Development Without Industrialization? Household Well-Being and Premature Deindustrialization

Abstract:

The effect of premature deindustrialization on the distribution of gains from growth has thus far been understudied. Using census data from eleven countries spanning five decades and shift-share analysis, I find evidence of persistent gaps in multidimensional well-being in household categories defined by employment type and urban/rural location, and a cross-country pattern of less improvement due to expansion of industrial employment over time, without adequate replacement. Taken together, these results provide evidence for a negative relationship between improvements in household well-being and premature deindustrialization on both an individual country case and in a cross-country sample.

Keywords::

• deindustrialization
• multidimensional well-being
• inequality
• economic development

JEL Classification Codes::

• I30
• I32
• O14
• O10

Notes

1 The surveys compiled by IPUMS are a combination of samples from national census bureaus. While some of the surveys and the sample size provided by IPUMS may differ, all samples are nationally representative, and all calculations are made using appropriate weights as provided by IPUMS.

2 Employment status of household head was used to assign households for data and technical reasons; the well-being score as calculated is only assignable to households, not to individuals. It should be noted that there are inherent limitations in linking households to types using household heads, given that the employment status of other members of the household do not enter into the analysis. There is also a gender aspect to this issue, as “household head” is likely to be husband/father, and so sectoral employment patterns will likely be affected by gender, including as specifically related to deindustrialization (See Greenstein and Anderson 2017).

3 These averages count each household as one unit. This process therefore does not take account of household size. While this approach makes the most sense given the nature of the measurement, it should be acknowledged that weighting households equally could introduce a bias where household size is correlated with other variables of interest. For example, poorer or rural households sometimes have larger household sizes (See Table in appendix).

4 Adjusted growth rate calculated using the formula:

$\{\frac{[\log (100-0)-\log (100-FinalValue)]}{\log (100-0)}-\frac{[\log (100-0)-\log (100-InitialValue)]}{\log (100-0)}\}/Number\text{ o}fyears$ This type of approach is commonly used for bounded variables, see for example UNICEF (2008), Degol Hailu and Raquel Tsukada (2011).

Additional information

Notes on contributors

Joshua Greenstein

Joshua Greenstein is an assistant professor of economics at Hobart and William Smith Colleges.