Productivity in a Changing World seminar series: Productivity and Skills - Progress and Policy Learnings from Five Generations of Research

Sholeh Maani (00:00:01):
Kia ora, and greetings. I'm Sholeh Maani, Professor of Economics at the University of Auckland. I'd like to thank the New Zealand Treasury for the invitation to give this seminar. The title of my seminar is Productivity and Skills - what have we learned from five generations of research? The seminar is a part of the New Zealand Treasury Seminar series titled Productivity in a Changing World. Today I'm presenting an overview of the economics research on skills and productivity. Our learnings, and how they have changed our understanding of relevant policies in the skills productivity space. I'm now going to share my slides and go through this progression.
(00:00:54):
First of all, I'd like to discuss why this question of productivity and skills link matters. Then I'll go through various studies in these five generations of studies, and the learnings and policy implications of them. When we look at the National Income or GDP of various countries, we noticed that the share of labour in the production of the National Income is very significant. For example, here, we are having OECD data from 1980 to 2019. For Australia in blue, and New Zealand in orange. We see that during this time period and periods before that, somewhere between 80% to 60 some percent of National Income has been contributed to the role of people or labour's contribution.
(00:02:06):
The link between the skills and productivity of the labour share is something that is quite important and would like to learn more about that. When we think about this link, an important place in economics literature to start is the development of Human Capital Theory. For example, going back to the work of Gary Becker and many studies that have followed it. In this approach, the economics incorporation of the role of education and the skill was changed to consider education and skills as a form of capital embedded in people that can enhance productivity.
(00:02:56):
The implications of the theory where that both macroeconomics and microeconomics were adapted to incorporate this important inclusion. For example, the growth models started to include human capital as an independent and separate form of capital, which enhances productivity. And microeconomic models looked at decisions to increase education as an investment that affects productivity of individuals, productivity of firms, and also earnings resulting in the labour market.
(00:03:39):
With these changes, a number of empirical studies considered this question and then verified the changes in labour market outcomes and earnings of individuals with different education levels. For example, age earnings profiles looked at lifetime earnings of people with different levels of education. And they showed differences and increases as education increases. And then there was evidence of higher lifetime earnings with increased education, indicating impacts also on wealth accumulation and spending during retirement.
(00:04:25):
For example, this analysis here based on Statistics New Zealand latest data is in this period of this type of analysis where we have individuals' highest degrees and qualifications from having level one to three post-school qualification in this diagram, to levels four - six, but below a bachelor's degree or equivalent and postgraduate. Many analysis like this for median weekly earnings for example, have shown that at a given point in time when we look at the population, we see that there are differences and increases in median earnings with higher levels of education.
(00:05:18):
Then analysis of age earnings profiles with groups by age categories, for example, also show a similar set of results. That for every age group, we notice that there's a difference in median earnings, again, by level of education. With the highest curve here representing bachelor's degrees or higher, and the darker blue curve showing level one to three. Implications of this sort of analysis were that we can expect differences in productivity not only at a given point in time, but throughout the person's lifetime.
(00:06:03):
Studies further showed that greater job stability could be shown across the business cycle with higher levels of skill and education. For example, in this diagram, which is based on New Zealand data, again, we find that over the Global Financial Crisis, the unemployment rates by education, they're different. This sort of analysis has been shown across countries and across recessions. What we find is that during recessions, the rate of unemployment is lower the higher the educational qualifications are. In this case, for example, during the Global Financial Crisis, we had the highest levels of unemployment with the case of no school qualifications compared to the case of having a bachelor's degree or higher.
(00:07:09):
Another thing that these studies showed in these early studies as well was that rates of return to investments in education for women were higher based on the data. For example, this is showing this sort of analysis for New Zealand. Again, we have median weekly earnings on the vertical axis, and we have various levels of education, from no qualifications in yellow, to postgraduate qualifications in dark blue. We find that for women and men, both, education increases are accompanied by higher average earnings, or in this case, median weekly earnings.
(00:07:58):
But for women, the rate of return or the comparable return by having a university degree, either bachelor's or higher, compared to the alternative of not having higher degrees was much greater compared to for men. For some time, we knew that rates of return for women are higher and we expected that investments in higher education would be beneficial for women, and then we expected rates of participation in higher education to increase for women.
(00:08:39):
As a result of the findings and the interest in this question, the expectation was that participation in education is affected by the expected returns to education. A number of empirical studies of returns to education looked at finding methods that looked at these questions more precisely. What these studies showed was that there was a significant lifetime return to increased education. For example, some of the early studies in the 1990s and 2000s have found 8 to 10% return in terms of earnings per year of additional education. We expected higher returns to educational investments for women as a result of these findings and also higher participation. Another finding was the widening of the skilled, lesser skilled earnings gap with increased knowledge-based jobs. So, throughout the 1990s and the 2000s, this gap started to widen as knowledge-based jobs became more significant.
(00:10:05):
Some of the policy learnings from these studies, well, these rates of return studies, were that there was expectation of expansion of participation in higher education with returns of 8 to 10%, which compared very favourably with other investments. The expectation was that there would be a higher percentage of school-leavers who'd want to participate in higher education. As a part of that expectation, policies were formed to facilitate educational investments. And then the student loans, allowances, fee policies and so on were all a part of this learning for anticipation of higher participation rates.
(00:10:57):
One other learning from these studies was the importance of school completion for access to higher education. What was found was that when knowledge-based jobs became much more significant in a number of Western and other economies, individuals or population groups who had not had a chance to complete secondary school with qualifications were at a disadvantage in taking advantage of participation in higher education. One of the policies that also became centre stage was consideration of ways to increase school-leaving with qualifications.
(00:11:48):
The second generation of the studies questioned whether the results that we see in the link between higher education or education skills in general, and earnings as a reflection of labour markets reward or productivity, were the ability question. The question of how much of what we observe is innate, how much of it is learned or how much of it is nature or nurture. A number of studies looked at ways of answering this question.
(00:12:30):
The twins’ studies and sibling studies were among the earlier influential research that tackle this question. For example, famous twins’ studies such as the one by Ashenfelter and Krueger showed a 12 to 16% returns to educational investments even when they looked at identical twins. Siblings’ studies that also followed showed that rates of return to education were not diminished when they were controlling for the sibling or the twin conditions.
(00:13:19):
Other studies found other ways of tackling this question. For example, looking at opportunities when there were changes in school-leaving age or the ability to take part in higher education. There was a draft lottery where some people with very similar ability were by chance chosen to take on military service and others were not. This way it was possible to observe individuals who would have otherwise been similar in taking up university degrees or not.
(00:14:00):
The studies that leverage changing in the minimum school-leaving age policies also looked at this difference across people where they chose different years of education because of change in legislation rather than their own decision. These studies further found that the learning part of the educational skill and productivity is very significant, that it's not diminished significantly.
(00:14:33):
Another group of studies looked at signalling and credential effects of studies. But before I go to that, I'd like to also mention that another advancement in analysis of the ability versus the actual learning question was helped by the development of panel data. Longitudinal and panel data and panel data studies where we were able to use econometric methods like fixed effects, random effects, instrumental variable estimations and so on. So that econometrically, it helped us account for similarities and differences in people across time. Sometimes observing the same people making choices in education investments or people who were very similar otherwise. These helped, again, tackle this question of ability and unobserved factors related to individuals.
(00:15:40):
The signalling and the credential effects of studies were questioning how much of the skills productivity or skills earnings, the link that we see, is because of having a credential or a degree. How much of it is the actual certificate, how much of it is the learning? Again, the number of studies tackled this question by separating the effect of the actual years of labelling from having completed a degree or not. These studies, again, show that degree completion does matter. But again, the greatest part of the returns to education that we are observing are because of the education or the learning that was taking place.
(00:16:38):
The policy learnings in relation to ability question and signalling gave some indications. For example, the importance of degree completions and policies that have facilitated it. Having a degree completed, again, not only give signal to employers about the ability to finish a task or to finish a degree, but it also made it more profitable for individuals to invest in education.
(00:17:16):
The third generation of studies considered the heterogeneity of skill and productivity outcomes. What was observed was that not everyone has the same returns to productivity or returns to earnings as a result of having increased the skills. Again, we had known that for some time, but the group of studies started to incorporate that understanding into the analysis. For example, if I go to this diagram of the age earnings profiles we were looking at before for the two degrees, the bachelor's or higher, and also lower degrees. What we had always known was that we are looking at averages of returns on the estimated regression lines that we were looking at, but there was always some distribution around that. There was some who were, for example, earning significantly more and there were some who were earning significantly less.
(00:18:34):
These groups of studies started to consider the factors that explain these differences. "Why is it that some, for example, would be having returns that are at the 90th quantile of the distribution? Why is it that someone would be at the 25th quantile?" And the factors that contribute to that. Understanding this was important. Fine-tuning of methods, for example, quantile analysis and penalised quantile regression that we've used in some of our studies. They use panel data and also quantile analysis so that they adjust for the ability question, but also, they adjust for this heterogeneity question.
(00:19:32):
The general conclusion from these groups of the studies is that in explaining some of the variation that we see, there are different factors, but degree choices of individuals are important. So not only how many years someone has been studying, but what degree choice was made and what you could call the quality of education is important. Quality of education could refer to the actual learning that had taken place, the reputation or the standing of the institutions that were engaged in teaching, and related factors.
(00:20:21):
For policy learnings, some of what we learned from these studies was that when we consider population groups who are at a disadvantage when it comes to returns to the investments in learning, in skills and education, the choice of the field of a study is important. And also, the quality of the education that is offered is important as well. We also learned from these analyses that the lower participation rates that we observe sometimes by some population groups can be explained by differential earnings returns. So, if the returns are expected to be lower, then investments are also expected to be lower. This has been an important understanding in terms of investments and returns to education and skills.
(00:21:34):
The fourth generation of studies considered this heterogeneity or differences that you see. But took a different approach by looking at skill, education and occupation mismatches. The idea behind this group of studies is that a person has investment in education and skills. And they would possess skills that are valuable in the labour market and to the economy. But, how much of it they can use also depends on the occupational job that they hold. The match between a person's skills and the job skill requirements of the job that they're holding became the focus of this type of analysis. The observation that there are mismatches in some of this education-occupation match across the economy. For example, one type of mismatch is vertical mismatch where a person may have more or less education than what the requirement of the job is, under-education or over-education.
(00:23:04):
A different type of mismatch has been observed where there's horizontal or credential mismatch. For example, the person having a degree in architecture, being engaged in retail of a product that's not related to housing, let's say. Both of these types of mismatch are possible, and then there's been evidence that they do exist in the economy. The reason why it matters is that when we think about productivity and skills, we would expect that the highest degree of productivity would be possible if a person is able to use all of the productivity that match and align with the job skill requirements. This type of research has looked at several questions in this space to see whether there are links associated with mismatches. If so, how large are they and what are the long-term effects of them.
(00:24:14):
Starting with the vertical mismatch or the over-education and under-education, and skill mismatches. What this sort of research has done is to separate years of education of a person into the part that is required for the job that they are holding. And then other components either additional or less education in that respect. This type of model is referred to as ORU models, over-education, required education, under-education models. There are a number of studies in this space.
(00:25:01):
Right. Okay, we are right now, yes. For example, to formulate this, if a person's attained or actual years of education are shown by schooling A, and the job that they're holding has required years of education that we could show by schooling R. Then the difference between the requirements and what the person's actual earnings are would be determining whether someone is over-educated or under-educated. For example, if a person's acquired or actual skills or education schooling are greater than the required level, then the person is over-educated and then this can give us years of over-education. If it's below the required level, the person is considered to be under-educated, and the difference can be measured.
(00:26:15):
For example, what it does to the earnings models that are estimated in this space, what many studies do is that they consider log of earnings of a person at the time. For example, in this case we have individual, I, at time T. Normally in these studies on the right-hand side, we would have had either years of schooling or the highest degree that the person had. This sort of ORU model separates years of schooling into three components.
(00:26:56):
The part that is required for the job that the person is holding, the years of schooling that are additional or over-education years. For some people instead of that there would be years of education that are missing, that are under the required level. Other characteristics of the individual and individual effects and so on, across individuals and across time. So then what this sort of analysis does is that it can separate rates of return to education for required years of schooling and over-education years of schooling, for example. This can give an indication of why individuals who look like they have the same years of schooling, and the same degrees are having different earnings. And if so, how much of that is because some are in match jobs and some are in jobs that don't quite match the requirements of their level of education.
(00:28:01):
To consider New Zealand, if you want to see how we are doing in terms of how many years of education people have and also the tertiary education, this is OECD data. They're from 2019 just before the pandemic. It shows individuals with a number of doctoral degrees, master's degrees in blue, bachelor's degree in light green, and short-cycle tertiary. New Zealand is actually next to the OECD average where we have a significant number of people having tertiary education now.
(00:28:40):
The difference that we have is that we have a much higher percentage with bachelor's degrees compared to other countries in this category. For example, New Zealand in general, in terms of overall tertiary education participation is just close to Denmark, Spain, and France and close to the average. But New Zealand has a higher level of bachelor's degree completions and a smaller percentage of master's and doctoral degrees compared to other degrees. OECD data, again, this was in 2019, is reporting two types of educational mismatch. The first column would be equivalent to a horizontal credential or field of study mismatch. This is the percentage of the population that could be categorised that way.
(00:29:44):
The second column here is showing vertical mismatch. And then these next two columns are dividing over-education, and the percentage under-educated. New Zealand is close to the average for the OECD and higher in some ways. But OECD distinguished New Zealand's over-education to be less of a problem and under-education to be a greater issue for New Zealand. This would indicate different reasons such as school-leavers who have not completed with the qualification and may not have had a chance to increase their skills through other means. And other populations who could have under-education in the jobs that they're holding.
(00:30:50):
A study by the Ministry of Education in 2020 used the 2016 survey of adult skills to look at over-education and under-education in New Zealand. This is an interesting study. They looked at different levels of education from lower secondary or below to postgraduate. Individuals who are matched are represented by dark orange here, the group that is over-educated is with medium colour, and the under-educated with a lighter colour.
(00:31:35):
They found that a large proportion of individuals with a bachelor's degree are actually matched. They also found that for level four to level seven non-degree qualifications, over-education was the highest for that group. This was an interesting finding of this study. They also looked at occupations and they concluded that occupations that required a bachelor's or a higher degree, a postgraduate degree, were in general and mostly filled by people who did possess those qualifications.
(00:32:35):
Looking at all studies, they are mainly in this space of looking at the over-education, under-education or vertical effects. ORU models are used in our studies and two of our studies in particular look at the questions related to the consequences of over-education and under-education. First of all, in this sort of analysis, defining what constitutes over-education or under-education is important. There are at least four ways to measure that, and various studies have covered these. Some of the studies, including ours, have used all four methods, for example.
(00:33:23):
One of them is to use Industry Standards. New Zealand and Australia have Industry Standards of the skill requirements for different occupations and the level of educational skills that are required. Another method is to use the mode, which is the level of education that most people have in a particular occupation. The main method looks at the average level of education in an occupation plus and minus a standard error. Self-reports are reports of the individuals saying that, "I'm over-educated for the job that I'm holding," or "I feel that I'm under-educated for the job."
(00:34:11):
These four methods each show a different way of measuring the same question. Our studies find that they are highly correlated, they tend to show the same sort of results with some variation. Then for skill mismatches, it is mainly based on self-reported information for a person indicating that they feel that they have more skills than the job requires, or they have lesser skills compared to the job's expectations.
(00:34:48):
One of the questions in these studies is, "Who is more likely to be over and under-educated? What are its consequences from the firm's point of view, and from the employee's perspective and is it temporary?" There are few studies that have looked at the impact of over-education and under-education from the firm's perspective. For example, there are two studies, one based on Belgian data and one with German data, that have looked at productivity of the firm. The value added per capita as explained by the proportion of the firm's workforce, which is over-educated or under-educated.
(00:35:47):
These studies concluded that over-educated workers have different effects for the firm from under-educated workers. For example, they concluded that under-education is costly for productivity. But they concluded that firms actually benefit from over-educated workers in particular where there are higher levels of skill required, when there is uncertainty, and when there is technological knowledge intensity involved. They've concluded that when these conditions exist, or in types of occupations or industries where these conditions exist, that firms benefit from over-educated employees. One of the observations for example, was that over-educated workers were willing to learn and adapt more, which was beneficial in this sort of environment.
(00:36:54):
A number of other studies have answered this question from the perspective of the employee or labour market outcomes as reflected by earnings, quit rates, job satisfaction, a number of other factors from the point of view or expressed through information on the employee. There are more studies in this space and our study is also in this space. I think one of the reasons why there have been fewer firm level studies could be that it has been harder to find the data that belongs to the firms and can accurately control for the variables of interest in the firm.
(00:37:42):
I'm going to cover findings of a few of our research papers in this space. We have about eight papers in recent years by using New Zealand, Australia and US data, looking at various aspects of this question. I'm going to talk briefly about the motivation and findings. In this type of study, we have generally used panel data. And using the longitudinal aspects of information on the same people so that we could observe them, for example, moving from jobs where they were matched to jobs that might have been mismatched. Or when they had changes in their circumstances, and productivity or skills, and then see the outcomes as well.
(00:38:35):
In the econometric modelling and objectives, we have adjusted for adjustments for ability questions, heterogeneity, characteristics of people who may otherwise not be known through data sets. And then we have also considered questions of indigeneity. For example, factors that could be explained by a person's choices but cannot be observed or they're related to other characteristics in the model and using instrumental variables or other ways of correcting for that.
(00:39:17):
We have been interested in these studies in productivity and earnings questions as a result of mismatches, career trajectories of individuals, their job satisfaction, quit rates, and in a recent study, the effect of on-the-job training. The reason why these factors, for example, job satisfaction and quit rates, are relevant in this sort of analysis is that one argument is that people may be in these mismatch jobs because they have other criteria, or they have other choices or taste.
(00:39:59):
One of the ways that we can get at that question about whether it's a choice factor or whether they're demand side, the factors that are affecting it is through job satisfaction and quit rates. If someone is very happy to have a job even that's just a job mismatch, then it would support the idea that it's their own choice. If the group that is mismatched has less job satisfaction and they have higher quit rates, then it indicates that there must have been demand factors in the economy that have resulted in the mismatch.
(00:40:32):
In general, individuals make choices for investments in education based on their information on possible outcomes, their interests at the time of investment and so on. The jobs in different occupations are also determined separately from various factors in the economy that determine the number of jobs in those occupations. It is quite likely that the match would not exactly happen, and there would be some demand side factors where everyone who has a certain degree of qualification may not be able to have a match on it. This sort of analysis would indicate whether there are also demand side factors that are explaining this.
(00:41:23):
Two papers of ours in this space, for example, look at the incidence of over-education and under-education. Explaining who is more likely to be over-educated or under-educated, are there any characteristics that we can identify. There're three findings that stand out in these studies. One is that age is an important factor. As we have had more investment in higher education in the past few decades, in general, the younger generation is more educated or has higher years of education compared to the older population.
(00:42:13):
The two groups have had different comparative advantages. The senior workers would have the experience and the younger group would have the higher degrees. The two in the labour market are a part of this question. What we find is that younger people are more likely to be over-educated in our samples. This is both for Australia and New Zealand. Then we find that those who have general degrees are more likely to have over-education or horizontal mismatch versus a specialised subjects or majors.
(00:43:02):
We find the higher incidence of over-education among immigrants, in particular immigrants from non-English speaking backgrounds. In countries like New Zealand, Australia, Canada where we have a skilled migration policies, it is not unusual. We would expect that migrants would have a higher level of education as a result compared to the average person or the general population. What these studies of ours also show is that age of arrival matters, especially age 12. Those who have arrived after the age 12 from a non-English speaking background country are more likely to be in a job where they're over-educated. We also find that country of highest degree matters. Having a New Zealand or Australian degree, for example, the highest qualification is associated with being less likely to be over-educated.
(00:44:12):
This table is from our paper in New Zealand Economic Papers which use the Household Labour Force Survey, HLFS. In this study, for example, we were able to show percentages of the population across different occupations that were over-educated, under-educated, or had a matched or adequate level of education. That time, for example, we found that women had a lower rate of over-education compared to men. But since then, our studies as we had expected, show that now women have a higher rate of over-education.
(00:45:06):
In this study we've done with US data and looks at horizontal mismatch, we found the following findings. One was that STEM workers are 25% less likely to be horizontally mismatched compared to those who had non-STEM or general degrees. In these studies, I must mention that economics is a part of the STEM degrees. Then there was a finding on state dependence. We found that there was a 19% higher probability of mismatch compared to those who had not experienced any mismatch. It was a condition that seemed to continue over time. We found that women with a STEM field degree had about 11% lower probability of horizontal mismatch.
(00:46:13):
Then in two other studies we were looking at consequences of over-education and under-education. For example, we were very interested in testing career mobility theory, which indicated that over-education should be short term, that individuals may be over-educated for a while, then they will change jobs and they will find the matched jobs later on. We wanted to test this and then having data that follows people over 18 or 20 years, in some cases you do it 10 years. We wanted to know the trajectories of individuals who are educationally matched and mismatched and see if there's a difference. If those who are mismatched actually have, if it's temporary for them. We checked occupation and earnings advancement. In this paper we also checked quit rates within a three-year period.
(00:47:23):
For this we used the Australian data, HILDA, which is a longitudinal survey of households run by the Melbourne Institute and the Australian Bureau of Statistics. Very rich data set that included many of the variables that we needed. And we used dynamic random effects models so that changes in one time period could affect the results in the next time period, making it more realistic that way.
(00:48:02):
These papers are available online through various sites, but I just wanted to bring the attention due to a few aspects of this analysis. In this paper, we were able to separate individuals into different groups. Those who were over-educated, but they also said in addition to that they're over-skilled. Those who only were over-educated, then we had a well-matched, those who were only under-educated for example. We had this variety, six different possibilities.
(00:48:45):
The results showed that the group that is most affected and could be considered as most mismatched in this case were the group were both over-educated and over-skilled. This group. They were 5.4% of the group that we had of the employees that we had compared to about a fifth that were over-educated. For this over-educated group, about 20% of them, 19.5% said that then they had both conditions. They were both over-educated and over-skilled, and 80% were only over-educated.
(00:49:32):
The results of this study showed us that the quit rates of the over-educated and over-skilled group was much higher than the group that was matched. For this group, the quit rate within a year from the previous survey was 11.1% compared to 5.3% for the matched group. The under-educated and over-skilled group also had a higher rate of quit, indicating that the mismatch was affecting the individuals. Then the results further showed that matched workers experienced wage growth, which was significant. It was 13.7% in our analysis. This was both within the occupation or across the occupation if they change jobs.
(00:50:35):
But for the job mismatch workers who quit their jobs, their probability of occupational advancement decreased within three years. By 23% for those who were over-educated, 35% for the over-educated and over-skilled, and another 34% for the group that was under-educated and over-skilled. These results suggested that the educational mismatch that was observed was not temporary. The career theory as a conclusion was not supported by these studies. What we found in this paper was that even when workers quit their jobs, the results were not in their favour. Contrary to career theory predictions, educational mismatch was not found to be temporary. Those who quit and were mismatched experienced lower wage growth and career mobility. And retention of over-educated workers based on these results can potentially enhance productivity. These are of policy relevance.
(00:52:11):
The next paper looked at the quantile aspect of educational mismatch and the earnings effects of that. In this paper, again, we were using the HILDA data. We found that when we compared Australia native born with English-speaking background immigrants and non-English speaking background immigrants, there was a difference in rates of over-education and under-education. Given Australia's skilled migration policy, immigration policy, we would expect that immigrants would have a higher level of education.
(00:53:01):
What we found was that that was the case, but we also found that the group from non-English speaking background countries had the highest level of education and they were in jobs that were requiring higher levels of education compared to the Australia native born. Then a higher percentage of non-English speaking background immigrants were over-educated, and a lower percentage was under-educated compared to the other groups. The English-speaking background immigrants were also having a higher rate of over-education, but not to the same extent as non-English speaking background immigrants.
(00:53:47):
These results are comparing returns to years of over-education and required education for the Australian born and the non-English speaking background immigrants. The comparison of these gives an indication of what sort of magnitude of disadvantage are we talking about. What we find in these analyses is that individuals who are over-educated are receiving a positive returns in earnings both for the years of education that is required and for the years of education that are over-education years. Here we are having the quantiles as well. We can see the group that is at the lower end of the returns and the group that is at the higher end of the returns.
(00:54:41):
What this comparison shows is useful in showing that for the years of education that are required by the job, the rate of return per year of education across the quantiles is somewhere higher than 7% to 8% per year, because it's still a significant return. This higher line is for the Australian-born group with the diamond signs. Then for the years of one's education, which are over-education, the rates of return are 7 to 8%, but they're lower than the level that is for the required years of education, and in some cases they can even be higher.
(00:55:36):
For the non-English speaking group, the years of over-education are not rewarded as much, you can see, compared to the years of required education. For the years of required education returns are somewhere between 6 to 6.5% while over-education is receiving about 5%. So, it's a positive return, but it's not as high as whether the person had the match job. That can explain some earnings differentials between the groups that you observe in returns to education. From the point of view of the employee having a match job has advantages and it would give the highest returns to the person's actual years of education.
(00:56:26):
When we look at the country of the highest qualification, we find that, again, it matters whether a person has an Australian qualification or an overseas qualification. Here the comparison of the results of the study are for the two groups from English-speaking background countries and non-English speaking background countries. With an Australian qualification, we see that the returns to education for the English-speaking background group of immigrants can be somewhere from 9% to over 12%, which is significant. That's a group that is doing very well in getting a reward that is higher for years of over-education. For the non-English speaking background immigrants also, there's a group from the 50th to the 90th percentile that is having a significant return to the years of over-education. Although this could be lower than the years of required education, but it's a significant return compared to having an overseas degree which does not receive the same sort of return.
(00:57:53):
We also found that for immigrants from non-English speaking background backgrounds, language proficiency is important. For example, it can result in 15.8% higher earnings at the 50th percentile and 17.8% higher at the 90th percentile. Of the learning from these analyses was that having... I can just focus on this... that having a degree in the host country, for example, in Australia in this case, enhances the possibilities of having a match job and also the returns to each year of education in required part and also the over-education part.
(00:58:54):
The last group of studies that I'm going to talk about, we looked at further reasons why people quit if they are over-educated and over-skilled. And what could be done to retain highly educated employees at firms. How could their job satisfaction be affected, and how can the retention increase? That was the motivation of our analysis since we had also found that retention of over-educated employees along with others, would enhance productivity. In this study, we still had the same categories of over-educated and over-skilled. Only over-educated and only under-educated, and this is our matched group. What we found was that the group that was over-educated and over-skilled had had a higher rate of quit compared to the well-matched group, and that was the group that we were concerned about most.
(01:00:11):
We looked at job satisfaction levels of these employees before they quit their jobs in the survey before the quit had happened. We found what we had expected that the group that was over-educated and over-skilled had significantly lower job satisfaction. For example, job satisfaction of the highest level is 8, 9, and 10. This group had a significantly lower rating at those levels compared to the group that was well-matched that had had a very significant group at this higher levels of job satisfaction. This group also had a higher level of very low job satisfaction. The same was true for only over-educated group, but the effect was more significant for this group.
(01:01:19):
What we found was that over-education alone or accompanied by skill under-utilisation, in combination with lower job satisfaction, increased incidences of job quitting. But on-the-job training decreased the likelihood of their quitting. And over-educated and over-skilled workers had close to a 12% greater likelihood of obtaining on-the-job training. Initially we thought that this was counterintuitive because we thought that, "Well, if someone is already over-educated and over-skilled, would they be more likely to receive on-the-job training?" But what we found was very clearly that this group is very willing to learn. They're more likely to receive on-the-job training, and then that increases their job satisfaction and retention with their employment. And then it was resulting in higher career mobility later on. What we found was that over-educated and over-skilled employees may stay with their present employers if they had a chance to have on-the-job training and have a higher job satisfaction.
(01:02:51):
Looking at Australia's government's financial support for training, their industry standard requirements and established training providers, we find that all of those facilitate on-the-job training. New Zealand also has a significant participation in on-the-job training. For example, 6 out of 10 employees in New Zealand undertook work-related training in the last two months. They know... About half had done a course or study that was paid by their employer, according to Statistics New Zealand latest information. New Zealand workers have high rates of participation in work-related education and training in comparison with other OECD countries. But participation rates vary across the workforce. For example, those in more high skilled jobs and more professional degrees, they are more likely to take on on-the-job training.
(01:04:02):
This brings us to the fifth topic as part of this seminar. That's the fact that now we are, again, at an important time where changes are happening in the skills and productivity area. We are very aware that with generative AI, we are at a special point in time. It is definitely a change that is a market disruptor. It is going to shape markets in ways that we had not experienced to the extent that is likely. It is going to affect skills and productivity in various ways.
(01:04:52):
What we do know is that there are many changes happening, but we do not know the exact capability of generative AI at this time. So, it's not possible to know with certainty what the exact effect is going to be. But what we do know is that there will be the importance of adaptability across the labour force in terms of the changes that are happening at this point in time. Two important factors are the productivity gain and the job displacement effects of generative AI.
(01:05:37):
It is very much expected that there's going to be productivity gains across several occupations with generative AI. That can enhance the productivity of current skills of individuals at rates that we may not have expected previously. But there's also the job displacement effect of generative AI. Two are significant and important. We remember the time when with the browser and also with email, the cost of sending information or receiving information decreased significantly. And then we also saw some of the effects and the changes that happened in job displacement, but also productivity gains. This is definitely a change in time at this point, and the exact degree of what the effects are going to be are going to become more manifest in the next few years.
(01:06:51):
What we do know is that when there is a difference in innovation. Whether it is a product innovation or whether it is a procedure innovation, affects the job displacement and productivity gain effects. When there have been product innovations, usually new jobs compared to the displacement have been more significant. When there have been procedure innovations, quite often the displacement effect has been the greater effect. For example, we have seen what has happened with post offices, with book printers and a number of other areas when these changes have happened. But we have also seen the productivity gain in a number of areas when the procedure innovation has happened.
(01:07:54):
At this point in time, generative AI has more features of a procedure innovation rather than a product innovation. So, the job displacement effects are likely. But what will be important would be the importance of adaptability of both firms and individuals because we still don't know the exact capability of AI and what is possible. One of the things, we have already seen a number of firms following the uptake of generative AI. For example, in call centres and so on, and there are more examples around. But what I also hope is that firms and organisations do not rush, because we have not yet experienced what is possible in terms of productivity gains with the workforce. In particular in jobs where the knowledge base has been an important factor.
(01:09:08):
Another effect is that the existence and the progress of generative AI is likely to change some of the ways in which the learning aspects of skills, and also the use of it will change. But the link between the two, if anything, will still become stronger. What we expect is that with generative AI, we'll see a greater productivity that will complement skill and knowledge-based information. There are also parallels in other types of production, including manufacturing and industry, which can also be affected.
(01:10:02):
In conclusions, five generations of study across a few decades have established the link between skill and productivity that is strong. Recent studies still show returns to education and the skills that vary between 4 to, 10, 12% depending on the type of degree and education gain. We expect that some educational mismatch will always be expected. A part of this is because as we have been having more and more participation in higher education, we expect that our younger populations will have higher levels of education compared to the existing. But we also expect that there will be some mismatch because there will be some differences between the types of jobs that are available at any point in an economy and the number of people who have degrees in those areas.
(01:11:14):
As a policy possibility, for example, we often report rates of return to education for school-leavers to have an idea about what to expect in terms of earning returns with different majors for different degrees. What, for example, is possible is to also report match percentages by degree. Because if you look at unemployment rates, for example, that mask some of the questions about whether a person is mismatched or not. So, it's possible for graduates to have jobs where they may have a match or it may not be, it might be a mismatch. So that gives additional information about the degrees at any point in time.
(01:12:10):
What we have also learned is that over-education results in earnings that are higher than not having a degree. So, in that sense, it is beneficial to employers and it's also beneficial to the employee. But the results and returns to years of over-education, as we have seen, are lower than if a person could have a match job. On under-education, it continues to be a concern for a number of countries. One of the ways in which it can be addressed is through school-leavers and options for further educational learning. New Zealand does have a number of opportunities for levels of education beyond school to level seven.
(01:13:07):
With the ageing population for New Zealand and the rate at which technology, including generative AI is changing, continued learning and lifelong learning and on-the-job training are continuing to be important. In our analysis we have shown that in fact, on-the-job training and lifetime learning can also help alleviate some of the issues with rates of over-educated workers, and in creating more stability of employees at firms.
(01:13:51):
This brings me to the end of the presentation today. The papers that I've discussed are all available online. If you have questions or comments, you're very welcome to contact me. My email is here. Thank you for listening.