory System Example Regression: Dummy Variable Dependent Variable: Height (mm) Coefficient t-statistic Calories Former British Colony Mean Temperature (C) Constant 0.45* (0.2) 2.25 20 -1 10 200* (10) -0.5 (0.5) 300* (30) Significant at 5%. Suppose U.S. average temperature was 20. What is expected average height? 0.45(2200) + 200(1) - 0.5(20) + 300 = 1480mm A non-British colony with the same data (say, Japan) is 200mm less Econ 113 (UC Berkeley) Lecture 3 9/2/2010 22 / 34 Industrialization Regression Multiple Regression Rise of Factory System Were factories efficient? Artisanal shops were displaced by factories during industrialization Facilitated by mechanization in some industries (e.g. cotton textiles) Mechanized: used machinery for all aspects of production Factory expansion entailed specialization and division of labor Larger factories often employed unskilled labor Was the transition to the factory system efficient? (Sokoloff, 1984) Did transition produce economies of scale? Econ 113 (UC Berkeley) Lecture 3 9/2/2010 23 / 34 Industrialization Regression Multiple Regression Rise of Factory System From artisanal shop... Econ 113 (UC Berkeley) Lecture 3 9/2/2010 24 / 34 Industrialization Regression Multiple Regression Rise of Factory System ...to the rise of the factory Econ 113 (UC Berkeley) Lecture 3 9/2/2010 25 / 34 Industrialization Regression Multiple Regression Rise of Factory System Manufacturing firms grew in size from 1820 to 1850 Source: Sokoloff (1984). Econ 113 (UC Berkeley) Lecture 3 9/2/2010 26 / 34 Industrialization Regression Multiple Regression Rise of Factory System Share of women and children greater in larger firms Source: Goldin and Sokoloff (1982). Econ 113 (UC Berkeley) Lecture 3 9/2/2010 27 / 34 Industrialization Regression Multiple Regression Rise of Factory System . More Efficiency or Scale Economies? Economics of scale versus efficiency in production Econ 113 (UC Berkeley) Lecture 3 9/2/2010 28 / 34 Industrialization Regression Multiple Regression Rise of Factory System Specialization is in TFP - Estimate growth accounting Production function: Y = AK a Lb Divide through by L for value added per worker: Y L Y L = AK a Lb L = AK a Lb La L1-a = A( K )a Lb-(1-a) L Let c = a + b - 1 Y = A( K )a Lc L L Take logarithm of both sides: ln( Y ) = ln(A) + a ln( K ) + c ln(L) L L Equation lends to a regression equation, where a and c are estimated ln(A) is ln(TFP), is unmeasured, and is treated as a constant Regression of log(output per labor input) on a constant, log(capital per labor input) and log(labor) Econ 113 (UC Berkeley) Lecture 3 9/2/2010 29 / 34 Industrialization Regression Multiple Regression Rise of Factory System Did larger firms experience economies of scale? Regression augmented for economies of scale for larger factories Adds a dummy variable for whether the firm has 6 or more workers Adds interaction regressor that is the dummy multiplied by log(labor) Interpretation: nonlinear output gain for factories of 6 or more? For factories of less than 6, effect of labor is c For factories of 6 or more, effect is c plus interaction coefficient Economies of scale if the interaction estimate is positive Econ 113 (UC Berkeley) Lecture 3 9/2/2010 30 / 34 Industrialization Regression Multiple Regression Rise of Factory System Were larger firms more efficient? Regression also indicates whether larger factories had a greater TFP TFP is measured by constant, ln(A) Dummy variable changes TFP estimate for larger factories Interpretation: efficiency gain for factories of 6 or more? For factories of less than 6, intercept is just the constant For factories of 6 or more, intercept is dummy estimate plus the constant TFP increases for firms of 6 or more if the dummy estimate is positive Econ 113 (UC Berkeley) Lecture 3 9/2/2010 31 / 34 Industrialization Regression Multiple Regression Rise of Factory System Larger Factories: greater TFP, no economies of scale Source: Sokoloff (1984). Econ 113 (UC Berkeley) Lecture 3 9/2/2010 32 / 34 Industrialization Regression Multiple Regression Rise of Factory System Interpreting Sokoloff (1984): Table 5 Column 4: most robust regression Economies of scale estimate: zero for larger factories Larger firms had no economies of scale: 0.236-0.240 TFP estimate: positive for factories of 6 or more Larger factories were more efficient: 3.181 + 0.402 = 3.583 Relationship is statistically significant Econ 113 (UC Berkeley) Lecture 3 9/2/2010 33 / 34 Industrialization Regression Multiple Regression Rise of Factory System Conclusion 1. Lower productivity of women and children in grains relative to cotton led Northeast to industrialize earlier than South 2. Shift from artisanal shops to non-mechanized factories led to efficiency gains, likely due to division of labor and specialization 3. Regression will come up every lecture throughout the course! Econ 113 (UC Berkeley) Lecture 3 9/2/2010 34 / 34 ...
View Full Document

This note was uploaded on 02/06/2012 for the course ECON 100A taught by Professor Woroch during the Fall '08 term at University of California, Berkeley.

Ask a homework question - tutors are online