Showing posts with label education and growth. Show all posts
Showing posts with label education and growth. Show all posts

Thursday, July 26, 2012

26/7/2012: Soft Skills - Survey of Evidence


A superb survey of the literature on soft skills (major component of human capital) published by IZA (DP No. 6580 May 2012) and written by my old prof: James J. Heckman and Tim Kautz, titled "Hard Evidence on Soft Skills". H/T to Professor Liam Delaney for spotting the paper.

The paper summarizes recent evidence on 
  • what achievement tests measure; 
  • how achievement tests relate to other measures of “cognitive ability” like IQ and grades; 
  • the important skills that achievement tests miss or mismeasure, and 
  • how much these skills matter in life. 
Core conclusions are: 
  • Achievement tests miss, or perhaps more accurately, do not adequately capture, soft skills – personality traits, goals, motivations, and preferences that are valued in the labor market, in school, and in many other domains. Incidentally, this is straight confirmation of the Nozickian bais (see here).
  • Ssoft skills predict success in life, causally produce that success, and 
  • Programs that enhance soft skills have an important place in an effective portfolio of public policies
Awesome to see this work summarized.


Sunday, September 25, 2011

25/09/2011: Returns to Education in Europe

CEPR working paper No. 8568 (link) "RETURNS TO EDUCATION ACROSS EUROPE" by Daniela Glocker and Viktor Steiner, published last week, provides some interesting (and intuitively consistent) evidence on the overall structure of the market returns to education.

Education is generally considered to be a key driver for economic growth and as such forms a specific target in many policy programmes for growth and development, such as the Europe 2020 strategy.

Since the seminal work of Gary S. Becker (starting from his 1960s papers) from an economic perspective, "the optimal level of education depends on the returns to education". Individuals invest in education if the life-time returns to education exceed the cost. These returns drive at least some of the differentials in education outcomes found across the countries. The CEPR study compares "the private returns to education across selected EU countries to explain cross-country differences in educational attainment."

The analysis is based on the 2008 panel data from the EU Statistics on Income and Living Conditions (EU-SILC) which provides comparable micro data for the member states of the European Union. The authors "estimate separate augmented Mincer-type wage equations for Austria, Germany, Italy, Sweden and UK, countries which differ significantly regarding both their education system and labour market structure."

"While the Austrian and German educational system are broadly similar and differ significantly in terms of enrollment rates in higher education from the other countries considered here, labour market outcomes in the two countries are quite distinct. Whereas Austria's unemployment rate is persistently one of the lowest in the European Union, Germany has one of the highest rates. Italy also features a relatively low enrollment rate in tertiary education, but does not have the system of vocational training prevailing in Austria and Germany which is said to be an important factor contributing to the relatively low levels of youth unemployment in these two countries. While Sweden and the United Kingdom both have relatively high enrollment rates in higher education, its financing differs significantly between these two countries and they also differ markedly in terms of labour market outcomes."

Table below - reproduced from the paper - summarizes some of the difference in outcomes across various countries.


The study estimates returns to education by country and by gender. Across countries the study finds that:
  • The direct effect of education on wages is positive and significant for all countries.
  • Education has a negative effect on unemployment duration, with effect being the strongest in Germany, and lowest for Swedish men where it is not statistically significant.
  • The probability of an unemployment spell is lower by up to 23 percentages points, for those with 16 years of education (university level) relative to those with nine years of education (basic education). The highest decrease in probability of unemployment spell is observable for German and Austrian men, and the lowest for Swedish men and women.
  • Similarly, the unconditional expected length of the cumulated unemployment declines with education. "For German men the decrease in the expected unemployment duration is the highest with six months, and the lowest for Swedish women"
  • Wage decreases due to time spend in unemployment reduce hourly wages in Germany, Austria and Italy, so that "education has an indirect effect on wages in these countries."
  • "The returns to education are positive and significant for men. Comparing the gross returns to education across countries, the UK has on average the highest returns to education with an increase in the hourly wages by 9 percent with an additional year of education. Sweden has the lowest gross returns to education with 4 percent."
  • "The effect of the expected cumulated unemployment duration is negative, but not statistically significant for Sweden and the UK. Although the level of schooling has a significant effect on the cumulated unemployment duration in the UK, the expected cumulated unemployment duration itself has not a significant effect on wages. The indirect effect of education on wages through the channel of the cumulated unemployment duration is the highest for Germany."
  • Focusing on the net returns broadly confirms the above results for gross (pre-tax) returns. "A slight change occurs when comparing Austria and Germany. While Austria has slightly higher gross returns (7.2 percent compared to 7 percent), Germany has with 6 percent 0.2 percentage points higher net returns. Looking how the returns to education change when comparing gross and net hourly wages, the UK has, on average, the highest reduction, i.e. by roughly 2 percentage points. In Austria, Italy and Germany, the respective net returns are approx. one percentage point lower than the gross returns. Sweden shows the smallest change with 0.7 percentage points. Interpreting this difference between gross and net returns as the "social return to education", the UK benefits the most from a high level of education in the population."
  • "For women [data shows] significant positive returns to education as well. As for men, the cumulated unemployment duration is significant for Austria, Germany and Italy. The combined gross as well as the net returns to education is highest for UK and Austrian women with 9 percent (and 7 percent when considering the net returns). As for men, Swedish women are estimated to have the lowest returns with respect to education."
  • "Comparing the returns of education by gender across countries, [the study] finds that there are no significant gender differences in the UK. While the returns are slightly lower for women in Germany and Sweden than for their male, the opposite is true for Austria and Italy."

Monday, July 19, 2010

Economics 19/7/10: Urban growth, education & knowledge intensive services - part 2

In the previous post (here) on the issues of growth, education and knowledge intensive sectors, I showed that
  • There is a strong positive resilience in income per capita levels across urban economies, with almost 94% of variation in income per capita in 2007 being associated with the variation in income per capita found in 1999. This strong persistency in GDP levels over time implies there is only a weak (but positive) relationship between the past and the future growth rates.
  • Data also shows that there is a weak positive relationship between long term growth in education and long term growth in income per capita. Growth in education between 1999 and 2007 was able to explain just 0.54% of the overall variation in growth in income per capita across various regions.
  • However, over time, the relationship between the levels of education of the workforce and the levels income per capita is becoming stronger both in terms of education impact and the overall explanatory power as to the direct positive correlation between education and income. By 2007 over 14.5% of variation in income per capita across major urban regions was explained by variations in education, up from 7.3% in 1999. If in 1999 1% increase in the proportion of population with 3rd level education was associated with a USD336.53 increase in income per capita (PPP-adjusted), by 2007 this effect rose to USD730.92.
  • Lagged period education levels were shown to be a better determinant of income per capita than contemporaneous levels of education, which suggests that causality flows from education to growth, rather than the other way around.

Chart 7 below explores the relationship between the levels of education in the labour force and two core higher value added sectors of economy: high tech manufacturing (HTM) and knowledge intensive services (KIS).

Chart 7
Consider the blue and the red lines. More educated workforce, it seems, is negatively correlated with high tech manufacturing role in the economy. And this correlation is becoming more negative over time (with lags). In other words, an urban centre that started with highly educated workforce in 1999 is more likely to see declining share of its economic activity accruing to HTM in 2007.

This can be related to the changes in manufacturing that took place over the last two decades, with manufacturing in general becoming increasingly more capital intensive. It is also likely to be due to the fact that with greater outsourcing of core activities and greater offshoring of manufacturing, much of higher value added activities related to high tech manufacturing, such as design and development, and marketing of new products, is now classed separately as services, and geographically removed from manufacturing activities, despite being physically embodied in the value of manufactured goods.

The opposite is true of the relationship between education levels of the workforce and knowledge intensive services role in the economy, although the positive correlation here is not becoming stronger over time (orange and green lines).

Knowledge economics – at least as proxied by education levels – is about the positive role of education in services, but it is not about the links between high-tech manufacturing and education. The dumber is your workforce (in extremely simplistic terms), the higher will be the importance of HTM to your economy… or so it appears…

Chart 8
A look at contemporaneous data reported in the chart above also confirms the previous chart 7 conclusions.

What is even more interesting here are the slopes of the two relationships.

First, the negative correlation between the degree of workforce education and high-tech manufacturing (HTM) had become more negative, from -0.1032 in 1999 to -0.1633 in 2007, while the overall relationship has strengthened (R2 = 0.0836 back in 1999 to R2=0.2341 in 2007).

This relationship is very robust and shows relatively less dispersion in the data than the relationship between education and knowledge intensive sectors.

Second, the slope of the positive relationship between the degree of workforce education and knowledge intensive sectors (KIS) has become weaker over time (from 0.5961, R2 = 0.3104 to 0.3877, R2 = 0.1396).

This result is surprising. Are we hitting diminishing returns to education in terms of increasing importance of KIS in the economy? Or are we simply at the flatter end of asymptotic KIS growth curve with much of knowledge economy already in place so that new education yields lower marginal returns? Alternatively, this might suggest that education is an imperfect instrument for skills and talent and that today, skills and talent gained outside formal classrooms matter more than before.

Finally, it is also worth noting that KIS results are significantly impacted by three observations which tend to drag the slope of the relationship down somewhat. The three, however, do not appear to represent statistically significant outliers.

Another striking relationship is shown in chart 9 below. Greater importance of high-tech manufacturing in the economy is associated with lower GDP per capita, and this negative relationship is strengthening over time, both in the explanatory power and in the size of overall negative effect. This again illustrates, most likely, the growth in capital-intensity of high tech manufacturing and disembodiment of the services-related components of high-tech manufacturing (such as R&D etc).

Chart 9
Lags in data confirm the above conclusion.

Chart 10
Chart 10 shows that identical conclusions to those presented in Chart 9 are warranted when we look at the lagged structure of economy with respect to high-tech manufacturing role in overall economic activity, so that regions that started (back in 1999) with greater share of HTM in overall economy tended to have lower GDP per capita 9 years later.

Lastly, unlike High-Tech Manufacturing, Knowledge Intensive Services are strongly positively correlated with GDP per capita, as shown in chart 11 below. This relationship is true for contemporaneous correlations and for the lagged one. And it is increasing in strength (slopes) over time, as well as in statistical significance. Furthermore, between 1999 and 2007 there has been an acceleration in the strengthening of the relationship. Finally, it is worth noting that lagged role of KIS in economy is almost as strong as 2007 contemporaneous relationship, suggesting that there is significant persistence in the positive effect that KIS have on overall income per capita.

Chart 11.


So let me summarize the main results:
  1. There is a weak positive relationship between long term growth in education and long term growth in income per capita between 1999 and 2007 across various regions.
  2. Over time, the relationship between the levels of education of the workforce and the levels income per capita is becoming stronger both in terms of education impact and the overall explanatory power as to the direct positive correlation between education and income
  3. Lagged levels of education are a better determinant of income per capita than contemporaneous levels of education, which suggests that causality flows from education to growth, rather than the other way around.
  4. More educated workforce is negatively correlated with the importance of high tech manufacturing in the economy. This correlation is becoming more negative over time (with lags).
  5. The relationship between education levels of the workforce and the importance of the knowledge intensive services role in the economy is positive. This positive correlation is not increasing over time.
  6. Overall, knowledge economy – as far as it is captured by third level education – is positively linked to services, and negatively linked to high-tech manufacturing.
  7. The negative correlation between the degree of workforce education and the extent of the high-tech manufacturing (HTM) in overall economy had become even more negative between 1999 and 2007.
  8. The positive relationship between the degree of workforce education and knowledge intensive sectors (KIS) has become weaker over 1999-2007 period.
  9. Greater importance of high-tech manufacturing in the economy is associated with lower GDP per capita, and this negative relationship is strengthening over time, both in the explanatory power and in the size of overall negative effect.
  10. Regions that started (back in 1999) with greater share of HTM in overall economy tended to have lower GDP per capita 9 years later.
  11. Knowledge Intensive Services are strongly positively correlated with GDP per capita. This relationship is true for contemporaneous correlations and for the lagged one.
  12. The positive correlation between income and KIS is increasing in strength (slopes) over time, as well as in statistical significance.

Note: you won’t be able to read this anywhere else – the two blog posts on urban economies, knowledge, education and the roles of high tech manufacturing and knowledge intensive services is an exclusive, just for the readers of this blog…

Economics 19/7/10: Urban growth, education & knowledge intensive services - part 1

As promised few days ago, here are the first couple results from an interesting data set from the OECD on regional economies.

Let me first explain what I have done to data in order to derive this (and the next post) analysis:

  • Out of 350 regions defined by the OECD, I have selected 50 regions that are directly aligned and dominated by capital cities and major industrial and commerce centers.
  • Countries covered are: Austria, Belgium, Canada, Czech Republic, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Korea, the Netherlands, Norway, Slovak Republic, Spain, Sweden, UK and US.
  • I tested for, and controlled for influential outliers (in particular Bratislava and Washington DC) where their presence distorted the overall results of estimations.
  • Time horizon covered by data is years 1999-2000 and 2006-2007.
  • Where the data was available for both 1999 and 2000, 1999 data was used. Where the data was available only for 2000, this data was used with 1999 label. Where the data was available for both 2006 and 2007, data for 2007 was used. Where the data was available only for 2006, it was used with 2007 label.
  • There were only 7 occurrences where pairs of 1999 and 2000, and 20006 and 2007 data were not available.
  • In addition to the OECD original data, I computed 1999-2007 growth rates.
The first chart below precisely plots what I call here Greater Dublin and South region (which, in the case of OECD includes Cork). OECD defines two regions for Ireland: Southern region and Border, Midlands and West region. Obviously, these are proxies for our more detailed traditional regional classification – perhaps, they are a hint that a country with 4.5 million inhabitants shouldn’t really have a Byzantine system of local authorities and Napoleonic system of regions that we have. Either way, the chart provides some very striking comparisons.

Chart 1
The size of each bubble corresponds to income per capita. This is not what I am after here. Instead, focus on change between blue dots – regional positions in terms of 1999 levels of education and the share of knowledge intensive services in overall economic activity, and green dots – the same data for 2007.

First, observe that while Dublin & South were clearly no better educated than the rest of the country in 1999, their share of higher value added knowledge intensive services (KIS) was much greater back then.


Second, notice that neither part of the country was anywhere near being in the leaders group in terms of either education or in terms of knowledge intensive services back in 1999 when compared to their peers worldwide. I always said that the claimed Irish advantage in terms of educated labour force back in the 1990s was nothing more than an urban myth. We were, frankly speaking below average in terms of education back then.


Third, note how dramatic was the increase between 1999 and 2007 in the levels of education in Dublin and South, especially compared to Border, Midlands & West. Within just 8 years or so, we moved Dublin & South out of the followers or laggards pack and into the lower end of the leaders group of better educated regional economies.


Fourth, notice that BMW region was rapidly catching up with Dublin in terms of its share of KIS in the economy, closing some of the earlier gap between 1999 and 2007, although still remaining out side the leaders group of regions.


This is interesting for a number of reasons, but chiefly, it is interesting since BMW levels of education did not rise as dramatically. There are couple of things going on here, which might explain this strange result. On the one hand, low early starting position in terms of higher value added KIS in BMW region might have resulted in a more significant growth during the financial services boom years. On the other hand, there might be a diminishing return to growth in education in the labour force present in Dublin & South region, especially as lower value added construction boomed during these years. Finally, one might conjecture that with a gradual decline in manufacturing in the country, BMW region saw increased inflows of less educated, but somewhat more experienced workers into KIS activities.


These are speculative reasons, but some are supported by the evidence presented and discussed below.



Chart 2
Chart above shows that there is a strong positive resilience in income per capita levels across urban economies. This implies that future income levels are strongly correlated with past income levels. Almost 94% of variation in income per capita in 2007 is associated with the variation in income per capita found in 1999.

This means that we have to be careful directly interpreting data showing, for example, that in a specific country, such as the US, a number of cities with highly evolved economic environment to support economic growth might be underperforming in terms of actual achieved growth their less advanced counterparts. In fact, across the 350 regions defined by the OECD, there is no statistically meaningful direct relationship between urban economies growth and levels of GDP per capita, neither in 1999, nor in 2007. Rich states in 1999 might have either lower or higher income growth through 2007 and vice versa.

Chart 3 below shows that strong persistency in GDP levels over time implies there is only a weak (but positive) relationship between the past and the future growth rates.

Chart 3

Chart 4
Chart above shows that
there is also a weak positive relationship between long term growth in education and long term growth in income per capita. A 1% growth in the proportion of population with 3rd level education between 1999 and 2007 accounted for 0.13% increase in the growth rate of income per capita. Growth in education between 1999 and 2007 was able to explain just 0.54% of the overall growth in income per capita over the same period. Although this contrasts the relationship between levels of education and levels of income per capita as shown in the chart below:

Chart 5

Over time, per above chart, the relationship between the levels of education of the workforce and the levels income per capita is becoming stronger both in terms of education impact and the overall explanatory power as to the direct positive correlation between education and income. In 1999, 7.3% of variation in income per capita across major urban regions was explained by variations in education. By 2007 this has increased to over 14.5%. If in 1999 1% increase in the proportion of population with 3rd level education was associated with a USD336.53 increase in income per capita (PPP-adjusted), by 2007 this effect rose to USD730.92.

Lagged period education levels are better determinants of income per capita than contemporaneous levels of education, which suggests that causality flows from education to growth, rather than the other way around. This is illustrated in the chart below. Notice also that this is true both in terms of explanatory power (R2) and the overall impact of education (slope coefficient). At the same time, compared to 2007 data (previous chart), this relationship (lagged 1999 education to 2007 income per capita) is weaker in the overall effect of education on income, suggesting that in recent years, there has been a significant shift in the importance of education in determining income per capita.

Chart 6

I will explore some of the possible explanations for these results in the next log post on the matter, so stay tuned…