We address the dynamic process of accumulating education and income generation across 396 Norwegian municipalities during the past four decades. The overall education level is high, and primary and secondary education are compulsory, but tertiary education has been varying significantly between regions. The education level is measured by the share of the grown up population with tertiary education. The quality of the tertiary education realistically is similar across regions. This is an important advantage compared to cross-country studies as education is a more homogenous commodity within a country. We are able to establish a measure of regional income per capita for the period 1972-2008 based on personal income from tax data. The income concept basically covers wage income.
Econometric analyses of long-term education effects face serious methodological challenges of endogeneity, in particular in regions with high mobility of people with higher education. It is difficult to find good instruments to predict the dynamics of education over the long period studied here. We therefore turn to distribution analysis and study the development of the entire cross-regional distribution of income per capita and education level. The analysis concentrates on patterns of income transitions and relations to the accumulation of education levels in the regions. Compared to the existing literature this is an alternative method to analyze the role of education. The investigation of systematic patterns between tertiary education and income transitions offers information of the plausibility of causal effects.
While we observe that highly educated concentrate in cities, the overall distribution dynamics is convergence of the education level across regions during this period. Kernel density estimates of education levels and estimated Markov chains show convergence of education levels over time. The share of the population with tertiary education has increased more in the periphery. The emergence of smart cities has been dominated by catching up from behind. The distribution analysis also shows convincing income convergence. The kernel density function of income levels is narrowing over time and first order Markov chains have ergodic distribution with single peak. However, the movements in the income distribution are unrelated to the accumulation of education. The hypothesis of equal income transition probabilities across subgroups of regions with different increases in education cannot be rejected. Catching up from low income is not driven by education and income growth has not taken off in cities with increasing education level. Overall we conclude that growth effects of human capital and knowledge spillovers are limited in Norway.