Unlocking Rural Innovation

Annex C. Impact of innovation, by regional characteristics

Table A C.1. Innovation and outcomes in rural versus more densely populated areas, fixed effects regressions on relatively rural regions (>= 75th percentile of degree of rurality)

Impact of ratio of patents to labour force on employment, productivity, household income, growth in value-added per worker and the Gini index, 2000-19

	(1)	(2)	(3)	(4)	(5)	(6)
Variable	First-step	Employment (log)	Productivity	Household income (log)	Growth in value-added per work	Gini
Ratio of patents to labour force, per 1 000		2.099***	0.913***	0.859***	0.220**	0.111***
		(0.215)	(0.244)	(0.189)	(0.096)	(0.010)
Productivity growth (1y lag)	0.006
	(0.006)
HH real income (1y lag)	0.000***
	(0.000)
Share of educated workers (1y lag)	0.000
	(0.000)
Elderly dependency ratio (1y lag)	0.002***
	(0.000)
Population density (1y lag)	-0.000
	(0.000)
Population density growth (1y lag)	0.000***
	(0.000)
Gender difference in labour market participation rate (1y lag)	-0.014
	(0.011)
Constant	-0.035***
	(0.009)
Observations	3 768	3 766	3 766	2 929	3 766	3 766
R-squared	0.107	0.440	0.434	0.390	0.204	-0.201
Number of clusters	221	221	221	169	221	221
F-test	60.90	166.4	124.1	86.67	40.48	29.46
Standard errors	Fixed effects	Fixed effects	Fixed effects	Fixed effects	Fixed effects	Fixed effects

Note: Predominantly rural regions refer to regions that are characterised as regions above the 25^th percentile of the TL2 rurality index. More densely populated areas are TL2 regions with a degree of rurality that is less than the 25^th percentile. Values include linear projections for years with missing values. The first-step estimation model is a fixed effects model on the level of TL2 with lagging independent variables. The second-step estimation includes controls for sectoral employment and value-added per worker. Controls include lags and shares in gross value added and employment in each major NACE sector. Regression is a two-stage least-squared fixed effects model. F-tests are all statistically significant to the 0.001 level. Standard errors are in parentheses.

*** p<0.01, ** p<0.05, * p<0.1.

Source: OECD Regional Demography (database) (OECD[5]).

Table A C.2. Innovation and outcomes in rural versus more densely populated areas, fixed effects regression on more densely populated regions (< 25th percentile of degree of rurality)

Impact of ratio of patents to labour force on employment, productivity, household income, growth in value-added per worker and the Gini index, 2000-19

	(7)	(8)	(9)	(10)	(11)	(12)
Variables	First-step	Employment (log)	Productivity	HH income (log)	Growth in VAPW	Gini
Ratio of patents to labour force, per 1 000		0.380***	0.539***	0.296*	0.126**	0.027***
		(0.094)	(0.140)	(0.156)	(0.059)	(0.007)
Productivity growth (1y lag)	0.019
	(0.043)
HH real income (1y lag)	0.000***
	(0.000)
Share of educated workers (1y lag)	0.000
	(0.001)
Elderly dependency ratio (1y lag)	0.002**
	(0.001)
Population density (1y lag)	0.000***
	(0.000)
Population density growth (1y lag)	-0.000
	(0.000)
Gender difference in LM participation rate (1y lag)
Constant	-0.018
	(0.029)
Observations	1 271	1 271	1 271	1 231	1 271	1 271
R-squared	0.099	0.604	0.367	0.446	0.151	0.175
Number of clusters	80	80	80	70	80	80
F-test	21.59	83.41	35.67	41.89	11.18	17.27
Standard errors	Fixed effects	Fixed effects	Fixed effects	Fixed effects	Fixed effects	Fixed effects

Note: Predominantly rural regions refer to regions that are characterised as regions above the 25^th percentile of the TL2 rurality index. More densely populated areas are TL2 regions with a degree of rurality that is less than the 25^th percentile. Values include linear projections for years with missing values. The first-step estimation model is a fixed effects model on the level of TL2 with lagging independent variables. The second-step estimation includes controls for sectoral employment and value-added per worker. Controls include lags and shares in gross value added and employment in each major NACE sector. Regression is a two-stage least-squared fixed effects model. F-tests are all statistically significant to the 0.001 level. Standard errors are in parentheses.

*** p<0.01, ** p<0.05, * p<0.1.

Source: OECD Regional Demography (database) (OECD[5]).