Wellbeing Report seminar series: Redistributive Effect and the Progressivity of Taxes and Benefits: Insights from the UK

Patrick Nolan:

We're at 10:00, so I might just introduce the event.

[speaking in te reo Māori] Tēnā koutou katoa. Ngā mihi ki a koutou, kua hui mai nei i tēnei rā, e tautoko ana i te ako. Ngā mihi nui ki a Stephen Jenkins, te kaikōrero o te rā. Tēnā koe. Nau mai, haere mai ki Te Tai Ōhanga.

Well, welcome, everyone to the Treasury and it's my great pleasure to be able to welcome you and chair this event with Professor Stephen Jenkins, which is the latest in our Wellbeing Seminar Series, which is being well organised by Diana Cook and the team. Just to provide a little bit of background, as you may remember, back in April the Treasury Secretary, Dr Caralee McLiesh, publicly launched the work programme for the Treasury's first wellbeing report, which will be published towards the end of this year. Te Tai Waiora will be a report on the state of wellbeing in New Zealand, how it has changed over the years and the risks to and the sustainability of wellbeing. Now as part of this, we are running this broader seminar series to bring in the key ideas, research and evidence, and to be able to draw on the advice of international experts, which is critical for shaping our thinking.

We've had a series of fantastic speakers so far and we've got a number of other seminars planned for the rest of the year and into 2023. But today, we are delighted to have Professor Stephen Jenkins join us and I'm personally thrilled. And as you all know, I am the Manager of the Analytics & Insights team, and the work that Stephen does is very relevant to the team and so I'm very much looking forward to hearing what he says. So Stephen is the Professor of Economic and Social Policy at the London School of Economics and Political Science. He is a quantitative generalist and his research interests are in implied economics from a longitudinal perspective with reference to the distribution of income and its redistribution through taxation, social security, and the labour market. His research has been published in a wide range of journals and he's published several books. He was named a Distinguished Fellow of the New Zealand Association of Economists in July 2019 and he is the elected President of the Society of the Study of Economic Inequality for 2021 to '23.

Today, our guest will focus on findings from his latest research on the redistributive effect and the progressivity of taxes and benefits as observed in the last four decades in the UK. This is very relevant to our work at Treasury as we continue to deepen our own research and understanding of the impact of inequality has on New Zealanders' wellbeing. So how the session today will work is that Stephen will present for around 45 minutes and then we'll have the remainder of the time for questions. As Stephen presents, please feel free to enter any questions you might have in the chat and I will ask them when we turn to questions and you're also welcome to vote on the questions that you liked. That way, we can make sure that those most liked questions will get asked. So make use of that facility there. So thank you again, Professor Jenkins. I'm really looking forward to your talk and I will pass over to you. So thank you.

Stephen Jenkins:

Thank you. I will try and share. That okay, Patrick? You can see fine? Yep? Good. Hello, everybody. Welcome. I should first of all say that this is joint work with Nicolas Hérault, who may be in the room. I'm not sure, he was planning to come but may or may not be here. Yeah, redistributive effect and the progressivity of taxes and benefits. That's a real mouthful but I hope I'll explain it as I go along. The first thing I should probably do is locate it in the area that Patrick mentioned you were starting, namely Living Standards Framework. Clearly, it's been revised and this big triumvirate of stuff on the left hand side and there are related reports. This is an excellent report I came across about what's been happening to wellbeing in New Zealand from all sorts of dimensions. So what the heck has my talk got to do with it?

Well, I guess the answer is it's basically a deep dive but only in one aspect of the framework. But I would say a rather important one, namely particularly focusing on income. And the other thing, the reason I've drawn the second bubble over here is that I'm going to be drawing a link between what happens to the income distribution and what the government does. In particular, what it does with taxes and benefits. So the key ideas then are to tell you about an approach that links wellbeing outcomes, one in particular namely income distribution, to the government's redistributive tools, the main ones, the tax benefit system. So the issue is what effects do taxes and benefits have on the distribution of income, both at a point in time, and my focus today, trends over time.

So secondly, the key idea enabling to do that, I'm going to use a framework proposed by Nanak Kakwani, a big guy in this field who worked at the University of New South Wales for many years, retired from there and is doing other stuff around the world. But one of the things that Nicolas and I do that is particularly new here is just exploit Nanak's framework to propose and apply a counterfactual approach to understand what the tax, which aspect of the tax benefit system played the dominant role in accounting for the trends that I'm going to look at.

What's new here in this field? What's new is in particular the detail combined with the length. So we've got yearly data for 40 years. That's the UK from 1977 through to almost, well, just before COVID, 2018. Nothing original about the data in the sense that I'm exploiting those that are used by the office for National Statistics or ONS in their official Effects of Taxes and Benefits, ETB article that they produce every year. So they produce this report every year but it's very rare that they take a long look and put things together.

So I guess what we're doing here is providing a fresh look which provides a more big picture perspective that hasn't been done before. And I think the only study that I'm aware of for New Zealand in this area is by Patrick's brother, Matt, part of his PhD. He has a recent working paper on a related issue. It's not quite exactly the same thing because he doesn't use exactly the same decomposition but it's very close. So it does show that there is potential in New Zealand for doing this sort of stuff, exploiting the HES in particular. So that one feature then is detail, long period. Second thing we do is that we look at cash and in-kind benefits, direct and indirect taxes, and most research looks at either direct taxes or cash benefits but not all those four components. But I'm going to focus today on direct taxes and cash benefits.

So having said we do all four stuff, it's going to be too complicated. We don't have enough time. I'm going to do, give you the hot take on just taxes and direct taxes and cash benefits. Also new in the paper, as I said, is the counterfactual approach. You'll see it rather basic. Well, I'm going to claim it's informative and I haven't seen it before. If we've got time, I'm going to focus on also very briefly on an empirical implementation issue and that's the definition of the so-called pre-fisc or reference distribution and show how it matters. So that's the background. So what is this Kakwani decomposition? It's been used a lot, but the idea is that the redistributive effect of taxes and benefits, which is just a fancy way of saying the degree to which inequality is affected by the tax benefit system, depends on three independent factors.

First of all, there's the average rate of taxes or of benefits overall. Secondly, is basically how the average rate of tax varies with income level. To what extent is it proportional or deviate from proportionality? And then thirdly, there's a residual, which is the reranking of individuals by the taxes or benefit system that takes us from the pre-tax to the post-tax income parades. So that's the overall bit in words. Now I'm afraid I'm going to jump into symbols to elaborate basically what's going on. So it works by summarising inequality in terms of the Gini coefficient. We summarise the income distributions inequality before the operation of government pre-fisc by the G, G for Gini, X for before. And the distribution after the operation of taxes or benefits, the post-fisc distribution inequality is summarised by the Gini Y. So redistributive effect then is effectively just how much inequality gets changed or reduced by the particular tax benefit instruments that we're talking about.

So the formula of the Kakwani decomposition if we're talking about taxes is that redistributed effect overall. R of T is equal to vertical redistribution plus horizontal distribution on the right hand side, the D term, that's essentially reranking. The core thing in here is where we get both the first term is summarising or based on the average tax rate, so total taxes paid divided by total income. That gives us lowercase T. But this capital K here is Kakwani's tax progressivity index, which is just a fancy summary way of summarising, in one number, the extent to which, how average tax rates vary with income.

So that's essentially what's going on. If we do it for benefits instead, then we essentially change the Ts to Bs. Little B for example, lower case B is total benefits, cash benefits paid out as a fraction of income and so on. And the absolute value of this is the progressivity of benefits. Again, relating to the extent to which benefits are proportional to income. And again, there's a reranking term. You can combine taxes and benefits and so the term gets elaborated. I'm not going to talk about that at all. There are various analytical reasons why it gets really complicated if you want to do that. So I'm just not going to do it. I'm going to look at taxes and benefits separately. The data, as I said before, we're using the ONS data, just their data in fact. So nothing original there. The coverage is, it's basically the data nationally represented of UK private household population and the variables are exactly the same at the ONS'.

So any contribution that we've got is to basically how we play around with the data rather than the data themselves. That becomes particularly important were I to be looking at indirect taxes and at non-cash benefits, but I aren't. I'm on rather non-controversial ground, I think, for what I'm going to be doing. So those are the data. Where do they come from? They come from the so-called Living Costs and Food Survey or the LCFS as it's been called since 2008. And before that, it used to be called the Family Expenditure Survey.

You can think of that being like New Zealand's Household Economic Survey, the HES. Though the UK actually also has a Family Resources Survey, so we're lucky in the sense. Bigger country got a couple of data sets. But essentially, I'm working with the UK equivalent of the HES. And in fact, its sample size is rather lower than the HES' is at the moment, around 5,000 households. It works for financial years. Before that, it was calendar years that I'm just going to refer to a year. It doesn't make any difference. The data use sample weights. Basically, we use them. It's not a big issue.

Okay, so we're going to be looking at taxes and benefits. So we need some income concepts. I'm going to focus on the three at the top. So the first one, O, is O for original or otherwise known as market income. People might call it factor income. But essentially, this is income that people get from the labour market, their investments and the savings and includes private pension income. On top of that, you could have gross income, G for gross, which is basically original income plus cash benefits. And in the UK, that is taken to mean state retirement pensions as well. They're included in there. So it essentially grosses some of those.

And then we end up with disposable income. It takes us from gross income where we take off direct taxes. And in the UK, direct income taxes include, basically, income tax payments, employee national insurance contributions, and any local taxes which are essentially council taxes, which are a local property tax. I won't go further than that, but it's a local tax and it's conventional to include them in direct income taxes.

Now, we have this problem of course that 5,000 pounds a month is not worth the same to a family of four as to a family of two. So to make things comparable, we've equivalised those incomes, adjusted them using an equivalent scale. We use the one that's most common in the UK, namely the modified-OECD scale. And we look at distributions amongst all individuals. So we're basically being like everybody else, both in terms of the definitions of the income variables and in terms of the equivalent scale. So for example, the Canberra Expert Group, their recommendations, reports and statistics by Eurostat, OECD use these income definitions in Brian Perry's wonderful MSD reports that have been coming out for a few years. Been paused. I'm waiting for the next one. Brian uses these same definitions as well and uses headlines on a New Zealand specific equivalent scale rather than modified OECD one. But that's not such a big deal.

Okay, the next thing is, well, redistributive effect is the looking at the difference in the inequality between the pre-fisc distribution and the post-fisc distributions. So now, I have to fess up about what those pre and post-fisc distributions are going to be. What we're going to do is when we're going to look at the redistribute effect of cash benefits, we're going to compare original income, market income with as the pre-distribution and use gross income as the post-distribution. Because remember, gross income is original income plus benefits. And then correspondingly, when we're going to look at direct taxes, we're going to take gross income as the pre-tax distribution and disposable or net income as the post-tax distribution. Okay, because the difference between those two distributions, the G and the D is capital T. Now doing this, we're not out of line with the literature at all, but it can be questioned. And if we've got time, I'm going to come back and show what difference it makes later on.

A lot of the theoretical literature people don't worry about this, they can of course... Then they can just write about pre and post-fisc distributions. Unfortunately, us empirical guys, when you have to implement these things, it becomes a choice. So this is why I've parked it and I'll come back to it. Okay, put things in place. Let's first look at Gini coefficient for the various distributions. I've got them there for five distributions at the moment, but throughout, I'm going to focus on three, the O, the G, and the D. Essentially, those in grey. On the vertical axis, I've got the Gini coefficient, if you're unfamiliar with this, it runs between the zero when there's no inequality at all and would run up to 100%, 100 on the scale if there were complete inequality, one person had all the income.

On the horizontal link, access is income going from 1977 as I said through to just before COVID. And familiar story here, whether or not you're using these ONS data or the other official series from the Department of Work and Pensions, essentially, you can't see off the left hand side of the picture but statistics going back to 1961. So basically, inequality fall up to the late '70s but then rapid rise through to around the beginning of '90s and then a flattening off. I'm using that generic description because you can see that essentially, the patterns are broadly the same, whichever income definition you choose. But I'm going to come back to it because it's actually the differences between these series, which may look rather small in the scale that we've got here. The essence of the talk. The other important things to take away, and this is similar for whatever country you're looking at, is that clearly that you see that the distribution of market income, original income is much more unequally distributed than the post-tax post-benefit incomes.

So for example, whereas we add in cash benefits, we get a drop off in Gini points of at least 10 percentage points, that's large number. And then as we have the operation of the tax system as well, we drop another five or so points and that's brought... Though of course, that gap is changing over time, which is the essence of the talk. Okay, so let's look at the differences. So remember, redistributive effect is the difference between different series. And this is, again, going through exactly the same thing on the horizontal axis. On the vertical axis, we're now looking at differences, redistributive effect, differences in the Gini, it's called Gini points because it's percentage points. Focus on cash benefits and direct taxes. And what we can see is that it's, for cash benefits, it's a series of rolling hills or waves, if you will. This dash line at the top.

And basically, we end up at the end around 10 Gini points as where we were at the beginning. But what's going on in the middle? And I'll alert you right early on is that there's no coincidence that these little hills are there. They coincide with recessions in the UK in the early '80s, in the early '90s and the early 2010s, after the GFC, great recession and so on. So essentially, this is flat but with some waves. Whereas if we look at direct taxes, the redistributive effect of taxes has been bobbling around a bit, but there's a more distinct increase over time from around two Gini percentage points, getting closer to six. So quite a large increase it appears over time. So this is the left hand side, the outcome variable that's going on. We want to see the extent to which, which aspect of cash benefits or direct taxes is responsible for what we're seeing there. What's driving the trends and the levels?

Well, so let's look at the components. So remember the components in the Kakwani formula, the first of them includes the average rate, either of taxes or of benefits. So if we focus on the grey line here, you can see for cash benefits, hey, we get those hills again and the hills are pretty much in the same place. On the other hand, if we also look at direct taxes, well, this is much flatter than what we were seeing before. We're not seeing it. We don't see hills that are mimicking what we see for cash benefits. Okay, that's averages. What happens if we look at progressivity indices? Here are the two things to look at. If we look at direct taxes, what you can see is that the number, it is just a number, is essentially increasing basically after it gets to the beginning of the '90s, relatively flat, begins to increase and then is finally flat again in the 2010s with some bobbling around over the time.

So the fact that it's going up means more progressive. That is the extent to which tax or average tax rates rise with income is getting steeper. Now what's happening with cash benefits? Okay, we have negative numbers here because remember, cash benefits work the other way. So a more progressive tax system is one where benefits decline with income. Which is of course what you would expect in most economies because cash benefits goes to the poorer people at the bottom and get withdrawn as you move up the income distribution. Okay, so you get negative figures. What you can see is those, they're getting more negative over time, so becoming more progressive, but around the beginning of the 2000, everything gets flat. Okay, so those are our components. The final one we have to think about is reranking, which I was referring to as a residual. And I've expressed there's more stuff in the paper, but I've expressed it here as a fraction of total redistributive effect.

And the things that I'm interested in, cash benefits and direct taxes, you can see for most of the period, this is less than 10% and basically flat for most of the period. So essentially, I'm going to ignore this reranking thing as an important thing to take in account for the rest of the talk. You might say, "Well, hey, what about direct taxes at the beginning here and it's unclear exactly what's going on. The smoking gun is the fact that the UK changed in 1990 to having the individual as the tax unit. Before that, it was the family. Either a single individual or a married couple. And so I think there's something going on there, but in a sense, it's never... This residual reranking term is never particularly big. So I'm going to ignore that going on.

Right, I've shown you the pictures. So we have to now think more about the trends. Was it the average rates that were driving things or was it progressivity changes that were driving things? So here's how this basic, very basic counterfactual exercise works and there're in fact two of them. The first one, remember this is the formula, what we do is fix the average rate and reranking so T and D at their historical average values, but allow progressivity, this K thing, to change as we observed it. And if counterfactual redistributive effect calculated, using these counterfactuals, tracks observed redistributive effect, this indicates that progressivity accounts, changes account for the observed trends. The second counterfactual exercise is just to turn things on their head and do it the other way around. We fix progressivity and reranking at the historical average values but allow the average rate to change as observed. And by contrast with the exercise one, if the counterfactual calculation tracks the observed redistributive effect, this indicates that the average rate changes are accounting for observed redistributive effect trends. And you can do this separately for taxes and benefits. Okay, so what happens?

Okay, so in this figure here, I've done started off for benefits and what we can see, so here again, time on the horizontal axis, redistributive effect on the vertical axis and there are three series here. The solid line in black is the actual series. If we fix the average rate, we get the grey series, which does not track it at all, but amazingly fixing the prep, fixed progressivity series, we get almost a really good tracking of the actual series. So what's going on here is that we can say that the trends in the redistributive effect of cash benefits are largely associated with cyclical changes in average benefit rates, those little Bs. So that's total. Remember, B is total amount paid out of benefits relative to total incomes. What happens if we do the same exercise for direct taxes? Here, we've got the same things on the axes, we've got the actual series again and the solid black line. The grey line is to fix the average rate. The dashed black line is the fixed progressivity one.

And now we see the opposite. So if you see fixing progressivity, we get a series that doesn't look at all like the actual one. On the other hand, the grey one, fixing the average rate series counterfactual tracks the observed one pretty damn well. I'm amazed that these numbers come out like they do. So what we can say here is that trends in redistributive effect of the direct taxes are largely associated with trends in progressivity of direct taxes. You might have said that from seeing the charts that I showed you earlier, that these are the numbers that clinch it. Okay. Seeing as I seem to be going rather fast here, I'm ahead of time, I'm going to take a bit of time just to talk about the issue about sensitivity to the definition of the reference distribution. And here, I'm going to focus on direct taxes.

So remember when I was assessing redistributive direct taxes, the reference distribution that I used was gross income. That's fairly standard, definitely standard in the UK. And the way you might justify that is that well, gross income is the sum of market income plus cash benefits. And it so happens that in the UK, quite a large number of cash benefits are in fact taxable. So it makes sense if you're going to think about progressivity of taxes and so on, to assess what taxes are doing against a distribution which includes benefits, many of which are taxable. On the other hand, if you were to go back to many of their original studies, including Kakwani's own studies for Australia and his other applications, including early ones for the UK, important work by Peter Lambert and others, they use the distribution of market income. Well, does it make a difference? This is what this chart is trying to summarise.

So again, time on the horizontal axis, redistributive effect on the vertical axis. So we've got expressing things in percentage terms here, so they're comparable. So the black lines are effectively what we were seeing before, the dark lines. So as it says down here in the legend, this is redistributive effect when we use what I was showing you before. Progressivity was the thinner dash lines, this one. The average tax rate was the thin one. Now, just redo the exercise but now use original market income to do the calculations. So we have redistributive effect here rather than the dark line, progressivity, the short dashes is here, and the average tax rate line is up here. So what do we see? I remember the original income on average is less than gross income. So effectively, that's shifting the average rate line up.

But what we see in terms of redistributive effect is that the lines, comparable pairs of siblings of lines, tend to have the same shape as before, they just moved up and down. So if we look at redistributive effect, they're in parallel, but what's really rather important is where they've moved to. And you can see that if you move to using original income as the reference distribution, redistributive effect moves down. That means there's less of it, that's saying that direct taxes have less of an effect on inequality as measured by the Gini coefficient. In particular, you can see at the beginning of the period up until around the 2000s, they have no effect. Okay, and then there's the other shifts around that I was saying before. So you get rather a different perspective about what the taxes, direct taxes are doing in terms of inequality if you change the distribution.

So which approach is right? That's a difficult question. I've made out the case for using gross income. It's standard. It's what Nicolas and I decided to do, but I don't think anybody's experiment would be using these two definitions. And lots of people have used original income as well, particularly in other countries. And we've shown that it does make a difference. Not to the decomposition of trends. If you were to redo all those counterfactual exercises, you'd still get the same culprits as coming up as being important. But what you're doing is decomposing a different redistribute effect. And that's rather important. This chart is just the same stuff with reranking. And again, it's just shifting things around. The slopes are the same. You can see that reranking hardly changes if you were to change. It's flat.

So this leads me to draw four sorts of conclusions from this work. First of all, I've argued that Kakwani's decomposition formula provides a nice way of doing things. It summarises changes in redistribution and progressivity over a long period of time. The UK, in this particular case, almost 40 years. We see that reranking plays a relatively minor role, that horizontal redistribution component. But then there's vertical redistribution and that is clearly very important. But then we say, "Well, is it changes in average rates or is it changes in progressivity?" And the answer for taxes, direct taxes and cash benefits, depends on which one you're looking at.

If we were to focus on cash benefits, we've seen that redistribute effect at the end of the period as much the same as the start, most noticeable is its cyclical nature. And this reflects the changes in average rate rather than changes in progressivity. So basically, the intuition is just the automatic stabilisation role of cash benefits and this is really important. So the economy goes into recession, cash benefit spending for workless people rises, cash benefits form a large fraction of household incomes on average. And as the economy recovers, the automatic stabilisation role goes in the reverse. So that's effectively driving what we saw there. I'm going through it, emphasising that, because one of the big discussions in the UK is that there's been a long run decline in the real value of benefits for workless people relative to average earning and that's continuing to go down and down and down. But if that factor was going to be having an effect on average rates, the lower case B, it's not showing up here, it's not showing up as a driver of trends in redistributive effect at the moment. It's dominated by the cyclical aspect.

On the other hand, things are different with direct taxes. Redistributive effect, relatively stable up to the beginning of the '90s but then it got bigger, almost doubled just up to the pre COVID period. And these changes are basically reflecting changes in progressivity. So average direct taxes are rates are rising with income to a greater extent than previously. And what's driving these? Well, here, it gets rather complicated to say exactly what's going on and that's because direct taxes are of multiple kinds. There's taxation on earnings, there's taxation on other forms of income, there is employee National Insurance Contribution rates. So in the lower bullet points, I've picked out a few elements, for example, just to illustrate some of the many changes that were underlying what was going on here.

So basically, in the '80s and '90s, the top rate of income tax, top marginal rate in the UK was capped at 40% in 1988, having been previously between 40 and 60%, but thereafter, the maximum rate was around 50%, slightly different depending on which source of income, and then down to 45%. Okay. But over the whole period, National Insurance Contribution rates, might think of as a silent tax, more silent tax, but increased gradually from just under 6% in 1977 to almost 12% in 2018 as well as there being an additional higher marginal rate. A percent more in 2003 going to 2% in 2011 over this period. So if we actually look at the average tax rate that is paid by different income groups, split the distribution up into 10 groups, the 10 decile groups, we can see, for example, not much change at the bottom, but if we look at the top, the steepening between the 9th and the 10th decile groups, so right at the top, widened quite a lot.

So my final conclusion is basically to step back and think about the analytical issues and potential limitations. As I've just said, the analysis does not directly reveal, precisely which tax change or which benefit change was responsible. There are lots of direct taxes and boy, I didn't tell you about the 30 odd cash benefits that are there. Now, Kakwani's formula can be used to look at these in detail, but frankly, I would only want to do that in one cross section. To try and do it over time, would be a heroic endeavour.

So I think to make better progress and try and pin down precisely what each tax does or each tax change, policy change does, ditto with benefits, what you should do is be less heroic, reduce the breadth in terms of the time span, and focus on particular tax benefit components. And this brings us much closer to the work that people conventionally do with tax-benefit micro simulation models and the counterfactual analysis that's based on that. And Nicolas, my co-author, has done this sort of thing and I'm sure Patrick and his team are doing quite a lot in the Treasury indeed before every budget when taxes and benefits get changed.

Something I haven't talked about also is what determines the changes in the pre-fisc distribution. For cash benefits, I cited the business cycle as responsible for the changes in the distribution of market income, but that's hardly a particular deep interpretation of what's going on. It’s simply describing what's going on rather than saying what was driving things. So what's changing those pre-fisc distributions? The market income distribution is particularly important. It's something that Thomas Piketty and others are also emphasising at the moment. The hard-bitten economists among you will of course also be waving your hands about and saying, "Hey, what about behavioural responses?" There ain't any in here. This is arithmetic calculations.

I would argue they're informative but yeah, behavioural response could be important to take account of. So the bottom line is, or almost an ultimate bottom line is that this decomposition approach is a complement to other approaches. It's got advantages in terms of long term big picture looks but needs to be complemented. And the bottom line, I suppose I better say this in this seminar series, is that to take us back to where we were at the beginning. Income is not the only aspect of wellbeing we should be bothered about. So that's it, Patrick. Thank you.

Patrick Nolan:

Great. Well, thank you, Stephen. That was a really entertaining and useful presentation. I mean, you quickly ran through the different elements of the Kakwani decomposition, the average tax rate to progressivity and the reranking effect. You talked about the different ways of the defining income, so moving from original to gross income to disposable income, which was the focus of your work. I see a few of the chats, a few of the questions in the chat are about those broader definitions of income moving towards sort post tax and final. You then ran through the redistributive effect, the importance of actually understanding what counterfactual you're looking at and looked at a range of different Gini coefficients and the different trends way back to 1977.

So we've got a few questions in the chat. I would like to treat them in a chronological order because I was... Ben Ching asked the question early on, "Is the increase in direct taxes in the Kakwani and it's purely a function of more progressive rates or can it also reflect base broadening?" And you touched on this at the end, but I thought it was quite interesting. I was quite surprised that the UK... I didn't know the UK tax scale that well. Actually, my impression was that the earlier rates, the personal rates in the individual income tax scale were higher in the '70s. So I was surprised but obviously, there's the effect of National Insurance Contributions. But how much of it is just that change in the rates versus that widening of the base? You mentioned that shift in the unit from the families to individuals. So how much of it's rates versus bases do you think?

Stephen Jenkins:

It's a really good question and it comes to the real problem of trying to disentangle what happened over the time when so much happened. And one of the sad facts of Nicolas and I, our paper when we were trying to disentangle that is that there is actually very hard to find a catalogue of the changes that have been done over this 40-year period. We've relied a lot, we put together stuff from the Institute for Fiscal Studies catalogue essentially, talked to one of my colleagues, Kitty Stewart, who's worked on these sorts of things. I've got more detail in the paper, but I think what I've told you is essentially what those catalogues, I'm calling them catalogues, what our assemblage of information was, what's going on. Yeah, playing around with rates rather than broadening, depends what you mean by broadening, of course. But playing around with I think what would call be called New Zealand tax free area, the rate at which people start paying tax, is something that's going on as well. And the fiscal drag is something that the UK has used quite a lot.

Patrick Nolan:

Okay, great. There are also a few questions, I guess, in a related sense. So not just looking at earned income but thinking about of income more broadly. So Nazila's asked about including capital gains in the measures and I saw that there was another one related to... Oh, yeah, well, capital gains I guess was the only one I could... And indirect tax as well. So yeah. So that's taking you from that disposable income to a fuller, closer to final income.

Stephen Jenkins:

Exactly, yeah.

Patrick Nolan:

Is that something you've also been thinking about or is that... What do you... Yup.

Stephen Jenkins:

You got two main things there. Let me take them off, but before I do, I just saw a question from Hamed flashed up saying, "Please, can I have the catalogue?" All that we know is essentially summarised in our paper and the references that we cite. So there's an IZA Discussion Paper that I think I sent through that has just about published in the Journal of Income Distribution and a special issue in honour of Nanak Kakwani. But the journal, unfortunately, is very expensive and hard to get a hold of. So have a look at the IZA working paper which is free.

Yeah, go back to those questions. Yeah, capital gains, should they be an income? Good question. Debated a lot. The original Canberra recommendations do not have... They have it in... Should be there in principle of course from a [inaudible] point of view, but in practice, it's recognised that it's very hard to include them. Certainly, the UK has capital gains taxation of the sort that if you can put your income into that, you should be doing that rather than to employ earned income tax because of favourable rates. But the bottom line is that at present, there is no routine headline series from the Department of Work and Pensions or from the Office of National Statistics that routinely includes capital gains. Part of the problem is the usual ones about valuation and assessment of them and if American experience is anything to go about, the official statisticians are always also wary of it because capital gains tend to fluctuate a lot more than do other income sources.

So you get something that's jumping around and the issue about whether or not there's noise, but it's important stuff. If you look at Thomas Piketty and his team's work on the share of total income that's held by the richest 1%, certainly for the stuff for the US and recent stuff by other people for the UK, you'll find that the share of the richest 1% is rather higher if you include realised capital gains. If you're more interested in that, then Rich Burkhauser and colleagues have just had a paper published in Journal of Political Economy where they argue for accrued capital gains. Rich is a mate, but I'm not sure I support his accrued capital gains. Partly, if you look at, again, the valuation issue and if you're worried about cross time variability in the series, the accrued one really jumps around a lot.

Yeah, indirect taxes, that's another issue. If you have a look at the paper, we go through and produce all our series for indirect taxes. So that's primarily, of course, value added tax, which has increased over the... The rate has increased over the 40 year period. But it's also various other duties, not just stamp duty on taxes, but the in your face ones, the sin taxes on cigarettes and booze and all that sort of thing. More complicated to assess. And from an analytical point of view, there's a big issue about whether or not you should be assessing the progressivity of indirect taxes, which are levied on spending, which you might think of as consumption about whether or not they should be, according to the Kakwani thing, should they be in there and put against income because people's ranking in the income distribution differs from where they are in the spending or consumption distribution.

And so yeah, we've produced the numbers, we've saluted the fact that this is an issue. Again, it's another version of the what should be the reference distribution issue. But we focused on the other aspect of it. If you're interested in this, there is further... There is somebody who's just published a paper in fiscal studies, I forget his name, where arguing for look at progressivity of indirect taxes in relation to consumption rather than income and has an example. But again, doesn't view the 40-year stuff.

Patrick Nolan:

Great. Thank you. So we've had a few other questions, one by Tim Hughes around subpopulations. And I know this is something Meghan Stephens in my team is looking into. The degree to which changes in the income distribution over time can be attributed to things like population ageing. For example, we see that super annuitants are increasingly having not just New Zealand super as an income but also market income. And so that's actually changing the dynamics of actually how the overall income distribution. So is that something that's come up in the UK as well?

Stephen Jenkins:

Yeah, I mean, those things are important but because they're slower moving targets, population ageing is a real thing, but it takes place at a lots more glacial pace than there's a Chancellor of Exchequer standing up and say he's going to abolish all the top tax rates overnight, which you may have heard of. Another way of thinking about this point is that we should be thinking more about the drivers of the pre-fisc distribution. It's, for example, related to old age populations in the UK. Another issue that's important for tracking the income distribution is a growing propensity for people who may not have any income at all. Not just because they're rich, it's because they don't actually need income, they just draw down their assets either because they're really rich or they can just draw them out or if they're poor, they're taking it out of private pension schemes and so on. So actually, for that group producing a problem later down the track.

In terms of the subpopulation thing, I guess you could do these exercises for different subpopulations, but the sort of, you'd have to do them consistently for each subpopulation because you couldn't add them all up again. The problem is that the income distributions for different groups overlap with each other and that's a real problem for the Gini coefficient. And we know, for example, the income distribution among old people is... Or the, let's say 60 plus is a very unequal distribution as much as those for the under 60s that are both rich pensioners and poor pensioners, which complicates things.

Patrick Nolan:

Okay, great. Thank you. Yeah, some of the big mega trends, aren't they, the big questions, how we think about these longer term changes. Also, so Tim has also asked, I'll read out his question, "The OECD publishes figures for gross, original, and disposable Gini so we can compare redistribution across different countries. How much can we rely on the picture painted by these comparisons? And do you think, can we conclude much for how New Zealand redistributes compared to other countries?" And Hamed has also name dropped my brother nicely. So was my brother's findings for New Zealand in line with what you presented for the UK? So basically, can we trust the international comparisons and where do we stand?

Stephen Jenkins:

Yeah, sure. I should have looked that up. I didn't. So Hamed, the problem with Matt's decompositions is that it's not actually the Kakwani decomposition. Matt's focus is on some gory details, in particular, trying to distinguish a concept of horizontal inequity from that overall reranking term. And to do that following the approach espoused by Peter Lambert and a colleague, it means that essentially, it's not directly comparable. The other thing is that the time period over which he's looking is relatively short by comparison with us. So I'm not sure. So I'm aware of the work but I didn't bother to do a direct comparison because I didn't think they were really comparable. But on the other hand, maybe Patrick can go home and twist Matt's arm and say, "Do some Kakwani stuff just to see."

OECD. Yeah, the OECD team produces all these numbers. So you can do all this stuff to look at redistributive effect. What you would have to do, however, to go and do the decompositions is that you would have to... That in the way that Nicolas and I have is that you would then need access to the unit record data for each of the countries to do the decompositions. Now, maybe you're not interested in that, you just want the overall estimates of redistributive effect, maybe. And at that level, I think yeah, OECD, it's fairly reliable. Another source that you could go to is the LIS data centre, so called. What used to be known as the Luxembourg Income Study, and have a look across a lot of different countries. The advantage there being is that they LIS-ify, i.e., harmonise data from multiple countries all around the world, increasing number of middle income countries as well as the usual suspects. They have the GOD, the G-O-D distributions.

So you could actually do the exercise for LIS but not New Zealand. New Zealand doesn't participate in the LIS and has long not done so. Many years ago, Helen Stott, who used to work for Statistics New Zealand, was involved with somewhat trying to see what was feasible but went through, let's say, a black period in New Zealand data. So let's cross our fingers that hope that NZ can re-enter the fold down the track. It's in the OECD figures, of course, though not all of them. Not all of them.

Patrick Nolan:

Yeah. Great. Thank you for that. No, it often comes up that we're missing from some of these international surveys and it seems like it should be the thing that we should be putting more into, but that's above my pay grade, I'm afraid.

Stephen Jenkins:

Well, if the worry is about having access to the unit record data, remember that the way most of the OECD stuff is derived is by a country correspondent working to a template. So it's not as if... If we can finance somebody to go into your secure data centre, then it's straightforward.

Patrick Nolan:

So getting into some of the data issues, so Meghan has asked on the reference distribution for tax, have you considered using taxable income, accounting for the fact that some benefits are taxable but others aren't?

Stephen Jenkins:

Love to, Meghan. Too difficult because the data sets that we have don't allow the disaggregation in the clear taxable versus non-taxable. And as I say, there is... Well, on the show card for the Family Resources Survey, there used to be 37 of them variation. Yeah. So I have to hands up, Nicolas and I opted out. I mean, we would like to be able to do taxable income rather than gross income. This is our first approximation.

Patrick Nolan:

And a related question, so Meghan also asked another question, and this is something that challenges us in A&I, is have you come across any issues to changes in survey data collection processes? So we find this a particular challenge when we try to do our work and we try to look over the income distribution over time. We do see some breaks in the series. So I'm just wondering if that's been... Especially when you're going back to 1977, there must be some challenges there.

Stephen Jenkins:

Yeah, well, you saw the big break in the sense when I changed the name from Living Costs and Food Survey from the Family Expenditure Survey. So there've been, in fact, there was something in between that was called something else, but they're essentially the same thing. I think the income questions have stayed pretty stable over time, particularly at the aggregate level. And the main adjustment, and this comes back to Meghan's earlier question, is benefits come and go and they change their names. A big change over time and the issue of, for example, over this period is the interaction by Gordon Brown and his new Labour government with tax credits about how to treat tax credits. And the UK position on this is basically to treat them as like cash benefits. And so they go in there after all their tax pool. I suppose it's not actually the survey’s issue over time, it's actually dealing with the real world out there that keeps changing, which is a problem.

There've been more changes, I think, on the data side for collection of expenditure data and the way in which the modes that are used for that. The one final thing related to potential changes over time is of course related to COVID. And so for example, I'm not sure what happened to the Living Cost and Food Survey, but I think it was the same as for the Family Resources Survey. Mainly, it switched almost immediately to telephone interviewing and is still not... Which is rather than face to face. And interviewing every adult in the household. So that was a huge change. And on the DWP Advisory Group that I sit, we spent the last year or so trying to work out whether or not the numbers that were coming up from this COVID mode were sufficiently reliable to release new headline statistics. And basically, bottom line was headline statistics are okay but the gory details that are normally published, no.

Patrick Nolan:

Okay, great. And I see Meghan has responded to your answer to the earlier question. She said, "Good to know that the taxable income reference would be useful. We could do that so I could compare." So there you go. She's offering to do some work for you if that would suit you. So there you go. So that's good. So Luke has asked, Luke Symes from my team, he's the one who's done a fair bit of work looking at the Gini and the impact of house prices on the Gini coefficient, which we were discussing before. And he's asked, "Could we do a similar decomposition analysis of redistributivity for the absolute Gini - average dollar gap between pairs?"

Stephen Jenkins:

I guess so. I mean, absolute Gini is just normal Gini divided by the mean. The trouble is that you'd be doing redistributive effect. Do you want to look at differences in absolute Gini? I'm sure you could do it, it's just they'll be normalising by a series of ratios of means to get you to do the washing out of those things. I haven't seen it done before. You could do it. Or if you really wanted to get fancy, you could also do a generalised Gini. Okay, so use an index that was more sensitive to income differences in different parts of the distribution. But yeah.

Patrick Nolan:

Okay, great. And so Tim Hughes has asked again, "I know you've also looked at the relationship between inequality and intergenerational mobility in the past. What kind of inequality is most closely related to immobility and how solid is this relationship?"

Stephen Jenkins:

Oh, gosh, Tim, take me back.

Patrick Nolan:

Good questions.

Stephen Jenkins:

You've flustered me now. Good inequality and bad inequality. Now, we're getting difficult. My colleague at the LSE, Chico Ferreira, likes to make... His analogy is based on the difference between inequality of opportunity and inequality of outcome and talking about... Because he is particularly interested in inequality of opportunity, saying that that's essentially the bad one, whereas inequality of outcome might be forgivable if the incentive aspects that are often associated with people reaping the rewards that they earn has an effect. So essentially, to work out the bad inequality from total inequality, you have to work out how much inequality of opportunity there is using. And there are standard ways of doing that. I could elaborate if you like.

The problem is that it's not directly the same as social mobility as you and I might normally think about it in terms of the correlation between outcomes across from parents to child. The concept of inequality of opportunity and intergenerational income mobility are the same only in some rather special circumstances, including, for example, whether or not parental income is the only thing that you want to take into account in the parental generation. But usually, it's other stuff. It's also sex, ethnic group, and of course, that's particularly relevant to New Zealand, and a load of other factors. So not just trying to avoid your question, Tim. I'm trying to talk my way around it.

Patrick Nolan:

Okay. That does raise another question which Hien Nguyen from my team has put, not in this chat, but in my team’s chat, she's put, "What is the rationale of the ONS' modified OECD scale to use two adult households instead of one adult household?”

Stephen Jenkins:

So for a bit of background, for those of you unaware of it, essentially, well, we know what the modified OECD scale is, we know what the Jensen scale is and all those things. We know about the per capita scale. And indeed, in the UK, there used to be a thing called the McClements scale. So the issue is when we look at inequality, we look at poverty, we're typically interested... We're doing... We use an equivalent scale to do a proportional adjustment. So we set the equivalent scale rate equal to one for the household type, reference household type, and then everything is a multiple of that. Now if you change the reference household type from a single person to, say a childless couple, it doesn't actually affect anything in terms of using relative inequality measures, relative poverty indices and so on and so on. It's just because everything is scaled.

So why is it done then, you might say? Well, I think the answer is history, UK history. What they used to use the McClements scale, the tax unit used to be married couple or an individual. So maybe I would think in terms of childless married couple, but actually, it doesn't make any difference. It just makes it damn complicated in doing comparisons in terms of equivalent scale rates across the country. You'd have to divide everything by .61 to get it back to the reference adult being equal to... Sorry the childless single adult living alone back to being the reference household. I think that's all it is Hien and Meghan or is there something on top of it that I'm missing?

Patrick Nolan:

Yes, no, I think that's a persuasive answer. So can I ask a question? I mean, we've been talking a lot about, I guess, kind of the detail of the approaches. I mean, taking a step back, one thing I think in A&I we're quite excited about is the increasing potential and the availability of data and we were discussing before the session, the degree to which data is linked and what we can now do with linked data. If you were to say, think ahead five years, what do you think, where do you think the new areas of inquiry would be? Where do you think we should be prioritising our investments? And given the fact that we're now able to see this administrative data and actually do these increasingly powerful decompositions, where would you recommend that we start investing?

Stephen Jenkins:

Well, I think we'll join the... Well, begin following... New Zealand, small place, but it is actually... The IDI is quite innovative and there are already countries in Europe, of course, the Nordic countries are out there already with linked registers where they're linking together everything. And importantly, linking together not just income, but also education registers, household registers so that you can actually get individuals together within a household. But Norway, Sweden, Finland, and Denmark are a bit unusual. The Netherlands is doing more. So I guess I'd make a distinction between countries for which you can do standard income distribution analysis without surveys at all.

Just work with linked registers and those where we're going to be enhancing surveys and enhancing surveys for two reasons in particular. One, we know that collecting income data of any kind and many other sorts of stuff is a respondent burden. So we can save time. Saving time also saves money. And you can collect information and do data substitution. So a pioneer in the field was, for example, the Canadian Survey of Labour Income Dynamics, that's SLID. Basically, the idea of the way the data was collected in the SLID on income was to go into the household and say, "Hey..." I translate roughly. "Hey, we're going to ask you a load of boring questions about your incomes and earnings that'll take 20 minutes or half an hour." Or you could say, "Can we link your data to your tax returns and other registers that they have?" And so they've got a significant proportion of people signing up to the linkage. So that's a form of data substitution.

So I think that form of data substitution is going to be increasingly used. I should also say that it's incredibly context specific. It seems okay in New Zealand to do it. It's clearly okay in the Nordic countries, in some European countries. The UK has only just recently changed its Statistics Act. That means that, for example, in the Family Resources Survey, if you want to link data, you no longer have to ask a question at the end of the survey that says, "Do I have your permission to link your data to the data held by Her Majesty's, it was Her Majesty then, revenue and customs?" And to which some respondents, about half of them would say no.

So under the new Statistics Act, there is a public benefit idea so that some form of linking is done. Not using the equivalent of tax IDs, National Insurance Numbers, because they're too error ridden to collect in the survey. But you can use various probabilistic linking, ways of linking. And so that's going to be the way ahead. It's going to be... But that takes a change, a societal change where people are somehow comfortable with doing this stuff. And it's a big deal. And maybe when you walk around a country when there are cameras everywhere checking on everybody, then to complain about linking of your consumer data or the data that you give away to Facebook or places like that. Then where data security and those issues are of a different order. But yeah, linking is going to be the way ahead.

Patrick Nolan:

Great. Well, thank you. I'm reminded of a quote. Was something that Meghan wrote, she got a quote, it's in the context of measuring poverty, describing it as the art of the possible. And I think it's very true as well for measuring inequality. Actually, anything to do with income distributions. And what is possible is changing over time. So it's great that we’re able to benefit from your expertise and your experience, Stephen. So I really appreciate you making the time today. As Meghan's already offered to do work for you, so we'd love to keep in touch. We're learning, but we're trying to do more in this space. As I mentioned, we're doing, increasingly thinking about how we can start to measure final income. So looking at total fiscal incidents. So include measures of spending in kind as well as indirect taxes and that to our micro simulation modelling.

So hopefully, that would be something we could keep in touch on and that you might find of interest. So I just wanted to thank you for making the time today. Thank you to all the attendees as well, and particularly those people who asked questions. Please do keep a lookout for upcoming events in the seminar series. And in particular, there's a milestone event coming up on the 24th of November. We’re hoping to host a seminar to launch the Treasury's first ever wellbeing report. So please be sure to head to our website and reserve your spot to be amongst the first to hear the findings. So it'll be excellent. So I'll just finish with a karakia.

So, [speaking in te reo Māori] Piki te kaha. Piki te Ora. Piki te Wairua. Mauri ora ē.

So thank you again for attending.

[speaking in te reo Maori] Mā te wā.

And I guess I'm not quite sure what the protocol is. If we're in a room, we'd all clap and thank Stephen in the usual way. But if you could please thank Stephen maybe with the reactions and we'll close the event there. Thank you.

Stephen Jenkins:

Yeah, if anybody feels free to email, that'd be great. And thanks very much for the really intriguing comments. I value them. Good to see you guys. And I'll be back in New Zealand next year, so I'll be... See you then. Bye.

Patrick Nolan:

Thank you. Thanks.

Stephen Jenkins:

See you Moira. Bye.

Patrick Nolan:

Bye.