foreign is i'd like to start by congratulating the city of trento for holding this festival i can tell you if we were to hold a festival like this in new jersey we would get nearly so many people to attend but and especially on a rainy day like today so i think it's a tribute to what high esteem people in this region hold education in what i'd like to do is talk about the economic value of education and i'll draw on research that i've done and others have done and also i'll advertise my book a little bit i have a book which collects some of the articles i've written on education called education matters the outline for my presentation is that i'll begin by presenting summarizing the economic framework for education doing large part to gary becker viewing education as an investment i'll talk about the payoff for that investment the return to education i'll talk about evidence on trends in educational attainment and student achievement and then i'll talk about how school resources in my view translate into student achievement the themes that i'll emphasize in this lecture are first there's no free lunch in education i put in parentheses usually one exception may be higher education in italy i'll focus more on elementary and secondary education and often it's claimed that there's a magic bullet out there that will solve our educational problems and my impression from the available research is that we could probably improve the performance of the education system but i don't think we should expect extraordinary results especially in a short period of time one of the reasons why i say that is i think a very important factor in how much students and adults learn is how much time they devote to learning time on task how much time they spend in school number of years of schooling that they have amount of time that they spend doing homework amount of time that they spend attending lectures and so on and it's difficult to greatly increase the amount of time that children spend studying and it's also costly to do that the third point i'll emphasize that inputs do matter inputs like the number of students per teacher remember teachers that a school has the quality of the teachers and i think the school inputs especially matter disadvantaged children then finally i'll show you some evidence that the economy is demanding more highly skilled workers i suspect this will continue into the future and that's one of the reasons why i think the topic of this entire festival is of the greatest importance so next in this slide i want to summarize the economic framework to thinking about education as an investment and i have a quote here from adam smith from the wealth of nations in 1776 which really does capture the idea of human capital gary becker of course wrote a very important book in 1964 which extended and generalized the idea of human capital with particular emphasis on education here's what adam smith wrote back in 1776 a man educated at the expense of much labor and time to any of those employments which require extraordinary dexterity and skill may be compared to one of those expensive machines the work which he learns to perform it must be expected over and above the usual wages of common labor will replace to him the whole expense of his education with at least the ordinary profits of an equally valuable capital so becker uh so excuse me adam smith as well as becker had in mind the idea that when people think about investing in education they think about how it will ride how it will raise their income and they can make judgments about investing time and resources into education versus other pursuits and in equilibrium we would expect that the payoff to investing in education to obtaining more education for learning more skills had at least as much a return as a return to investing in machinery and equipment well that's all fine at a theoretical level the question that i'm most interested in is the evidence what is the evidence on the rate of return or in lay terms the payoff to education it's been known for a very long time at least a hundred years that there's a strong correlation between people's income and the amount of education that they have now correlation does not necessarily mean causality and for a long time economists have debated what this correlation means and i'll show you a little bit of evidence on this relationship and why a correlation between income and education what i mean is that people who tend to have more education also tend to have higher income well perhaps the higher income is a result of getting that extra education that's what adam smith would have predicted or perhaps it's picking up reflecting other factors what economists call omitted variables for example perhaps people who come from wealthier families tend to get more education because their families can afford the extra education but perhaps they do well in the labor market not as a result of that education but because of their family's connections maybe they go into the family business and that's why they make higher income so unless we can control unless we can somehow adjust for the background factors it's not clear whether the payoff the education that we observe is truly a result of the education per se we're merely reflecting some other features of individuals interestingly a factor that cuts in the other direction is that when we interview people and we collect data from them they sometimes misreport their education sometimes people will forget about a couple of years or they'll exaggerate and if that's the case if we have noisy data then we might think that the correlation is actually obscured the relationship between income and education it's causing us to understate the strength of the relationship and i'll come back to that because i think in modern data sets reporting errors and mistakes in the data actually play quite an important role so let me show you a little bit of evidence on the relationship between earnings and years of schooling this is from data from the united states and this is a sample of around 30 000 men drawn from the current population survey what this diagram shows you is for each number of years of education 7 8 9 10 11 so on up to 21 years of schooling those would be the phd students the average of the logarithm of the hourly wage for people with that level of education so don't be intimidated by this just thinking of this as how average income tends to rise as you look across people with higher education and you can see a very strong relationship here between average earnings and average education in economics we like to take logarithms there's a a theory behind why logarithms might make sense here for the logarithmic earnings i think the stronger argument is that it tends to fit the data pretty well and you can see this looks pretty much like a straight line the points cluster pretty closely to the line i have a friend who likes to say that we take logarithms because it makes us seem scientific i think there's a stronger logic than that but also if we do take logarithms we can interpret the slope of this line as the proportionate increase in earnings associated with an additional year of education and because the line fits the data so well you can actually summarize the entire relationship between earnings and education with a simple number just the slope of this line and for the united states the slope of this line is currently about ten percent each additional year of education is associated with about ten percent higher earnings in the 1970s that was more like six percent so we've seen an increase in the payoff to education which i'll come back to and you've experienced a similar phenomenon here in italy more recently and you'll notice degrees do seem to matter so completing secondary school which is where it says high school grad well that's a lot better than leaving high school in 12th grade or finishing a bachelor's degree or getting a jd seems to be a little bit above the line but not tremendously above the line so you're really not doing a great injustice to the data by fitting a line through the points to summarize the overall relationship and if you were to look at the data in this way in other countries i've done it for germany for sweden others have done it for italy the fit seems to be pretty close to a straight line this is not unique to the united states now let me show you some more international evidence this is from a study which looked at several european countries and try to estimate that line the slope of that line for each of these countries and you can see somehow in the translation to italian the green arrow that pushed a little bit to the left that's italy to the right of the green arrow so for men in italy a year of education was associated with about 6.2 percent higher earnings in the mid 1990s for women which are shown in pink a year of schooling is associated with a little bit higher increase in earnings of 7.7 percent those numbers are slightly below the average for all of europe but the italian data are net of taxes these are net wages not gross wages and most of the other countries are gross wages if an adjustment is made through taxation then italy is very close to the average of all of europe in terms of the relationship between earnings and higher education so in that in essence italy looks like the bulk of europe now you'll notice the uk looks a little bit of an outlier here a higher payoff to education in the uk in general the countries that have more dispersion in their income distribution like the united states and the uk tend to have a higher payoff to additional education countries like sweden the other scandinavian countries whichever oppressed wages less variability tend to have a lower payoff to an additional year of education but one of the things i think this suggests is that across all of europe additional education is associated with higher earnings it's still an outstanding question what that means we know that people who get more education on average tend to earn more but we don't know if it's because of that education the key question in my mind is whether education and earnings uh are causally related um and a variety of strategies have been used in the literature to try to tease out the reasons behind the relationship between education and earnings to try to probe whether the higher earnings associated with additional education is causal and i would say it's a rarity in economics because almost universally this literature says yes as far as we can tell this looks like a causal relationship and i'd like to give you a flavor for the different types of strategies that have been used to address this question the earliest studies tried to statistically control for factors like intelligence as best we could measure it for family background and so on those studies tended to find that education was associated with higher earnings maybe a little bit lower than what you would see if you didn't make that adjustment but there was still quite a strong relationship another line of approach used happenstance geography people happen to be happy to live near a school they're actually more likely to go on to college or community college just using this kind of geographic variability with those studies tended to find as well was that education was associated with higher earnings i've done some work on twins compulsory schooling and comparisons across countries that i want to go into in a little bit more detail so the work on twins it might be the last festival i went to was the twinsburg twins festival in twinsburg ohio i think this is a subtle difference between italy and the united states but in the united states we have festivals about twins and here you have festivals about economics so in any event in the mid-1990s four years in a row i went to twinsburg ohio with my colleague whirly ashenfelter and a team of graduate students and we interviewed hundreds of identical twins and i have to tell you this was a unique experience i personally interviewed identical quadruplets and we had a clever idea in this in in this data collection because we asked each of the twins not only about himself or herself but also about their twin and this was important because we asked each twin how much schooling do you have and how much does your sibling have so for each person we had two measures and this is one way of overcoming the reporting errors and they didn't always agree in fact we also asked them about their father how much education does their father have since they had the same father she would think they'd all report the same level of education but the correlation was only 0.85 less than one and identical twins tend to be very similar so a lot of the differences in the education that you see are probably reporting herbs that's one of the things that we discovered actually when i interviewed the quadruplets it was kind of a data set and a microcosm because we asked each one of the quadruplets about herself there were four identical women in their early 30s and then we asked about each of her siblings so we needed four forms for each of them and after i interviewed one she went back to the group i couldn't figure out which one i interviewed the other thing that strikes me about this family was that after the parents had two children they then had quadruplets so they had six children they opened up a daycare center and i suspect they were their own largest client anyway let me show you some of the data this is from a paper that worley ashley melter and i published in the american economic review and what we did was to line up the twins and take the difference between their education and since we didn't know if twin reported the correct education or the sibling reported the correct education we took the average that was a simple way of trying to overcome the reporting problems that arise and what we found as you can see here is that there's a lot of noise around the line that we estimated a lot of scatter around the line but the twins the the the member of the pair of twins who had a higher level of schooling on average tended to have higher earnings and when we use the statistical technique to estimate the best fitting line through these points what we found was that each additional year of schooling was associated with about 12 percent higher earnings so even more than what i mentioned before which was 10 and identical twins represent a really good experiment here because they have the same genetic structure they're not clones but they are quite similar they come from the same families so there's no difference in family background which could possibly be driving the relationship that we see in the data so our study of twins and other studies of twins have also shown that earnings tend to rise with education another thing that you'll notice here is there's a great deal of dispersion around the line now i will use my pointers if you look at the point which says zero here these are identical twins who have the exact same level of schooling they don't all lie on the line there was a big difference in earnings often between the pairs of twins what that suggests is that there are factors beyond education which matter for individuals earnings education is only one of many factors some twins might have jobs which have unpleasant working conditions and they get paid more as a result of that some might have been lucky and landed a job a higher paying job some connections might have helped in some cases and not in others speaking about the role of lot i heard a story recently that jk rawlings the author of harry potter books before she published the first book she didn't have an agent so she went to the library to look for an agent she looked at several names and she looked for a name which sounded like it might be a character in her novel so she chose christopher little who must be the luckiest man in the world uh so clearly luck has some role to play in in the determination of earnings but education also matters now when we look at twins we don't know why one twin has more schooling than the other one another line of work tries to use variability in education that comes from factors that we think are random where we can explain why one group has more schooling than another so i want to describe a little bit of this work that i did with joshua anders where we looked at the impact of compulsory schooling and the logic behind this is as follows children start school depending upon whether their birthday is before or after the cut off that year and in the united states in the period that we were looking most children needed to turn a six to start first grade by january 1st if they were born december 31st they just made the cutoff and they could start if they were born january 2nd they would have to wait an entire year before they began then when you combine that with compulsory schooling laws we have something of a natural experiment compulsory schooling requires children to go to school typically until they reach their 16th birthday i understand in italy the compulsory schooling law just increased to 15. it recently increased to 15. in the united states it varies across states it's either 16 17 or 18 but in all cases it's until you reach your birthday so this sets up kind of an experiment where depending upon accident of date of birth you're required to attend school longer if you were younger when you started now i would say unless you believe in astrology i don't think date of birth has much to do with your inheritability or the background characteristics which i mentioned earlier which might affect individuals earnings irrespective of schooling so this gives us something of a natural experiment and i have to say when joshua agress through that mit and i first came up with this idea i was really surprised to see that it worked pretty well in the data so we used data from the u.s census and this looks at men who were born in the 1930s and the 1940s and we've also looked at men in the 1950s very large samples over a million people are represented in this graph and what we show you here is their average education and steadily education has been rising but those who were born in the beginning of the year january february march the first quarter of the year they tend to have a little bit lower education on average see little blips down and those who were born at the end of the year october november december they tend to get a little bit more education and you can see this more clearly if we remove the trends from the data if we deviate the ongoing trend this shows you in the solar lines the black satellites are the deviations from the trend education level for those born in the first quarter of the year and you can see they're typically negative and in fact when we look in 29 out of 30 years we found that those who were born in the first quarter of the year got a little less education than those who were born at the end of the year as best we can tell this looks like it's due to compulsory schooling i think we could rule out a lot of other explanations for one thing this only arises for levels of education below below high school if you look at the chance of getting a college degree or getting a phd you do not see a seasonal pattern the way that we see it here and there's some other evidence which i think rules out other possible explanations i remember gary becker once raised me the possibility that people might have waited till the beginning of the new year so they could get a tax deduction uh which was a primary explanation but i think it goes the wrong way since those tended to be the lower income ones anyway the next thing that we did was to look at their earnings and i have to say this really did shock me when we use the census data to look at these individuals earnings by date of birth what you see here is year of birth and we indicate the quarter of the year in which people were born you see the little blips in education translate to blips and earnings the same regular pattern arises now people tend to have higher earnings as they get older which is why you see this overall pattern but you see a large number of the blips down tend to be those in the first quarter and again i think we could rule out some other explanations so if we look just at college graduates we don't see this pattern at all we don't see these negative blips for the first quarter and if you take the ratio of earnings for those born say if you took the difference in earnings between those born in the fourth quarter of the year and the first quarter and you compare that to the difference in their education you could get a rough estimate of the value of an additional year of education kind of a way of scaling these flips that we saw in earnings in education and the best estimate we come up with is about 10 higher earnings associated with an additional year of schooling and if you think about that if our interpretation is correct it's really a remarkable result because it says that the people who are compelled to go on to school longer than they want to they're only in school because the law requires them to go to their 16th birthday they benefit in terms of getting higher earnings later on from that extra schooling which they otherwise would not have had received now that schooling might have been very difficult for them but it does seem to have a high payoff for them later on and in fact as best we can tell it looks like those people benefit more from the extra schooling than the average person does from an additional year of schooling and one of the conclusions i emphasized earlier one of the themes i told you i went back to is that education seems to be more valuable for those who come from a more disadvantaged background my interpretation of that pattern is that people from a more disadvantaged background have a much higher cost of attending school not only monetary cost of attending school monetary costs because there may be tuition when it comes to college or there's opportunity cost of the time their family may need them to work but also a higher psychological cost i think students from a more disadvantaged background find school less interesting they find it more difficult and that's one of the reasons why they tend to complete less schooling but because their costs are higher it seems that if we push them further or we can figure out a way to keep them in school longer they actually benefit from it even more than the typical person does oh i wanted to point out that this result also was anticipated by adam smith this is one of my favorite quotes from the wealth of nations adam smith said the difference between the most dissimilar characters between a philosopher and a common street porter for example seems to arise not so much from nature as from habit custom and education so smith clearly thought that education had the potential to raise people's opportunities to raise their living standards and change the types of professions they would qualify for and so on also i suspect he was poking fun out of himself since adam smith was a philosopher so he chose to use his own profession there okay now another active area of research has been to look not across individuals at education and look at how individual income relates to education but to look across countries and at the individual level we tend to think of the relationships that we're estimating as giving us information on the private payoff to education how much do you personally gain from getting additional education the social return could be greater or less than the private return the social return can be greater than the private return of people who are better educated tend to have positive spillover effects for society as a whole there's some benefit for society that they don't individually capture we call those externalities the social return could be less than the private return if all that education does is to signal who would have been a good worker anyway in that case education could become something of a rat race and even though individually it might be a good decision for people to go on and get additional education it might not be the case for society as a whole now i tend to think that the results for compulsory schooling weigh against that interpretation because if all education did was signal individuals ability their credentials if all education was was a credential who has high ability then extra education that was obtained because it was obligatory because of compulsory schooling would not convey very much information but in general it's proved difficult to distinguish between the private returns education and the social return to education based on data on individuals alone so one approach is to look across countries and say the countries that have higher education tend to have higher higher gdp per capita or higher income and the answer is certainly yes this shows you a graph where each point here is a country and on the horizontal axis its average years the schooling and the vertical axis is gdp per capita and you can see that clearly the relationship is upward sloping and it's pretty strong correlation then a question arises how does this relationship compare to what we find at the individual level suppose we started at the individual level and just aggregated up is the relationship between earnings and education at the individual level what we would predict we would find at the country level and i think it's much harder to do analysis at the country level there are so many things that differ across countries and we don't as of yet have enough data to look at the kinds of natural experiments that we've been able to use when we look at individuals like date of birth and compulsory schooling and there are no or very few countries which are like twins where we can control for very much by looking at the differences between them so i think our knowledge base is much weaker when it comes to looking across countries but i'll summarize what i found in work that i did with michael lindahl who's at stockholm university what we found is that there's a very strong correlation across countries in a point in time between earnings and education and in fact the slope of that line is a lot greater than what we find at the individual level but if we look at changes when countries increase their education level does their gdp per capita increase and that's what this graph shows you you get a much noisier relationship and if you look over short time periods five years ten years you get a fairly weak relationship between the increase in education and the growth in gdp per capita if you look over a 20-year period you actually get a pretty strong relationship and the data we have at the country level just like the data on the tunes that we collected are actually fairly noisy not all countries do a good job collecting data on educational attainment and when we look at changes you tend to get a lot of changes which i think are just mistakes in the way the data are recorded or the way we estimate average education which is not what one would expect at the country level because one of the advantages of using country level data is that you're averaging over a lot of different people yet the data is still i think quite weak and a lot of this work i think is more at its infancy than the stage of the work on research at the individual level my sense is when we look at long periods of time say over 20 year period increases in education in the country are associated with higher gdp per capita and i would go one step further and say the increase is about what you would expect from the studies at the individual level that each year of education is associated with about 10 higher earnings now that conclusion is somewhat controversial and it depends a little bit on what else one controls for in the model but i think it's hard to reject that conclusion and what that conclusion suggests is that the payoff to education that we estimate at the individual level the private payoff is probably about the same order of magnitude as the social payoff because if there were benefits to education that go beyond the individual who got the education for society as a whole that would show up when we look at the country level that would generate a stronger relationship a steeper relationship a higher payoff to education at the country level than at the individual level but my sense is it's very hard to reject that the two are the same now there is a controversy in this literature because one body of research has found that what really matters or what they've argued is what matters is not so much the change in education but the level of education that countries are starting from countries that have a higher level of education people like paul romer and then fellows have argued tend to have faster growth and they've gone on to develop theories that say things like well if you start from a high level of education you can adapt new technology from abroad and that has a benefit for society as a whole which goes beyond the individual level now in the work i've done i found that conclusion to be pretty sensitive to the type of statistical model that one estimates in particular if you look at high income countries if you look at the oecd for example you don't find that the initial level of education matters for growth that conclusion is being driven by the low-income countries moreover you might argue that the payoff to education is different in different countries that's something i showed you earlier maybe the payoff to the initial level of education is different in different countries so if you allow the starting point to have a different effect in different countries what you find is for the average country the starting point doesn't matter very much so my conclusion which is somewhat tentative and probably controversial is that you do pretty well just starting with the micro data and aggregating from there so that's the reason why i'm going to focus on trying to draw inferences from what we find at the individual level and at the individual level in both the united states italy and much of the world the payoff to education has been rising over time this shows you a very simple illustration of that conclusion this compares average earnings for those of a college degree or more to the average learnings of those who have exactly a high school degree and in the 1950s people with a college degree on average earned about 38 more than people with a high school degree by 1970 it was about 60 percent more that fell in the 1970s my phd advisor richard freeman actually wrote a book at the end of the 70s called the over-educated american based on that dip the end of that book it's little known he actually predicted that it would rebound i don't think he ever anticipated what happened in the 1980s and you can see the payoff education continued to grow in the 1990s and in the 2000s it's been pretty pretty stable italy looks a little bit different but it's certainly not the case that the payoff to education today in italy is lower than it was in the 1970s this shows you for men based on a study by brunello comey and sephora the payoff to a year of education based on those kinds of relationships that i showed you earlier which all of the research suggests gives a reasonably good estimate of how education helps people in the labor market it shows you separately for the private sector in the public sector and you can see that since the early 1980s in the public sector the payoff to education has risen considerably in italy there's been growth in the private sector since the late 1980s but not nearly as great it's certainly not the case that the payoff to education has fallen over this time period in italy and my interpretation of the evidence is italy started from a very low level in terms of the payoff of education from education and it's grown to close to the oecd average since then if you look at women you see a much bigger jump in the in the early 1990s uh compared to what you see here even though here you also see a pretty big jump for men in the early 1990s it's an interesting question why the public sector is so different from the private sector and that's something i'd be happy to discuss in questions although i think people in the audience probably know much more about that than i do given that the payoff to education has increased that the reward given to education is much greater now than it has been before one really has to worry about educational attainment one has to focus on educational attainment and i also think that gives some clues about the forces underlying forces in the labor market that are behind the payoff to education and i'll make that a little bit clearer in a moment i hope so this shows you data from the oecd on average education for 25 to 64 year olds in various oecd countries and you can see i point to italy on the far right only turkey mexico and portugal have a lower average level of completed schooling than italy so if you think about having an economy where the payoff to education is growing and that's not only true in italy is also true in the u.s it's also true in the uk and in many other countries italy is starting through a pretty disadvantaged position in terms of having a low level of average education compared to the rest of the oecd that's the bad news the good news is that italy's school enrollment has been growing very rapidly much more rapidly than most countries in the oecd and i guess if you go back to what i said earlier what's more important i believe is the change in education than the initial starting point that's good news for italy this shows you school enrollment secondary school on the left tertiary school on the right and you can see a very sharp trend here school enrollment has been growing very rapidly since the 1950s in italy and only greece ireland korea and spain have seen faster growth in education than middlemen okay so let me try to put this picture together we have an increase in the supply of educated workers in italy which is a quite rapid increase compared to the rest of the world and at the same time the payoff to a year of education looks like it's either rising or staying the same probably rising slightly in recent years what does that suggest in terms of supply and demand well the increase in school enrollment means that a higher share of the workforce is better educated the supply of higher educated workers is increased and this diagram shows you a supply curve in red and demand curve in blue and i have on the axes the relative employment of highly educated workers and the relative wage of highly educated workers so the supply of highly educated workers has shifted out because of the increase in school enrollment other things equal we would expect that to lower the payoff to education but in fact the payoff to education has not fallen in italy it's actually risen that suggests that the demand curve has increased more rapidly than the supply curve and italy looks very much like the united states in this regard in the us we've also had an increase in the supply of educated workers but that increase was not fast enough to keep pace with the increase in demand in my view and that is an important factor driving the increase in the payoff to education in the united states and it seems to be the case in italy as well so understanding what's going on to the demand for skilled workers why is it increasing is very important you can fill up a whole festival with talks on the role of technology globalization and the demand for skilled workers i won't go into this in very much detail but i do believe that changes in technology at work are a driving force behind this increase in demand for highly skilled workers but i also want to emphasize that the job mix is hard to predict and i'll do this by telling you the story about the movie the graduate i assume that many of you have seen the graduate you'll remember that benjamin was advised to go into plastics i think about 1968 so i checked before i came here and job growth in plastics is a little bit below average so he should have been advised to go into especially since he was in california into the high-tech industry instead of plastics i'll give you an illustration of how hard it is to predict the topics i did a little experiment where i took the set of occupations that we had in 1960. the census bureau in the united states has a set of codes where they code all of the major occupations very detailed and then i said how much of the job growth between 1960 and 2000 was due to occupations that were not part of the 1960 occupation codes occupations that were not even important enough to actually have their own code maybe they didn't even exist at that time more than half of the job growth was an occupation set for new occupations that did not exist in 1960 so it would have been really impossible based on the 1960 job codes the set of occupations that we were using to predict where the job growth would come from because by and large they were very small occupations which were not so important to merit their own code or brand new occupations which we didn't conceive of at the time so the job mix is continually changing one of the great strengths of the us labor market is it is flexible enough to accommodate a new set of jobs and also i would argue the education system is flexible enough to prepare workers for those jobs but we have seen one steady trend in the job mix which is this is from a study that was done by david otter frank levy and richard bernane if you break down jobs and say what types of skills do they use do they use cognitive skills do they use routine manual skills like digging a ditch or working on an assembly line what has increased are non-routine analytic skills non-routine interactive skills and what's gone down are routine cognitive skills the types of cognitive skills that can be replaced easily by a computer and manual routine manual skills that suggests that we need to prepare workers assuming that these demands continue we need to prepare workers for an environment where they're continually adapting where they're learning to do new things on the job they're not always doing the same thing with day to day also i think given the changes in the world economy the fact that so many countries are now competing in a global environment and the ability of the internet to have people deliver some of their output from really remote locations we need to focus more on interactive skills interpersonal skills what my colleague alan blinder calls personally delivered services okay let me next move uh to alternatives to just years of education i will make a segue here by pointing out marion jones the great sprinter did a campaign for nike which had the tagline education the more the better it dawned on me there's a limit to that statement because of course there's a cost to additional education so i think a question that we should address is how can we improve what we get for the education that we're currently getting how can we get more out of the current level of education and here i want to talk about evidence on test scores historically it was the case that educate that that that students cognitive ability as measured by tests seemed to work mainly through the amount of education that they got the schooling that that scoring better on standardized tests predicted whether people would get more schooling but it was the schooling that mattered not their test achievement now the test scores seem to have more of an independent effect and some research suggests that raising test scores from an initial level will also be associated with higher earnings there's a pretty strong relationship between average earnings and individuals test scores this shows you for the u.s for the national adult literacy survey their their reading test scores measured in lexiles on the x-axis and then average earnings and you can see a pretty tight pretty steep relationship here so how can we raise test scores and uh clearly understanding where countries stand in terms of their test scores is important uh italy like the u.s stands near the bottom this is from the pisa 8th grade math tests in 2003 and you can see italy is here the countries are ranked in terms of their average performance the united states is just two countries ahead russia is in between not very distinguished in terms of the overall world picture a couple of other facts from pisa which are of some relevance first family background matters quite a bit in pest performance especially in italy if you rank families based on their income occupational status and so on those who are in the bottom quarter in terms of socioeconomic status three times more likely to be among the lowest quarter of performers on the pisa standardized math test than family the children from families in the top quarter in terms of socioeconomic status so coming from a more advantaged background very strongly associated with test performance in italy uh that's true in the us as well it's cool all around the world but it's particularly true in italy in the us then secondly with the data from pisa we could look try to make a decomposition and say what part of the variability in the student test course is from schooling is from the particular school they go to and what part is due to other factors the component that's due to the school that students are attending in italy tends to be quite big a lot of the differences in performance in italy are coming about or at least are associated with the particular school that students are attending now pisa just gives a snapshot just shows you at a point in time how students are performing that's very important but it's also important to have some historical information to have a trend and i wanted to share with you some evidence that i've assembled to try to look at longer term time trends this is from the international adult literacy survey adults are given literacy tests in quantitative reading prose reading and document reading three different measures of literacy and what i did was to group people based on their birth cohort to try to convert these data into historical data now if people leave school and don't get any smarter or dumber this is a perfect approach you might worry that maybe when people get older they don't read as well they don't read as quickly there's not all that much evidence on that but that is a potential concern with these data but it's going to be less of a concern because i'm going to make comparisons across countries and you might think that the extent to which people become more literate or less literate as they age is pretty constant across countries so what do you see here for italy well first of all for all three of these measures of literacy you see an upward trend that means the more recent cohorts are much better readers than the earlier cohorts also the trends are about the same and i think it's a bit overkill to look separately at the quad at the quantitative literacy prose literacy document literacy because they all tell about the same story uh and to a certain extent you might say it's a good news story for italy this looks like progress each successive cohort to score in higher this is united states now i've scaled these the same way and the united states go backwards the united states started out at this uh considerably higher level than italy but then the united states plateaued and the more recent cohorts you can see are actually doing worse a little bit worse than the earlier cohorts in the united states and by the cohorts born in the 1970s performance on the adult literacy survey in italy and the united states is about the same which is interesting because that's what the pisa is showing the piece of data are showing as well now let me group some other countries for you the top line is sweden sweden starts out at a higher level of literacy and continues to grow the next line is the uk that's the red line uh and the uk had a very steep increase which is associated with the period when they increased the compulsory schooling age and then you can see the uk is kind of plateaued and then below that you have ireland and then italy and essentially all the countries are converging except for sweden at the same level so the last topic i want to address is how to try to erade how to try to affect the achievement levels and in particular the amount of money that's being spent on schools the effective school resources on student achievement now one economist eric hanishek has argued very strongly that spending money on schools does not produce output it does not increase student achievement this is a quote from an influential study that professor hanishek did he said the close to 400 studies of student achievement demonstrate that there is not a strong or consistent relationship between student performance and school resources at least after variations of family inputs are taken into account and that conclusion is very widely cited and i'm going to argue i don't think it's persuasive i don't think it's persuasive based on the evidence that professor hanishek produced and i don't think it's persuasive based on the best evidence that's available now i do think that this is a legitimate concern for italy because if you look at the data from pisa this shows you student achievement in eighth grade graft against expenditures on students up into eighth grade so cumulative expenditures while students are in primary school for each of the countries and italy is an outlier actually a slightly larger outlier than the us here is italy in the bottom right corner and the united states slightly above but both italy and the united states are below the line now the fact that the line is up for sloping suggests that as we look across countries those who spend more money on students through eighth grade do on average tend to have higher achievement scores on pisa now of course there may be many other explanations other than the effect of the expenditures um but i do suspect that the expenditures are playing a role here and then you have the united states and italy as outliers below the line now one of the reasons probably the main reason italy spends so much on schooling compared to other countries is that italy tends to have small classes the most important determinant of how much is spent per student is the number of teachers who are hired that's the main input in schooling and you can see italy tend to have fairly small classes compared to the rest of the oecd and it's expensive to have small classes because you have to hire additional teachers so the last topic i'm going to talk about is the effect of class size on student achievement since i think that's quite important for italy now the evidence that professor hanishek had in mind was based on a summary of the literature that he did recall in that quotation i read he said there were close to 400 studies or over 400 studies which found no strong or consistent relationship for 277 studies professor hanishek looked at the effect of class sizing student achievement and you can see that they're just as likely to have to show a positive effect of smaller classes as they are to show a negative effect of smaller classes and about 20 of the studies they didn't even report whether it was positive or negative which i've always considered a mystery and when you look at that it seems that well maybe there's a lot of support for this claim that having smaller classes does not help on average but there's something very curious in in these data first there really were not 277 separate studies or 400 separate studies there were only 59 studies and what was done was to take multiple estimates from some studies so from two studies done by the same authors professor hanashek took 48 estimates from my own study in the journal political economy he took one and if you look at the results based on the number of estimates that he took from each of the studies what you find is from the studies where you took one estimate overwhelmingly you find that smaller classes are associated with positive effects if you look at the studies where he took two or seven estimates overwhelmingly smaller classes were associated with better results and if you look at the studies where he took eight or more estimates which is just nine studies a small minority of the overall literature they're overwhelmingly you find that smaller classes are associated with worse student achievement so you have to wonder about these nine studies and i can tell you i've reviewed these nine studies they're not pretty many of those studies were not focused on the effective class size many of them did what i would call a kitchen sink approach they were interested in the effect of some other factor family income in some cases the students from more advantaged families have higher achievement and they just throw in all the variables they can think of often the specifications that they estimated made no sense it was very common for them in their statistical analysis to hold constant the effective class size as well as the effect of expenditures per student well if you do that what you're doing is comparing two schools to have the same expenditures per student but different class size how can they achieve that they could achieve that by paying the teachers less so the effective class size in that kind of a model is actually estimating the effect of having lower teacher pay rather than the effect of expenditures per student but reasoning is like rate is like racing and not like hauling and the single barbary steed can outrun a hundred dray horses and i think when it comes to empirical evidence it's often the case that the questions we're addressing are very difficult we might try to control for factors that we can't control for adequately and as a result the evidence is not very persuasive ideally what we would like to have is an experiment an experiment where we would randomly reduce class size for some students not for others and that's been done it was done in tennessee in the united states in the late 1980s 11 600 school children from 79 different schools were randomly assigned when they started kindergarten they were assigned to either a small class with 12 to 15 students a regular size class with 22 to 25 students or a regular size class for the teacher age and they were to stay in those classes for four years the teachers were also randomly assigned and i've been evaluating the performance of these students since the experiment began what do you see well in green is the average percentile rank on the stanford achievement test one of the standard tests we have of student achievement in the united states for the in green the students in the small classes in blue the students in the larger classes and then in the light blue the powder blue students who were in the larger classes were the teacher age and you can see that the performance was about five or six percentile points higher on average for the students who were in the smaller classes now that didn't increase the longer they were in the smaller classes instead at about five or six percentile points but it might be the case that if everyone moved to regular size classes it would have fallen or maybe even disappeared i've looked at some longer term outcomes for these students high school graduation was higher for the students from the smaller classes their act and sat scores were higher especially for the minority students from the smaller classes those are the tests that students take if they would like to apply to college team parenting was lower and the crime rate was lower uh especially for black males if they were in the smaller classes and college aspirations the chance of taking one of the college tests was higher you can see here's a press briefing where i had a very good assistant the assistance here has been excellent and i expect that the assistant i would sign here will do as well as mr clinton anyway um having been assigned to a small class does seem to have lasting effects let me skip over that slide to wrap up when you look at the data more closely it looks like being in a smaller class had especially beneficial effect for the more disadvantaged students for the students who were below the poverty line for the students who were minorities for the students who lived in the inner cities it also had a bigger benefit for boys than it did for girls i interpreted this result as suggesting that it helped to socialize students especially in their early grades now i don't think having small classes matters as much at the college level or the high school level as it does at the early level at the early grades but at the early grades i think it helps the teacher to maintain control in the classroom if he or she has a smaller class and the students learn to be better students if you try to do a benefit cost comparison looking at the gain in test scores how much is that worth later on what i concluded is that the benefits were about twice as much as the costs or another way of saying that is if you use an interest rate or discount rate of six percent in that case the benefits would be about equal to the costs so it's not an extraordinary large return i said earlier there's no free lunch it's about the return i think you would you would say that seems reasonable that's close to the return to capital for the more disadvantaged students the return is higher overall six percent seems like a reasonable rate of return so let me make a few comments about implications of all of this for italy and then i'd like to take some questions first in italy educational attainment years of schooling as well as achievement are growing and they're both growing pretty rapidly and i think it's important to continue that trend um i think it's difficult to raise achievement beyond the levels where italy is currently i think that's going to be a real challenge because class sizes are already small in italy but i think maybe focusing at the disadvantaged students focusing more in the underachieving schools because they seem to do particularly poorly in italy would be one strategy for improving the overall average for italy um and i suspect it's going to be the case that in the future the demand for highly skilled workers is going to continue to grow and probably continue to outstrip the increase in supply in italy uh which makes investing in education all the more important i want to stop there take some questions uh very very good talk very interesting talk we'll get a lot of your stuff and some others i have a couple of questions i think the rate of return estimates you get are all right i mean 10 12 i'm not sure exactly but that seems like the right ballpark but increasingly i think we've become aware that the return the earnings return is only a part of the total return and there's something we call the education premium paradox in sort of an analogy with the equity premium paradox that almost every variable i think has been looked at education helps if you look at health i think there we have very good evidence that more educated people are healthy and it's not just reverse causation although there's some of that look at marital stability investment in children voting you name it it's hard to find anything where education isn't positive so it's hard it's difficult to add all that up but it's going to mean that the total return on education is far i think in excess of 10 maybe 15 i don't know what the number is i say well how can we have an equilibrium where we're getting these high rates of return on education and we aren't getting more and more people going now you said well people don't like school i know they don't like school but they have to not like school a tremendous amount in order to explain this well they may be capital constrained but it's very hard in terms of what we know now for at least for the advanced countries to explain it by capital constraints and so i think it's a real paradox i don't have the answer to that i've worried about it a lot and i wondered what your solution is but i think you got to be careful just saying well people don't like school i mean that's that's like bring in some residual and saying well i close up the system by this variable which i can't measure and i don't know what its magnitude is so i'd like to hear your response to that thank you very much gary i think the first part of your question which is i had in my notes and discuss a little bit on crime and education and health and so on to where you mentioned like neglected to mention that is actually not necessarily the second word because in the second part all you have to do is point to the monetary return increasing it from six percent to ten or twelve and say why haven't we been why hasn't haven't school been flooded with more students in the us well but it's the increase that should affect i mean if we were initially in an equilibrium there's the increase that should cause enrollment would have to be that it raises those other things by more than ten percent i mean suppose suppose i think of the monetary returns i guess giving me overall well i see you're shaking your head so i'll say this with some questions um i'd often thought about the role of education in home production you know topics which you thought about very deeply and i can't tell you what it was specifically about my education especially k-12 which is beneficial to me on the job and i suspect that those same kinds of things are beneficial to me in home production in interacting with with my family and interacting when i go shopping and making decisions so that's a part of the return which is a much bigger part in terms of my my hourly quality than my work life uh which is not getting counted and i wouldn't even think 10 there's no reason to think it's not the same as what's going on at what would work but anyway i do agree with you that education has these benefits in addition to the narrow economic measure uh that i've been looking at when i did the benefit cross study for the star experiment if you added in all the results for crime if you ended in the benefit from and benefit from reduced teen pregnancy in some ways it was very small but the biggest part there that this when it came to the school resources was from higher achievement and higher earnings so to your main question which is why haven't we seen more of an increase in development we have seen some in the u.s the group that hasn't increased is african-americans african-americans in the u.s have had pretty stable educational attainment despite uh what seems to be very high increasing returns high and increasing returns for the next year education and i'm not sure the answer that because i'll take a little bit of exception when i when you say that saying that students don't find school interesting is a residual there's actually some evidence on that danny kahneman and i have been collecting data on how people feel about the activities they do throughout the day and what was remarkable to me is people do not find school interest and in fact the young men they find it painful one of the one of the emotions we asked them about was how much pain did they experience and especially when they were doing homework they were reporting high levels of pain i think part of it does have to do with capital constraints liquidity constraints but i'm not going to link too far on that i was actually quite cautious about saying liquidity constraints were all that important in the debate i have with jim heckman um so my my tentative answer is that we should think about ways of reaching the children who are at high risk of dropping out because it's difficult to find an alternative to help them get on track if they do drop out and i think trying to make school more interesting for them more value where they can see what that value is would be helpful i do think that school enrollment has increased in the us because of the increase in in the payoff education but not as much as i would have expected or achieved looking at the recommendations for italy regarding education maybe you kind of you are mentioning what one kind of return of education is the weight return that you can get but we know that as you mentioned there are externalities that are not captured by the wage dimension and in particular growth externalities so the effect of education on growth will be much bigger than what in the dfa will be on earnings now in italy maybe and you mentioned some work you know relating growth to levels of education and when you restricted to oecd data you didn't find so significant but maybe one one way to get significance is to distinguish between different types of education uh like lower education and higher education and to interact the composition of education with how developed the country is so for example for italy and france what is striking is that at the at the stage of development they are investing more in higher education would be uh would be very much growth enhancing particularly if you combine the investment in higher education with autonomy we move towards autonomy of universities so i was wondering what you think about you know isn't that kind of first order thing that italy could do you know to on the education side to to improve growth performance i certainly agree but not because of material i presented today that having more autonomy and higher education admittedly makes a lot of sense uh at least in some subset of schools but i tried pretty hard in this work i did with michael lindahl to look at higher tertiary education versus secondary education the initial level changes across the oecd and you can often get mush you know when when you try to include the levels in at the same time or in barrow's working separated out by sex and found completely bizarre results that women's education didn't seem to matter which i found to be a result which is not very robust i do think that for italy you know if you look at the individual well by the way i would say the evidence on the internality to my mind i think they're small i mean i think it's more presumption um for for oecd countries but i would say based on the evidence that if you want to look across different groups and so on investing in education for women in general makes a fair amount of sense the payoff to education is higher in italy uh they've also women have had faster growth since world war i believe and it may be that they're going to enter into more flexible sectors of employment i do tend to think that you know if there are going to be externalities it will probably be from higher education which will help to adapt new technology but so far i've been skeptical that that they're there and i tend to prefer parsimony so in my mind you know if you can explain the data pretty well without having to turn to spillovers then i tend to prefer that so maybe i'm just kind of stubborn about that but if i if that's where i started with a null hypothesis i hadn't seen enough evidence to move from that me foreign a let me first say that you have my sympathy because my wife is also a school teacher and i know that means many nights when she's grading papers and writing tests the question about how can we motivate young people to do better in school and stay in school longer which is very much like the question gary becker answered um one thing i would think about i i think you know as an economist i think trying to explain to people what the return to education means what it trying to relate it to them in their own minds and how there's a value can matter but i don't think that's going to matter for all students of course there is a lot of research that suggests that more disadvantaged students learn better in terms of contextual learning and when i worked at the labor department in the united states i had done some work trying to reform our job training programs and the direction we were trying to move in was to show a context for why the students were learning geometry why your carpenter you actually could benefit from geometry that will help to make you a better carpenter and so they can see the context connected with the substantive material and i think that is one way to try to make schooling more interesting to students i can also tell you i think it's difficult i've been writing for the last four years of textbook to try to teach high school students economics and it's a lot harder than it looks at trying to write in the language that the students can can access and to use examples that the students relate to is it is difficult so i have a great deal of sympathy for the teachers who do this because i think it's much harder than it seems uh the other thing i would add is one trend i don't know how well appreciated this is in italy since education of the parents generation has been growing so quickly in italy that should have some beneficial effects i believe that education does help to make people better parents will help to improve their their children's performance in school and i think that's something which can uh and i think you might expect to see improvements coming into pisa over time uh just because of improvement in the education background of pirates center proposed is they is i'm not familiar with the study that you've referred to so i don't know over which time period is referring i can tell you as a regular pattern those with higher level of education have much lower unemployment rates and that is true in italy the data that i looked at most recently as well as elsewhere predicting which sectors of the economy are going to grow i think it's very difficult i think the general pad this gets me back to plastics and i don't want to sound like the character which advises to go into plastics one of the advantages of education is i think it makes it impossible for people to make transitions and i do think that the changes in the global economy are going to lead to you know we clearly we've seen as countries grow they make a transition from agriculture to manufacturing from manufacturing to services and some of the factors that you mentioned about italy i think make the transition to services more difficult some of the regulation in italy makes that more difficult it's also one of the things that makes italy so distinctive such an interesting place to visit and i i suspect that that will continue at a broad level i suspect that the type of services that will grow will be more personally delivered services um because types of one of the things which we're just at the beginning of isis back of seeing offshoring of services but there's no reason why that shouldn't grow certainly companies have a very strong incentive to develop the technology so that they can offshore the provision of telephone centers for example to low-wage countries which have a high supply of skilled workers especially countries that speak the same language i think that's probably more of a risk for the u.s than it is for italy but if that's the case and i think what you'll see is more of the domestic employment going more of domestic employment going into industries where the service is delivered personally and where social interactions matter more one of the reasons why i think education does command such a high premium in the labor market is that people who have a higher level of education are able to adapt to changes in the economy and can switch to different sectors when those civil sectors tend to grow so i would say as kind of an insurance policy you might think of higher education as a way of maturing against fluctuations which are our iowa doctors are very difficult to predict foreign is regret first i'd be interested in seeing your study which uses the tim's and relates student achievement on that science and math exams to school resources and class size i think you have to be careful when you use data collected like the tim's which uses the class sizes that the school of ministers administrators happen to assign students to there's a very nice theoretical paper by ed lazier in the quarterly journal of economics on the production of school production and how schools should optimally choose new science students the classes the model he has in mind is kind of similar to what i concluded about socialization but he's much more analytic about it and the model he has in mind is that if you have a big class you're more likely to have disruptions and suppose that there's some chance that a student is going to disrupt the class well that slows down everybody in the class if you have a bigger class then you're just going to get more disruptions so in lazier's model the advantage of reducing class size is that it reduces the number of disruptions now if you have students who are prone to disrupt the class and students who are very well behaved and not going to disrupt the class the optimal assignment for the school administrator is to group the students who are well behaved and will not disrupt the class in very big classes i put the other ones in smaller classes and to some extent i think school administrators have this in mind if you look at special needs classes they tend to be very small and in a country like korea where students do seem to be better disciplined than say in elementary schools in the united states then i think they can get by with larger classes so you have to bear in mind the way that the school administrator is choosing to assign the resources that he or she has and the optimal assignment is probably to put the weaker students in the smaller classes where they would get more individualized attention and they would disrupt fewer of their classmates because the classes are smaller what i had in mind is that the chance of disrupting the class you can think of it in that way in lazier's model itself changes depending upon the student's assignment at the early grades especially a child with a disadvantaged background had no idea what the function of school is or how to behave in the classroom and that's one of the things the student is learning students learn that at most once so you think you would get a bigger benefit of first-year students or in a smaller class um so i had in mind that the socialization or the way to education class and how destructive they are would evolve depending upon what their experience had been to that point and if that's the case that i think you can get a bigger bag for the dollar in terms of class size of the lower grades than at the number older grades good evening i am an english teacher here in toronto and i just want to tell you something not about economic and economical level because i'm not it's not my field so i'm not but anyway um a couple of days ago i talked on the phone with a friend of mine who is a retired headmaster in dublin and he said he quoted me an interesting proverb in english which says praise youth and it will flourish we were talking about some good bad marks on my daughter's head in mathematics and he said praise youth and it will flourish what i think is really is that um we have to be as teachers i am a teacher we have to try to find the best relationship among us and our students in order to attain the same goal so we have to treat them as partners in towards something which will make them shine as a person worthy person you know that's all i don't know if this is quite on point but i want to mention something we have in the u.s called kip academies kipp which in some sense do try to make partners of the students and their families and the the students uh tips schools i think rob brings this right to charter schools so they're publicly supported schools but they can deviate from many of the regulations they tend to meet longer uh they sign more homework than students stay in the schools longer italy has done a long school hours but the family signed a contract that they're going to oversee the students homework they're going to make sure the student attends and then those types of schools do seem to be generating very positive results and i think that's associated with what i said earlier about time on task i think that is investment but i think if you think of it as partnering and making sure that the student and their family are delivering on their side of the park i would agree with that to help african americans and in my own experience with african-american students and i know some of it are very i've seen the women do a lot better even though they come from the same background and i'm wondering whether you take culture into account because we know historically african american women have been much more conscientious and they've played a much more effective role and could that also play a role in and let's say the fact that the males drop out whereas like the females don't and have you paid attention to that uh yes and you're certainly right if you look at rates of college enrollment or high school dropout um african-american women are much more successful than african-american males and um and i also and i think that you know can you trace this through where in most families male 1 is more than the female and that generates lots of other social problems down down the road in terms of the formation of african-american families one of the nice things about having an experiment like i had is that cultures held constant in tennessee about 40 of the students for african-americans it was a large group in the 11 600 and because children from the same school coming from the same background randomly assigned that does hold that aspect of culture constant and i found it encouraging that the strongest results were for the african-american males at least in terms of the test scores and the results for crime tended to be uh effectively if we were due to the african-american males um so i think it's awfully hard to change culture and i and i this is one area where i'm quite lazy fair and i don't think of kind of restructuring households and thinking about how to improve outcomes so i think here we have an accepted intervention which seems to help the group that has the most difficulties in school which is a contrast incidentally to preschool so i didn't talk today about preschool education because italy has very strong preschool education there's been a lot of research in the united states recently on preschool education or actually the research isn't so new but the enthusiasm for it is somewhat new and preschool education for disadvantaged children seems to have a very high return but that's particularly true for women particularly true for the african-american women the african-american male students don't seem to benefit nearly as much from preschool which is a puzzle why preschool is not helping them and i think it's important to bear that in mind as we kind of and i suspect we will move more in the direction of investing more in preschool i don't think that's going to solve the serious problems we have with the american african-american students american society um i'm a student and so on the other side um foreign personality well let me since you got an applause i would say i agree with you one thing that strikes me in school is we don't tend to value students time and i said earlier that reducing class size or the number of teachers per student is the most costly input but and truly the most costly input is the opportunity cost of the student's time and we don't treat students as if they had a meter going and we don't worry about wasting their time so i think that's a problem kind of a universal problem in education i do think one way of developing students talents as i mentioned earlier is more contextual learning teaching in the context of what interests them what they might be doing in practice later on uh i would also say in terms of developing talent in italy i was struck to see how few farm students attend higher education in italy italy is one of the lowest in terms of the oecd in terms of attracting students from abroad to college and graduate school and i suspect that that's more supply than demand in the us especially in grad physical a very high proportion of our students are from abroad they i think energize our higher education they bring a different perspective which i think is positive and i suspect in italy in part because of the lack of flexibility in higher education the lack of competition among the top research universities attending graduate school in italy is less attractive which i think is unfortunate because i think italy has many natural advantages which would make it a mecca as far as college campus goes so that's another area where i think we'll probably do a better job uh um um is and well i can't give a complete answer to your question about how research on education translates into public policy in italy but i can tell you that i think having a festival like this one is one way for the public to learn more about the research findings and through the public's own education uh to try to deal with public policy uh and i have to say i'm very impressed by the interest uh that's been displayed uh in the session and throughout throughout this festival uh and hopefully uh the politicians will eventually see and have translated to them the research findings so that they can then practice foreign so you