 Let's move on to the next session. So even though this is a very minimal role for me, I was very pleased to be asked to be involved in today's event because I think this topic is extremely important for central banks and also for the economics profession. Of course there's a strong overlap between central banks and economics but let me emphasise is central banks have many more occupations and professions than just the world of economics. So Andy's particular diversity challenge actually varies quite a bit across different parts of the organisation which ensures true at the Bank of England as well. And in this session I think we're very pleased to have Andy Holden from the Bank of England. Andy has been I think a leader in this programme of work globally in terms of diversity in central banking and I'm sure we will look forward to listening to his contribution today which is focused on the UK data. So without further delay, let me hand over to Andy. Well thank you very much. Round of applause for an even set of words. That's always nice. Thank you Philip for those kind words. Let me just start by congratulating the organisers for this tripartite ECB bank-fed conference. The second of its kind. It's been absolutely spectacular. Congratulations to all the paper presenters. I think as Gideon said in the previous session, some outstanding papers today that have really kind of built out our evidence base even further to make the case even more compelling than it was previously. I thought what I'd do today is talk about one dimension of this issue. I want to try and pose and hopefully attempt to answer the following question. You have someone who is otherwise equally skilled and equally experienced doing the same thing in the same place. They differ only by their gender or their ethnic background. Why should they pay differ at all? The answer of course is that it shouldn't in that situation. Those pay gaps, if you like, ought not to exist but do exist. What I thought I'd do is shed a bit of evidence on the size of those pay gaps using, as Philip mentioned, some UK evidence. How have they evolved over time? How can they be explained? Ultimately what might be done to shrink those pay gaps back to around zero, which is where they should be given the way I described them at the beginning. This is joint work with three colleagues from the Bank of England who are named here. You will see that we have a perfectly diverse set of co-authors. They did almost all of the work. My role is very much that of the token white male today. A question was asked in the previous session. How could we encourage more white men to come along and talk at events like this? Why am I here? I was invited, which is very nice. It's an issue on which I find intellectually fascinating, but more than that about which I'm very passionate, indeed proud to be talking about. Truth be told, we are all minorities some of the time. We've all been in situations where we felt our face hasn't quite fit. Diversity comes in all shapes and sizes, Rhonda said earlier on. And I think viewed in that way, arriving, coming and speaking at events like this should not be seen as a challenging thing for white men, but as a real opportunity. The last three or four years, most of my time, or a chunk of my time, doing something that central bankers don't do very much of, which is wandering around. Wandering around the UK, seeking out and talking to as disparate and diverse set of communities as possible. Not the sort of people that central bankers typically talk to. My one learning, one learning from that three or four year experience is that the key to engagement, the key to inclusivity is to put yourself consciously in a position of vulnerability. The words of situation where the agenda isn't set by you, the questions aren't set by you. We certainly don't have all of the answers. The issues come from the other side of the table. The language is set by the other side of the table. That you spend as much or more time, actually, listening as you do talking. You are leaning in to the uncertainty, leaning in to the vulnerability. And I think the more people, central bank or otherwise, who could act in that way, the more encouragement we could get to get them along to conferences like this, which I think is absolutely crucial. Let me turn then to the topic. And it's pay gaps. In other words, does the same work pay the same wage? This is an issue that's of some policy import and topicality, including the UK. Sarah mentioned in the previous session that since 2017, companies with more than 250 employees in the UK are required by law to publish annually their gender pay gap. There's a consultation out at the moment in the UK asking whether something similar might also be done when it came to ethnic minority pay gaps across that same set of companies. Similar initiatives are underway internationally in places like Iceland and Denmark and elsewhere. Different detail, but the same basic principle. That is to say, should we require, or at least encourage actively, individual companies to publish any gaps that exist between how much they pay, whether it's girls versus females or whites versus non-whites in their work force. At the Bank of England, for our part, we started publishing our gender pay gap one year ahead of schedule and already started publishing our ethnicity pay gap even though that is not yet required. I can tell you that publication listed a degree of outside interest, indeed a degree of outside criticism, some of which was probably the unfair side of neutral. But nonetheless, I can tell you, was a very useful nudge. But more than that was a useful dig in the ribs for us to explain why those gaps existed. Essentially, in our case, it was a question of too little senior representation by women and ethnic minorities in the organisation. But as importantly, it catalyzed us to craft and indeed put renewed energy between our efforts to fill those gaps, to close that pay gap over time. In other words, disclosures, in this case of pay gaps, served as a useful disinfectant, as an incentive device to individual companies and institutions to not just account for those gaps, but crucially to put in place a plan for closing them. Now, those disclosures are present company-specific and cover only a point in time. So I thought what I'd do is to take a different database and assess those pay gaps over time to see how they've evolved and additionally, at a national level. And third and finally, to try and explain why they might exist. Could it be justified in terms of characteristics of the people or of the jobs those people are doing? And that's roughly the plan for my presentation. Here it is set up with a bit just on the data. Then talk a little bit about these pay gaps, untouched, these unconditional pay gaps and how they've evolved over time. I'm going to cover off both the gender dimension but also look at the pay gaps for ethnic minorities in the UK workforce as well. Why do that? Well, of course, it's of interest in its own right but also there's a crucial interaction between ethnic minority and gender. A double pay problem, if you like, for ethnic minority females. Then I'm going to go on and seek to do some job of explaining what lies behind these pay gaps. How can they be accounted for in terms of characteristics either of the workers or of the jobs that they are undertaking? I'm going to use two approaches to do that, both of which have been mentioned and used in papers earlier on today. The first is a decomposition and of the many and various ways of doing that decomposition, the one I'm going to choose is this Oaxaca blinded decomposition. That's not for reasons of cystical robustness. The real reason is because I really, really like saying the word Oaxaca. You should really get your mouth around it. Even saying the word now, I can feel my pulse slowing, my heart beat dropping. A certain sense of zen comes over me whenever I say the word Oaxaca, which I will now say as much as possible during the chorus. Say to yourself under your breath now, Oaxaca. It helps no end. This is peak zen of the presentation. It's all downhill rapidly after this, by the way, because I'll launch into some of the data and the econometrics. But about two-thirds of the way in, I will come back to Oaxaca and a certain warm glow will cross the audience, a little dopamine rush as you hear the word. All the previous stuff will appear. It's just a bad dream, really. Oaxaca and then some cloudy golden type regressions of how we explain pair patterns over time. And then finally, what might it mean for policy? Given that I'm a policymaker, I want to put out there a suggestion or two about not just what has happened, but what might be done to close these pay gaps rather more rapidly than might otherwise be the case. So these are the data. They're for the UK. They're over the last 25 years or so on individual workers within the UK labour force. That gives us a slightly north of half a million observations on individual workers in places that's nudging up to around a million observations on individual workers. What this database gives us in particular, which is useful, is a very rich set of characteristics, both of the worker, their age, their educational background, their gender and ethnicity, but also of the job they are undertaking, which sector, which occupation, part-time or full-time, et cetera, et cetera. Let me talk very quickly about some of the characteristics of these data over time. This just looks at the extent of female participation in the labour force across the G7 countries. The UK here is in green. Broadly speaking, more or less, as most of you know, all of you, I'm sure, there's been generally speaking a rising tide of female participation in the labour force. The UK is no exception to that. In fact, having recently overtaken the US, we're now second only to Canada in the extent of that female participation in the workforce. The self-same thing is true of ethnic minority participation in the UK labour force as well. Other things equal. You'd expect that rising tide of representation and participation to have begun to chip away at pay gaps, whether gender or ethnic minority-based. There's also been a rising tide of educational attainment. This divvies up the sample four ways, white males, white females, and ethnic minority males and females. For all four cohorts, educational attainment has risen very rapidly over the last 25 years or so. Let's just take the bottom blob, the dark blue part of the pie chart. That measures those with a degree. Between 94, 2019, that blue bit has increased very materially. Indeed, it's increased much more materially for women and for ethnic minorities than it has for white males. To such an extent that as of 2019, levels of educational attainment among ethnic minorities and women are higher than those of white males in the UK. We sit at the bottom of the league table when it comes to educational attainment. That is not true if you take a different cut of this cake, which is looking at the skills, the professions of those same four cohorts. Again, the dark blue part of this chart refers to the highest paid managers, professionals, and associates component. That too has grown in size over time across the four cohorts. Interestingly, the numbers in that group are larger for white males than for either women or ethnic minorities, despite the fact that their levels of educational attainment are lower. At least suggestive that some pay gap, gender or ethnic minority basis might exist. Let me turn now to those pay data. This looks at the distribution of male hourly pay. This is a female hourly pay left and right on two dates, 97 and 2018. Those distributions shift as you'd expect to the right over time as pay picks up. Comparing those two distributions, it's very clear this peak here, can you see this peak? This peak here, which is the peak of lower pay among women, is maturely higher than this peak of lower pay along men. And equivalently, the upper peak of hourly pay is much fatter for men than it is for women. All of which is suggestive, at least of a gender pay gap. In fact, on this sample, that gender pay gap over the last 25 years or so averages just north of 20%. This looks at that gender pay gap over time, whereas we move from left to right. This is the sort of time series. I'm plotting now that the gender pay gap against female participation in the workforce. The basic story is of this gender pay gap, having fallen from peaks of close to 30% in the late 1990s down to around 20% as you enter the noughties. Strikingly, there has been precious little erosion or diminution of that gender pay gap pretty much since the dawn of the global financial crisis. We can cut this gender pay gap any which way. Let me cut it one or two ways to illustrate. This cuts it by level of qualification. That's running horizontally and also distinguishes gender pay gaps inside London and outside of London. The key takeaways here are those gender pay gaps tend to be larger. The lower is the level of qualification of a worker. They also tend to be a bit lower inside London rather than outside of London. Turning then to ethnic minority pay patterns, here again two distributions on two dates across the two groups. You'll see it's far less easy to discern much of a difference in these two distributions as between whites and non-whites across the workforce. That's pretty much borne out if we look at a time series of now the ethnic minority pay gap over time. Again, running left to right. Generally speaking, that pay gap has been at a lower level than the gender pay gap, average around 4% over the last 25 years. That's the good news. The less good news is that it shows far fewer signs of having reduced over that self-same period. Again, a few different cuts of the cake, rather different than on the gender side. If anything, the ethnic minority pay gap is larger among those with higher levels of qualification. What's more, it also appears to be larger in London than is the case outside of London, which by itself is a little strange given that fully 40% of the ethnic minority population in the UK is resident within London. There are some puzzles in this data as to what quite is going on. Rhonda just made in the previous session this very important point about disaggregation. This chart disaggregates the ethnic minority pay gap across different cohorts, and you'll see from them a very mixed picture among people with different ethnic minority backgrounds. For those with a Chinese or Indian or mixed race background, their average pay is actually higher than that of the equivalent white worker, roughly around 15% higher. In a way that's not true of, for example, Black and Afro-Caribbean, Pakistani or Bangladeshi origin workers in the UK. For them, a pay gap of around 15% in the opposite direction. So those are the raw data, the unconditional pay gaps, the like of which are being published by companies across the UK as we speak. But to get underneath the skin of what's really going on here, what I want to try and do now is to bring to the table some of the characteristics of either of those workers or of the job they undertake that might account and explain for those patterns and that evolution of a time, and those to construct some conditional measure of the pay gap using a worker-specific characteristics, for example, educational background, age, tenure in work, whether they have children under the age of two, and some job-specific characteristics such as the region of the job, the occupation, the sector, whether it's full or part-time, etc. To be absolutely clear, even if I can explain account for those pay gaps with these factors, I am not suggesting that makes them justifiable. If the origin of them is a huge gap, a gulf in educational attainment between, say, males and females, that is not justifiable. But what this accounting decomposition and explanation can do is be suggestive of where policy action might best be directed. And as importantly, if we can't account for these gaps, even taking account of these factors, we have a very pure measure of pay bias. You might even call it pay discrimination, either by gender or by ethnicity. So what do these decompositions or regressions suggest? Well, Andy, what type of decomposition are you going to try? Well, I'm going to do a wahacca, there you see. I told you, a little warm glow, depamine rush, a wahacca decomposition of, in this case, the gender pay gap. And on the right-hand side, I've broken that down into the bit that can be explained by those factors, worker or job specific, and the bit that even with those factors cannot be accounted for. You will see that it's roughly 50-50. We can explain some of that gaps in terms of looking at the left-hand side picture, in particular, the job specific characteristics, the sector in which the job is and the occupation that people are working in. Nonetheless, that still leaves a roughly double-digit pay gap that remains unexplained, a pure measure, if you like, of pay bias. This tracks it over time. The grey here is the unaccounted fall, unexplained pay gap. The good news is that that's been falling over the last 25 years or so. The bad news is that even by the end of the sample, that pay bias remains in double digits. If we do the self-same, that's just plot-set using regression techniques, rather than Whacka, it tells the self-stame story, roughly speaking, a halving of the gender pay gap over time, but remaining at around double-digit levels, even at the end of the sample. If we do this for ethnic minorities, the picture is actually rather different in the following sense. Those characteristics of the worker and of the job for ethnic minorities ought to be pushing their pay above that of equivalent white workers. That's what the left-hand side part of that picture is showing. In other words, the raw, the unconditional pay gap data for ethnic minorities understates the extent of the pay problem in the workplace. The unaccounted fall, conditional pay gap for ethnic minority workers is in fact as large as it is for women. The reason why an average ethnic minority should other things equal be paid more than the equivalent white workers is that relatively more of them work in London. That's the orange part here, and of course they're also on average better qualified educationally than their white equivalents. Plotted over time, the grey part here is the unexplained the conditional pay gap. It is around double digits and what's more and what's worrying, it shows far fewer signs of having diminished over time despite the rising representation of ethnic minorities in the UK labour force. So summarising if you like these conditional pay gaps blue here is gender that gender pay gap, conditioning of the stuff has halved over time but remains large in double digits the ethnic minority pay gap that too has fallen but by far less and it too remains in double digits and for ethnic minority females that's the yellow line here the good news is that that gap has also halved over the last 25 years but from a much higher level and remains maturely higher even today at closer to 15 percentage points. You'd also use this sort of analysis to explore the impact of different of those factors I mentioned on different cohorts by running different regressions on different cohorts white versus non-white females versus males we explore the relative impact of different factors on those different cohorts just to run through one or two of these educationally we know there's a very significant pay premium for graduates but it turns out that graduate pay premium is maturely larger for men than it is for women we know there's also an age related pay premium it turns out from our regressions that pay premium is maturely larger is for women at age say 45 to 49 that's around a 10 percentage point gap more paid to otherwise identical men than to women for part-time or non-permanent work interestingly the pay discount or pay penalty there is in fact larger for men than it is for women at the very start of the conference today for men and women with children under two there is a pay premium for men but a material pay discount for women particularly women who are working full-time and finally just the impact of the sectoral dimension the financial sector in which many of us work there is you'll be unsurprised to hear a pretty fat pay premium from working in finance relative to any other sector but if you're a man working in finance that fat pay premium comes in at almost double that of the otherwise identical woman working in that sector so much then for the diagnosis let me skip on with some reflections on what we make of this and most importantly of all what do we do about this what further purpose of policy action could be taken to make greater inroads into these still large persistent pay biases or pay gaps well I mentioned one or two things here one thing would be at present I mentioned this new regime this legal regime requiring UK companies with more than 250 employees to disclose their pay gaps of course many companies have far fewer than 250 employees around 60% of the UK workforce working firms with less than 250 employees so one obvious policy extension would be to extend the reporting to include a larger number of those small firms a second would be this consultation on ethnic minority pay gap reporting this analysis would suggest there's a strong case for making that not voluntary but mandatory and indeed to do some disaggregation of the type that Rhonda mentioned of the cohort effects within that ethnic minority population if you were being more ambitious still let's ask the question why don't we make this a standard internationally why don't we harmonize how we disclose these data and seek to do so across all countries there's a quite interesting analogy here in what's happened for example on a different area of disclosure which is on climate change disclosure there we went down the route of harmonizing the nature of that disclosure and those really worked wonders on putting this on the radar of a great many companies right across the planet if it's good enough for climate change why is it not good enough for issues of diversity as well and what's more perhaps central banks could play some of a leading role in doing just that in catalyzing action across the piece globally last point many of you in the audience will raise points about are you not overloading these measures of pay gaps they are necessarily imperfect in their construction that is all true they are deeply imperfect but my sense is that in this area something very very much beats nothing it serves as a prompt for explanation it serves as a prompt for accountability and it serves as a prompt for action and in this crucially important area we need lots more of each of those things and disclosure in this case of pay gaps will be one route to achieving that on that Philip I think I will start so Andy got through a lot there the underlying 35 page papers on the Bank of England website Andy is famous for being one of the best writers in economics someone was earlier on criticizing just the economists for not being very clear in their language on Twitter but in Andy's papers and speeches he's usually pretty clear I would say though usually Andy comes up with a pretty interesting title for his papers not so much in this occasion no and so let's see for the next version of it I don't want to take up too much time because we have a closing session but I do want to have a minute or two if there are comments or questions from the floor so please please here in the middle so this is a really well researched topic in the UK I think in the last two years just that I can think of off the top of my head that low pay commission have come out Tim Butcher has come out with a study using the words, the workplace data set and Ash Wendy Olson's got one for the Government Equality Agency looking using the British Household Panel Survey because it can do really accurate job of tracking actual work experience Melanie Jones has got one for the Labour Power Economics using the Labour Force Survey and splitting the gap between men and women Edinburgh have just come out one with Ash using the workplace identifier all of those agenda pay gaps studies all of them control for ethnicity all of them also control for self-selection because of course we're not observing the wages of women who aren't actually working in the labour market all of them include AHARCA decompositions and all of them can include statistical significance tests for the components of the decomposition I think this is actually a really important issue when you start moving into ethnicity as well because there's a very strong lower probability for women from Bangladesh and Pakistan backgrounds to be working in the labour market so if you include gender in the ethnic pay gap analysis then you're getting an even stronger distortion effect on the decomposition so what I'm wondering Andy is why are you not talking about those results as well and why are you not placing your results in the context of these other studies that are out there and telling us about where the contributions are lying and what's different in the findings okay thank you what I'm going to do is collect several comments and then just ask Andy to respond collectively so please a very quick question I remain curious about the pay gap at the Bank of England which you didn't show us the information okay any other comments or questions maybe I'll ask a connected question which is clearly given the public very important differences between public sector and private sector and in terms of policy making does this lead to particular questions for the Bank of England as part of the wider public sector so I'll ask for the Bank of England perspective on that okay over to you Andy thank you so on Karen's question so I didn't go into it in detail I hope the paper references if not all then certainly some of the studies that you mentioned haven't overlooked them I mean as you say some of those are using different sorts of surveys we use the Libreforce survey because it gave us a richer set of in particular individual characteristics over time and a time series in a way that's less true of Ash and the British House held panel survey but I see those results as very much being kind of complimentary rather than a substitute for the other stuff that you mentioned Karen if I had more time I'd have gone into what others have said about this but hopefully it's there in the 40 page paper that you've all had about half an hour to read while sitting listening to me over the last few minutes on the question about our gaps which I conspicuously skipped over that slide a bit too quickly so the latest numbers were on the gender side 23% and on the ethnic minority side just shy of 7% so those are the raw numbers they can be fully accounted for as I mentioned by compositional effects too few too little senior representation of women and ethnic minorities in our workplace we have made we have targets for senior representation among both women and ethnic minorities at the bank and a timeline for getting from here to there and that's our plan for closing the gap from the current levels that it is it is at to Philips point I mean across most parts of the UK Civil Service they're doing something similar already I'm absolutely of the view I know others from the bank many from the bank here including Joe who leads for the bank on this stuff but I can say the reason we went out early a year earlier than any other company was precisely because we see ourselves as having a key signalling and catalytic role to play we can serve as an exemplar for others not just within the public sector but within the private sector as well I mean putting yourself first also meant we were the kind of lightning rod for quite a lot of criticism early on price well worth paying for putting your best foot forward on this issue for disclosing and explaining and having a plan to back up how you'll rid yourself of this over time and that's the track we've taken just before I finish this session let me advertise some other work which is kind of connected so when I was at the central bank of Ireland as a supervisor had a lot of interesting access to fitness and probity applications so fitness and probity is a gateway to senior roles in the financial sector and the differences between male and female applicants or nominees for senior positions was extremely stark so if you haven't seen that work I would recommend that's another angle because again you point out this big difference in management so going to senior management what explains why when it comes to C-Suite and related roles these gender differences are so stark very stark it's quite interesting so with that let me thank Andy and we can move on to the final session so thank you