2nd Session, Meeting of the Group of Experts on Consumer Price Indices

Okay, welcome back, everyone. I hope you enjoyed the poster session. A big thank you to everyone involved in presenting. I thought it went really well, lots of interesting discussion. So my job here is easy, because I'm going to hand over to Jill, who's going to chair this session on household cost indices. It gives me great pleasure to introduce Jill, because she's been a driving force behind the development in the UK. And without further ado, I'll give the floor to Jill.

OK, is that better? I'll start again. Right. Good afternoon, everybody, and thank you, Chris, for the introduction. So this is on household costs indices, which, as Chris said, is a bit of an interest of mine. And we've got four lovely papers to listen to. We've already heard a number of references this morning to the importance of looking at how households experience inflation and the fact that the official macroeconomic consumer price indices don't always do the job that accurately. So now we're going to look at how different countries have looked and are developing household cost indices. And we're going to start with two presentations, one from the Bureau of Labor Statistics in the United States, and that will be given by Rob Cage, whom you already heard from this morning, so he's been worked hard today, and one from Sebastian Privet from the Federal Statistical Office of Switzerland. Now, for this particular one, with the agreement of both speakers, we're going to do things a little bit differently. Because they are both looking at the concepts of household cost indices, we have agreed that we will take the two presentations one after the other and then have a joint discussion session after that. I will now give the floor to Rob Cage for the second time today. Rob, the floor is yours.

All right. Thank you so much.

Good to be back.

The name of this presentation is Georges Ullomp, a conceptual framework for the dual purpose CPI. If I got the French pronunciation there wrong, please see Corinne Becker of Statistics Switzerland.

She was my tutor on how to pronounce that. Okay, so this is the agenda for this presentation, which aims to highlight the work that we've been doing in the U.S.

For the past seven years to improve.

Our population subgroup indexes and measure inflation heterogeneity across different household groups, inspired largely by the.

HCI work that is going on in the U.K.

But this is also a call to action. I believe a consensus has emerged on the major conceptual tenets for producing what I'll call indicator CPIs versus escalator CPIs. And the valuable research that continues in this subfield of inflation measurement, I think is really nice refinement around the edges, kind of a pursuit of a panacea, one solution that all NSOs are perfectly satisfied with. But I don't think there is a single remedy out there that will satisfy us all. But I think we know enough at this juncture to begin production of dual purpose CPIs. And we owe it to our stakeholders to start producing indexes specifically tailored for escalation of retirement benefits, which is the primary or a major escalation use of most CPIs. So as I present this material, I'd like you to ask yourself, This question, are we measuring inflation or are we making it? Are we actually contributing to its cause by producing a product ill-fitted for its intended use? Okay, so with that, let's launch in. I wanted to come up with a metaphor to start my presentation. We're in Geneva and sort of Switzerland in general. which is great bicycling, so why not pick a bicycle metaphor to start. So if you take your bike to the mechanic and you pick it up, the first thing you want to do to make sure it's ready to ride is to check the air pressure. And the tool you use for that is a air gauge, a pressure gauge. But if your tire is flat and you need to inflate it up to the proper pressure, you need a different tool, a pump. And with technological innovation, there's now these gadgets out there that are really spiffy. They're two tools in one. You know, they're a gauge and a pump with a lot of power. I mean, these little gadgets can inflate the--not only two bicycle wheels, but four wheels on a car. So they're really sort of fantastic devices. But they're still two distinct functions.

And I believe the CPI is basically the same.

Two distinct functions, one as a gauge, a principal federal economic indicator, and then one as a pump, as an escalator, and mostly for retirement benefits.

So I'm going to not keep you in suspense.

I'm going to cut to the conclusion right up front and give you the conceptual framework that I'm proposing for producing two lines, two separate CPIs for the principal economic indicator, You have the four tenets here. The population target should be the broadest measure that you can achieve within your country. For aggregation across households, you need a plutocratic approach. For aggregation across items, we need a superlative index approach. And for treatment of owner-occupied housing, a rental equivalence or other flow of service approach is suitable. Now, there are other aspects, other tenets, but I'm going to be focusing on these four. The distinguished speakers in this session that follow me we'll talk about some of these other issues. Now for your escalator CPI, I think it's critically important. And here I'm going to be focusing on retirement income escalation. There are others, for example, low income measurement for food stamp eligibility or food assistance in the U.S. is another important use. I'll be focusing on pensions for retirement beneficiaries. So you've got to get the population target correct. which I think this issue is often overlooked. So you and for your escalator, in our case, you want to be able to find the inflation experience of the retirement income recipients or the beneficiaries. Aggregation across households should be democratic. Aggregation across items, again, the superlative index approach. And here's the big one. Treatment of owner occupied housing needs to be an actual out of pocket expense or payment, not the rental equivalent or flow of service approach. All right. What's my motivation?

I'm going to walk through each of these four.

Tenets one by one and cite some of the prior research and thoughts on these topics that have motivated me. So here is the actual public policy that's stated by our Social Security Administration on the use of the CPI in the first instance to escalate retirement benefits in the US. Social Security is by far the largest retirement pension plan in the United States. Over 95% of the labor force participates and is eligible for this plan. And if you can see, if you read closely, you know, the policy behind the use of the CPI here to escalate their fixed annuity is to protect beneficiaries. So not all households, the beneficiaries, it's protecting their purchasing power, right? And it's helping them to maintain standard of living, very critical. That tells me that your index needs to be somehow tethered to a cost of living index approach, not.

Just a cost of goods approach.

So here, I mean, our stakeholders are.

Actually telling us how to produce this index if you read this very carefully.

Maybe I'm interpreting the words very, very strongly, but to me, this tells us what to do. So for the cohort definition, I'm doing something new that we have not done previously, which is to define the eligible set of households for this index to be those households that actually receive the pension actually receive Social Security income. And in the US, it's 35% as of 2023. That's a very large, large number. So you have sufficient sample size for this cohort in our household survey. So for a brief history, the existing index that we use for this purpose is actually based on the spending behavior of households in the labor force. And this is sort of a quirk of how the legislation was first passed in 1972 to tie indexation to the CPI. At that time, the BLS was only producing a CPI for this narrow cohort, wage earners and clerical workers.

And Congress has not acted to pass that legislation.

So that's the index that's currently used, our sort of old legacy product. Over the years, different indexes have been proposed. We have a research CPI-E, which is measuring the inflation experience of older consumers, age 62 and over. Congress has mandated that we produce this as a potential replacement for the CPI-W.

I think it's a good one.

But it's not quite precise. It doesn't necessarily zone in on households receiving Social Security incomes. You can be age 62 over and still be in the labor force, for example. So on the margins, it's not perfect. We also have the chained CPI-U, which we introduced in 2002. This index uses superlative aggregation across items to control for substitution bias. However, it's only available for all urban consumers.

We haven't yet to produce it for a cohort.

And here I am introducing for the first time, drum roll, the RHCI-SS, which is in development.

So it's.

Index specifically tailored to our retirement recipients, and we're defining that as any household that receives any Social Security income. So here's a histogram of what those households look like. I have a sample from our spending survey from 2016 to 2018, so three years of data, and I'm keeping in the sample households that participate for a full year. So I have full year spending on each household in the sample, the sample size is just shy of 10,000. So what this histogram is showing is the percentage of Social Security or retirement income as a percentage of all household income. So it's relatively stable and flat except on the far right. So these are folks.

That have 95% or more of their total income sourced from retirement.

It's 30% of this cohort. This is really a stunning finding to me. And, you know, one might argue this is the group that we're trying to target for the public policy.

Because these are the folks that are really reliant on the Social Security income to maintain their standard of living.

But if you threw out everybody else, now you have small sample size problems, and so why do that? So I include all the households as long as they receive some income, and I've developed this adjustment factor, which you see there on the right, RJK, which is the ratio of retirement income divided by total, and I'm going to use this as an adjustment in my democratic aggregation across households, which I'll show at the end. And I'm going to have to speed up here because I'm already 10 minutes into the presentation. So here's tenet two, and this is the plutocratic gap. We're all sort of familiar with this issue, which is, you know, if you aggregate households inflation experience, you can aggregate it in two basic ways by plutocratically weighting households by their total spending level or income or democratically where each household gets a weight of one. Angus Deaton in the late I like this quote from him, which he found in the US that the plutocratic CPI represents roughly the representative consumers around the 75th percentile of income. And then you have this quote from the At What Price panel in 2002, which basically claimed that it's hard to imagine that policymakers would want to make the decisions about pensions by tracking households at that level of income. So to me, these two quotes sort of proclaim loud and clear that the democratic approach to aggregation across households is the right one for Escalator. Okay, tenet three, this is an issue that I believe is settled in our line of work, which is that we prefer, if you have the input data, to calculate a superlative measure across items. So in the U.S., we do have that ability, and so this graph is showing the difference between Our superlative chain CPI, year-over-year change minus our headline CPI product, every bar below the red line there means that the superlative number was lower. It's expected to be lower because it accounts for substitution bias. There's a couple periods where it was higher, so it's not systematically lower. So that was the first period there on the left is the housing crisis, the mortgage crisis of 2008, and of course the pandemic, the chain CPI slipped a little bit higher for a couple of On average, this is about two-tenths of a percent difference. All right, I'm going to skip past these fun little cartoons that sort of raise tenant four, which is rent. And this is the big issue, right? A lot of people in the US right now think there's nothing going on but the rent. That's a rap song, or R&B song, sorry, from the late '80s, from Gwen Guthrie, if you're familiar with it.

And they think that the reason why.

Inflation has not come down, backed out to sort of the Fed's target of 2% is because rent prices and shelter inflation have, are continuing to persist high. And, you know, that is true. Of course, eggs is now really the, the big problem in the U.S. So maybe I should be talking about eggs, but, you know, it has such a small weight. I don't think you need a special treatment for it in your escalator CPI. So I'm going to skip over that and just real quickly show you some data. In the U.S., our all items less shelter measure is hovering around 2% right now. So that's good. That's our Fed target. But shelter remains high. And this is historically the case. If you see on the left side of the slide that shelter indexes run around 3% per annum. This was pre-COVID, with everything else running around 1%. So rent matters. It has a big weight in the CPI. It's 36%. Of all the shelter, we define shelter as four things, tenants and household insurance, lodging away from home, rent, and owner's equivalent rent. So rent is easy to measure. We have a household sample that we track, actual contract rents that are paid, and so that's easy. We use rental equivalents for our headline CPI as the principal federal economic indicator. One of the most difficult issues for CPI compilers to tackle, and if you don't believe me, the CPI manual professed this in 2004. I think our updated version also claims this to be a difficult issue. So the treatment options tend to be that you can exclude it from the index altogether. You can track home sales and home purchase spending outright, like acquisition cost approach sort of, or you can track the flow of service value, which is what we do. And we do it from, so we're basically taking the rented prices and assigning it to homeowners. And so this really causes a problem. And as stated by Michael Boskin of the Boskin Commission in 1996, if the purpose of having a separate price index for the elderly is to use it to adjust Social Security benefits, it should not be -- that cohort should not be compensated for capital gains on their housing. So as rent prices go up.

It makes sense that housing prices go up, and those that own their home are getting equity in their -- in their -- by owning a home. And so by -- you're sort of.

Rewarding them for that asset valuation, which you don't want to do in a escalation index. And so this is dramatically shown in this slide here, which shows the home ownership rate by our different cohorts. Just take note of the first line, Social Security recipients, over half -- half of these folks in the U.S.

Own their home without.

A mortgage. They have no payment.

And yet we're imputing this large rental equivalent payment for their -- in the headline CPI.

So this is the problem. It's not just a problem in the U.S. Here's a map of the home ownership rate by country. I was particularly surprised to see how high it is in Asia. So I'll be talking to my Russian Federation friends after this to explain that. So here's a comparison of our -- the owner's equivalent rent approach versus an out-of-pocket payment approach. This is for renters. So The owner equivalent approach and the payment approach for renters are going to be the same, so they line up along the 45-degree line here. Here are owners who own with a mortgage, and I'm at 15 minutes. I'm going to go about one minute over here. They have slightly lower weight. Here, the payments approach is I'm counting it as principal and interest and taxes paid. And here are the folks that own without a mortgage. You can see that they're clustered near the zero line because on average, these folks, if we use the payment approach, would only have a 3% weight for shelter. So what we're proposing to do is a payment method. And our proposed definition is mortgage principal plus mortgage interest plus adjusted property taxes. We want to remove any part of the tax payment that is servicing the equity growth in the asset, because that is investment. We don't want to classify that as consumption. Our loan debate at BLS is over mortgage principal. Should that be included in your payment approach, or should it be excluded?

I'm of the opinion to include it.

I think I'm in the minority view. Happy to have debates about that after the session is over and at dinner and at happy hour and whatnot. So here are some of the prices that we used in the prototype index over, these are average price relatives over the 2019 to 2024 period. And real quickly, I'm going to just show the results of our initial prototype calculation here. And these are independently showing each of the tenants impacts on headline measurement, aggregation across households. If it's democratic, we get a slightly higher result. Cohort definition matters, slightly higher result than the all urban index. Aggregation across items, a super approach, slightly lower. And of course, treatment, and this is my message, treatment of owner-occupied housing is the most critical piece if you switch to a payments approach. it could be up to one-tenth of a percent per annum lower. So the summary results are there on the right-hand side. Our official CPI-W over this five-year period was 4.3%. My prototype HCI for Social Security recipients coming in at 3.85 percent. That's almost half of a tenth percent lower. This is big business, and that equates to $5 billion in transfer payments in the U.S.

For this system.

So, I think getting it wrong, and that's why my cry is for us to start producing this index so that the Social Security Administration can use it. So, I'm over time. I'm just going to wrap up with this slide here, which talks about some of the practical considerations that we need to consider in actually putting this proposal into practice. If you have a household spending survey, here are the data elements that you need to collect. On the price side, you're going to have to start tracking mortgage interest rates. There is a way to control for elementary index heterogeneity. Don't have time to go into this chart on the right. So come see me after the after the session. And I think there's a way that we can do a superlative measure which is listed here. So I'm out of time, I've gone three minutes over, so I'm gonna see the, yeah. Also, we do have, a lot of our research is out there on the internet already, so some of the links are provided here. Our lead researcher is Bob Martin. He's done some nice work, which you can check out online. So, thank you.

Thank you very much, Rob. That was fascinating. Right, let's go on now to Sebastian. Okay, from Switzerland. Yes.

Thank you very much. So just look my clock. Okay, perfect. Thank you. So good afternoon, everyone. So I will directly start with something quite original. I wanted to start with a quote. Here in the picture, you see Joseph Stiglitz, who was a former winner of the Swedish Bank Prize in Economic Sciences. And he makes an important point regarding macroeconomic statistics. He says, as important as metrics are, our metrics are beset by problem. The choice of metrics is part of the debate. In what context he say that? Well, he say that in context, of course, of GDP and the problem of well-being. And he wanted to highlight that, as you probably know, that GDP is not the best measure to capture the well-being, and there are also other measures that exist. And you may be now wondering why I'm starting with this question about the GDP and well-being. is that because we have seen a little bit, if we look back in the history, something quite similar in the inflation statistics. It started first in 1996 with the Boskin Report, which popularized the question of the cost of living index. And then since the 2000, we have seen a new manner to measure inflation from our household perspective that started first in Australia, then more recently in New Zealand, and in UK with household cost indices. So I'm Sebastien Privet. I'm currently working at the the FSO. And I will begin with some conceptual approach about HCIs and then I will show some example in the context of--in the Swiss case. So let's directly start. And to me, it seems that it's really important to start with Some, the major foundation of the orthodox perspective of inflation, to me, there are three main foundations. So the first one is the use of the CPI. By that, I wanted to highlight that since maybe 60 years, we have seen more and more use of the CPI as a tool for targeting inflation, which means that it's more and more used for a macroeconomic purpose instead of a social policy tool. And this is really important to understand because it raises an important question about the validity to, to compensate also with the CPI, another important, um, foundation, it's a conceptual issue that is not really new also, is the question of the debate, uh, between average inflation and inflation by, by category of, of also the population. Of course, we already know that inflation does not affect everyone by the same strength. And so, yes, this conceptual issue, I like it also an important question about the validity of the CPI to compensate all sorts with some specific characteristics. And then finally, something that is quite also original and interesting is the practical need of this type of indices because they emerge after the discontinuation of the retail price index in the UK. So it was the initial social policy tool in this country. Beside this foundation, there is some differences, of course, with the CPI and the cost of living index. I just don't have the time to go deeply in all the details. It's more detailed in the paper. Here in the column of the CPI, it's more the Swiss CPI. So of course, in some of your countries, it could be some quite different. But anyway, I wanted to highlight three main differences that are involved here in the text. The first one is the aggregation method, as it was explained by Rob. So I don't want to go more deeply into the details. But yes, of course, in the HDI, there is a democratic aggregation. Whereas in the CPI, in the Swiss CPI, we use a bureaucratic aggregation, then there is also an important differences, and we will see it's really important in Switzerland, it's the difference in the insurance premiums, there is a growth rate method in the HICPs, whereas in the CPI, we use a net weight. method which basically we just take into account of the service portion of this insurance premium. And finally, something that is quite also important to understand the differences with the cost of living index is that basically they have pretty the same use. So the idea is to use it as a social policy tool, but we have this problem of the utility. So of course, the cost of living index is more challenging, whereas HCIs, as we have seen in some countries, it's possible to implement. So, beside these, uh, original, um, um, differences, there is, of course, uh, there are, of course, some challenges, the first one is about taxes, so in, in the literature, they highlighted that, um, as in the CPI, uh, they should include it, uh, regular and recurring taxes. and excluded income taxes, but there is some open question when it deducted at sources in some countries because it's already inside the household budget. So it's something that we should a little bit more think about. Another important challenges is the quality adjustment because some authors highlighted that in general, the quality, the upgrades about the quality, it's more a false quality. So also do not choose about this quality adjustment and that's so that in general, it does not reflect the actual payment made Because I don't know if we have a price increase of a computer of 4%, then we measure that we have a 5% of upgrading quality, then we will report only minus 1% of inside the index, but it's not really what I also pay, so there is also some… debate about discretion. And finally, there is the price heterogeneity problem that was already a little bit explained this morning. So I will not really go more deeply into details, but there is some discussion about discretion when also with different income levels, varying prices for the same product, and they substitute them also in different ways. So for some others, it could also underestimate sometimes inflation for these categories of also. Now I wanted to move to the Swiss analysis, I think that Switzerland is really interesting because we have seen at the beginning of 2000, also consumer indices by population groups, and also because there is a main critique of the Swiss CPI is the exclusion of the health insurance premiums. And so alongside with the Swiss CPI, we published since 1999, an health insurance premium index. Why we do that, it's because it's something quite important for us all, because in Switzerland, everyone must have a health insurance, and the premium is the amount you pay every month, even if you don't use any medical services, and basically it's offered the same coverage. And so there is some debate about these topics, and we have seen with this health insurance premium index that, Until 2024, it have reduced the growth of disposable income of households in Switzerland. So there is some impact of this essential spending index. So based on this problematic, I constructed analysis based on the household budget survey for the weights and the prices comes from the Federal Statistical Office. So I do not adapt to the prices for category of households. It's basically the same prices for every type of households. And basically, I will not show really a full HCI's approach, but more I will concentrate on two main concepts. The first one, so it's related to this insurance premium index with the growth rate method of the HCI's. And then I will also show indices for 10 also types plus the democratic aggregation. So the first result is is in this slide, it's about the insurance index in Switzerland, um, between December 2016 to November 2024, and basically, we, we, we see that we have a big differences, and I will quickly explain it, but, um, the decrease of 3.7, 3.7… It's basically what is taking into account in the Swiss CPI, so it's the decreasing of the prices of the services, so it's another approach, so here the quantities are constant, whereas in the HCIs, if we included this health insurance premium index with a growth rate method, we have an increase of 18.5%, and basically it shows something that that is not really new. Also in the statistics, here we have a quantity effect. So basically, the more people go to the doctor, get prescription and so on, the more premiums tend to increase. And so this approach highlighted this fact. But of course, there are pros and cons in the two approach. I'm not saying that is the best method, but basically it showed that, yes, of course, the concept have an impact. And we see that for a specific index aggregate, it should have a big impact over a small period. Now, if we move to the indices, for some households, you see that I have some income characteristic, I have also ages, and I have also some income quantities. And basically, here I show the 2025 weight for the ECO-COP group 4 for housing and energy expenditures. And basically, we see clearly that we have some differences. Of course, if you are a low-income household, you will spend or allocates more of your budget to these categories of expenditure, whereas if you have more money, you will spend less of your budget to this expenditure. But basically, what is more important to me is to highlight the difference between the democratic CPI and the plutocratic CPI. Just to remember, the plutocratic CPI is basically the Swiss CPI. So in the democratic CPI, it's this approach in the HCIs. And we see that For this specific group, we will have a 5.6% of increase for these categories. Of course, it's something quite extreme. If we look for all eco group for 2025, we see that it's not always the case for basically for essential needs, such as for health or food and non-alcoholic beverage, the democratic CPI's weights exceeds the Plutocratic CPI weight, whereas for the group seven, transport or 11, recreational and culture is the opposite, so it's something also not really new. And of course, if I don't add up the prices and I have just a different weight, I have also different, so various indices. And here I wanted to show just two examples. The first one is the difference between a low income also and a high income also in Switzerland. And it's something that is not really intuitive. So we've seen, for example, in the period that is noted too, it's during the pandemic in Switzerland. basically the higher income households face higher inflation than the low income households, so it's not really intuitive. But anyway, in the most recent period, for example, since mid-2024, it's the opposite, something that we could expect. And also during between 2016 to 2018, we have seen also the same. Something that is... I think quite anecdotical, but always interesting. It's a difference between a couple without children and single without children in Switzerland. And basically, we see that, yes, single also without children face higher inflation through the entire period than couple without children. Now, if we look across different household types through the entire period, we see that we have sometimes some differences, of course. Sometimes they are relatively moderate. For example, the difference between the democratic CPI and the autocratic CPI is just 0.1 point of percentage of differences. But for example, for elderly people, it's more important. We have a decrease of 0.7 point of percentage. And for single people, single households, sorry, we have an increase of 0.9 point of percentage. And just to conclude and to sum up, I would like to stress that, uh, alongside this, uh, this presentation, HCIs matters because, uh, first, they address the need for a compensatory measure of inflation, and I think that we understand with the beginning of this, this presentation that sometimes the CPI… tend to overlook this aspect of the compensation and the use of the CPI as inflation, as a social policy tool. They also demonstrate that also do not face the same inflation. We have seen some example in the UK, in New Zealand, in Australia, and now we have some example also in Switzerland, even if it still remains sometimes moderate. And also something that is quite original is that they prioritize the real life experiments of also, for example, if we think about mortgage interest, sometimes we say, okay, no, it's not consumption, but actually something that a lot of households pay every month. And with this prioritization of the real life experience of also, we see that there are some impacts on the results that we calculate. And for all these reasons, to me, it seems that they enhance the legitimacy and public understanding of the National Statistical Institute. So that's it for me. Thank you very much. I hope that it was useful. Thank you.

Remember this. Thank you very much, Sebastian. That was fascinating. I think what's fascinating also is to see where the two of your presentations were similar and where they differ, because countries differ. but there are some things that are the same. Anyway, oh, right, so I can see there are some. Right, well, let's go Italy, Mexico, can't see the one at the back, South Africa and United Kingdom. So, what did I say? Italy, Mex, SA, UK. So, Italy, the floor is yours.

Thanks for the presentation. Actually, I have a question regarding the treatment of the index proposed by Robert, especially in terms of the, you know, out-of-pocket payments for rent, because, you know, you show that the large majority of the recipient for social security is actually homeowner. But my question was about the 23% and how much of this 23%, so what is the share of reliance on social security payment for this 23% which is not, as I, home owner? Because on the one side, I understand your point of averaging, or this is smoothing the payment, but in the end, if you come out that this 23% is the one that is most reliant on the social security income, you're basically, getting them a rebate on their, which while they face the full increase of the cost. So, because in the end, you're trying to protect the recipients as is their mandate, but the most vulnerable should be the one deserving a little bit more protection, even though it means that, you know, the one more well off get, let's call it a payment for the appreciation of their asset. So.

Okay, excellent question, Luigi. I think you're right that, you know, by, you know, we're not really protecting those that heavily relied on the Social Security income by including owners in the measure. If this, if that cohort that's the 30% that have not, that rely heavily on Social Security are mostly renters. I don't have that breakdown with me, but I'm assuming that that cohort are more, that's where the renters are, most of them. Or not, they could be the homeowners that, you know, bought a long, long time ago. So I need to get that statistic and I can decompose it for you. But that's why I think it's important to calculate that adjustment factor.

I'm skewing.

The democratic aggregation toward that cohort by I didn't explain it in the in the presentation but I'm applying each household doesn't have a way to one in my proposal it has a weight of the proportion that they rely on their Social Security so I believe that adjustment will sort of solve the issue that you're raising and if not we'll have a debate after yes Mexico.

Next, please.

Thank you very much. This is for Sebastian. In 2012, my colleague from Germany presented the German CPI. And one of the questions was, what is one of your great challenges? And she said that some groups of people do not believe our data. CPI, no, because I have different inflation. So now with this presentation, my friend, are you suggesting that inflation concept of definition has to be changed?

Thank you very much for your question. I think, yes, it's a quite interesting question. But no, it's not really that. I mean, it's just what I wanted to show. So It's, it's just, I, I'm based my analysis on the UK example, basically, and so what I've, what I found really interesting is, is that it's really related to the purpose of the index, so the idea is just, for example, in the UK, is to have a macroeconomic index, so, for example, for the purpose of targeting inflation, we have one specific indicator, and if we wanted to compensate also to index wages and so on, we have another index, so basically, it's not saying all is false, it's just saying that for one specific purpose we could think about have a new type of measure or new concept to measure this inflation. I don't know if it's answered your question, but I have some.

Yes, very good, thank you.

Thank you.

Okay, thank you. Right, the next is South Africa.

Thanks very much. I found this part quite challenging presentations, and I think you're bringing us something outside of the field, perhaps, that we're accustomed to, which challenge is probably good. So this is not specifically to either of you could respond. I mean, how do we see what you're talking about, this household cost index is different to a cost of living index, which I think is well considered within the kind of consumer price index theory literature. You know, typically we, you know, we may talk about a macroeconomic indicator, we may talk about a cost of living index, and there's a lot of CPIs that fall somewhere between them or satisfy both requirements. What worries me a bit from the presentations is that you're kind of proposing an index that's not really consistent with typical CPI methodologies. I think it's absolutely important to provide CPIs for different population subgroups, different income groups, pensioners, retirees, whatever. I think it really helps us gain insight to the experience that people have. But I think that what provides a credibility through all of those is that we're consistent in our methodology. of the different types of indices that we produce. So it does concern me that if we start to say we want to produce something that's totally-- well, I won't say totally, but has some significant elements that are kind of outside of the framework, yeah, I'm not sure where that leads to. And I think, specifically, the inclusion of mortgage interest, for me, is quite a difficult one, you know, as an item by itself, just because when inflation starts to increase, central banks will start to increase interest rates and we will then get higher inflation rates through our household cost index. And that seems to me to be a circularity that might not be that positive. Thanks.

Okay. We've got several people who want to ask questions, so, and we've got a hard deadline of the coffee thing, well we can go over a bit, but what I'm going to suggest then is that we now take several questions together and then I'll ask both Rob and Sebastian to pick out which ones, what they want to comment on. So I'm going to ask, it's UK and then it's Australia, I think there was somebody else down there but we'll take UK and Australia and Sorry? Oh, the ILO, yes, okay. So we'll take UK. India. And India, right, okay. So four questions, if you could keep them as obviously as brief as possible. So we've got UK, Australia, India, and the ILO. UK. Yes.

Thank you. I actually have two questions, so I'll try and be really, really brief. So it's really interesting, mostly to Rob, in your analysis of the retired households, targeting that towards the retired households and the interaction with low income households. So they obviously have quite different spending patterns, especially with owner-occupied housing costs. So, do you think when we talk about other benefit measures, so focusing on low income households, should retired households be excluded from low income households to kind of remove their unusual spending patterns from anything that we then want to use the low income index for, like escalation purposes? And then secondly, sorry, the retired households again have an interaction with the low income households in what we've been discussing earlier with the cheaplation, like having, so in those senses they're kind of losing out with the higher price rises for those cheaper goods. So how do you see that in comparison to the loss that they're getting from not accessing, yeah, there we go, thank you.

Okay, let's go on to Australia.

Thank you. Yeah, well, it's one of the countries that has living cost indexes as well as CPI. I guess what I would say is that they're complementary. They're not one replacing the other. So in Australia, the CPI is used by the Reserve Bank for interest, you know, for making interest rate decisions. So as you say, you can't really have interest rates in it, but they impact people's payments. So Rob, I think we have mortgage interest charges in our living cost indexes, but we don't have payments. But we do raise it as mortgage interest payments get higher. So as people are taking out bigger and bigger mortgages, that feeds into how we calculate mortgage interest charges. So just maybe that's a a way that you might sort of have a compromise there. I would say they're really useful. We otherwise spend an enormous amount of time explaining to people why we don't have mortgage interest charges in the CPI. So having the set is, I find, really, really helpful to help people understand what we do. One last thought for you is around, you said that the ones that the aged Social Security one is running a bit lower than the CPI. So obviously the Social Security recipients might not be too happy about that change. So in Australia, it's interesting because we, the people on government benefits get the higher of the CPI increase or the increase from the living cost indexes, whichever one is higher. So that's the way that it kind of works together. So another perhaps tip for you there or sort of a question whether you've considered that. There was actually a long discussion too around whether wage price index would be better because going to your point initially about whether or not it's maintaining a living sort of maintaining a standard of living, there was a lot of discussion around, you know, wages often go up higher than inflation, so doesn't that mean that people who are retired are missing out in some way. So I think ours is actually a little bit of a combination of average weekly earnings underpinning as well as the other things. So yeah, not really a question, but some thoughts to share. Thank you.

Thank you very much. All right, we now go on to India.

Yeah, thank you. Congratulations to both the presenters for very thoughtful presentation. I have just two, three points to raise here. The first is that there are, I mean, what are you suggesting that should we go for so many segmented CPIs, like CPI for industrial workers, CPI for rural, CPI for urban, CPI for rural laborers, CPI for social security holders, or we should go for a unique, consolidated kind of CPI which should serve to all the segment of the population. So I think this is one question which is very important and which is, I mean, we are facing this challenge and we are not able to respond to this question. And the second is that when we say that inflation is a tool for targeting inflation or a social policy tool, So for most of the countries, NSOs are working for both these, are compiling CPIs for both the purposes. Like definitely central banks are using the CPIs for targeting inflation, but for the government, it is a tool for, it is a macroeconomic indicator for determining the social policy. So in view of that, this discussion on CPI versus the cost of living index is always on and there is always a confusion that should we not consider compiling this cost of living index also or should we not make this CPI such a robust index that it should be able to address the objectives of, I mean, both these objectives, like it should also address the targeting inflation and can also be used as a social policy tools. So whether there is, should we not work on, think on that direction that how we can club these two things instead of developing different kind of segmented CPIs. Thank you.

You. Thank you very much. And finally, the ILO.

Presenters for very interesting presentations. So a couple of comments on both presentations and maybe some questions. First about the objectives of the CPI. Initially, it was designed for adjustment of wages and social security payments. Use as a macroeconomic indicators came much later. So in most of the countries, recently we have done, I mean, a couple of years ago, a survey among all countries. And all of them, they have indicated that the CPI is used for adjusting wages and social security payments. And, of course, widely used as a macroeconomic indicator for measuring inflation. So it's not at the expense of indexation, CPI is an indexation index. The second comment is about the use of democratic and plutocratic weights. In the US case, BLS simulation, use of democratic weights would lead to cost saving for the government. But in the case of Switzerland, it's different. So what my point is that the decisions regarding any alternative index should not be driven by this cost saving, the purpose of saving and making saving for the government budget. The third comment is regarding the, in the US case, 35% of all beneficiaries are receiving, no, 35% of all the whole population households are beneficiaries of social security payments. But CPI is widely used as a indexation for indexation of wages. So we should not forget that. So either we have one for the social security beneficiaries, one for the wages, but most of the countries cannot afford to have too many indices. And then it's also confusing. One should have in mind that having too many indices is confusing for the users. and they will go for the one that is most appropriate is in their benefits. I mean, the lowest one, especially if it's an employer or the government or somebody else. And then my final comment is regarding the ages. You said in US, 35% are living for free. They have free accommodation because they have repaid. But that is very interesting facts of mortgage free. But this, that is something that one should be looking at. Maybe the aging of societies, we are getting older and older. Initially CPI was designed specifically for adjustment of wages, not even for social security payments. But at that time, I mean, 100 years ago, 90% of the population was employed or working. Now this ratio retired to working is deteriorating in most of the countries as the societies are aging. So maybe that is something that we should be taking into account to what extent it affects the other methodologies. Thank you.

Thank you very much. There were some very, very perceptive questions there. I don't know how long we can impinge in the coffee break, Chris, but it seems to be no problem. Well, no, hang on, it's not no problem. That's wrong. Coffee is important. But if Rob would like to speak first and then Sebastian, with any picking up what you would like from those questions, and then the debate can carry on afterwards.

Okay, so very quickly, those were some fantastic comments. I have some notes here. I hope I can do them justice with my responses.

So first.

Patrick, on this issue of whether the escalator CPI should be a COGE or a COLE, I think it needs to be a COLE. I think you have to account for quality change in this measure. Otherwise, you'd be overcompensating households for quality change, and so that would be an upward bias. Your comment about interest rates, yes, that was true in the early '80s. The US actually used a mortgage interest payments approach for our housing measure. And of course, in the early '80s, interest rates were sky high, 12% to 18%, and it caused these massive increases in the CPI, so which one of the reasons why we switched to the rental equivalence approach. I don't think that's much of an issue now, not just because interest rates have come down, but for this cohort and in the US, Most of the mortgages are fixed rate. 95% in fact are fixed rate mortgages. ARMs or adjustable rate mortgages are a very small percentage. So any increase in mortgages, most households aren't really facing those.

They're on fixed, fixed. Now that might be different in other countries and so your point is well taken. Comments from Tanya on the UK.

So I think you want to treat Your cohort index design for retirement households, completely independent and separate for any treatment that you need for low-income houses, even though they're related, especially those that heavily rely on Social Security income.

That's a small amount.

Those households are likely in the low-income cohort. Now, have I mismeasured this because I've completely not accounted for price heterogeneity of the elementary index? Yes, that's a huge issue.

My proposal, which I didn't have time to talk.

About, is that I think in our current construct, in our current survey operations, we can attempt to control for it by stratifying.

If you have a household.

Survey and you've collected, this is what we do, we collect our retail establishment frame for our pricing survey based on a survey of where people shop. So you can stratify that sample selection based on the characteristics of the households. And so my proposal is a two by two matrix, Social Security recipiency status, yes and no, low income, yes or no.

And then you've got four subgroups based on that two by two matrix.

And you can use overlap maximization sampling to control for price change. heterogeneity to sort of improve the measure. And yes, that would expect it to go up if lower income households are experiencing. So if that did not make sense, we'll talk after. Australia, fantastic comments.

I agree with almost all of them.

I can't really read my notes, but we'll talk.

India, yes, I think what I'm proposing is that NSOs produce two.

Indexes, two tools, a gauge and a pump. You've got to produce an index for the macroeconomic indicator as a barometer.

That's what the central banks want.

But that index is completely ill-suited for escalation. So we have to be producing the second index and using different methodologies underneath them.

So that's really what I'm proposing. The ILO comments were spot on as well, agree with all three of them.

Our CPI was originally developed as a wage contract escalator. So yes, and over time, it's morphed into this macroeconomic product that now undermines that first use of the index altogether.

So this is why I'm saying that we've got to fix this by producing the second product as an official product.

Respect to the different cohorts, agree that we still -- what I'm saying is that we want to agree on the methodology for the escalation purpose and then produce several of them for each cohort that's needed. So we need one for the retirement cohort. We need one for the wage earners, for wage contract escalations and so forth. And I think -- and we need one for low income, because in the U.S., we have this 1950s.

Measure of what the.

Cost of a minimum nutrition is, and that measure has been escalated by the CPI since 1950. 1950 to establish the eligibility for income assist for food program assistance. So to keep that measure more accurate until it's directly measured, we need to measure price change as.

The low income households experience. So I think that would target that and maybe that's.

Yes, thank you. And you have already. well responded to all of the question, but just maybe to be to to sum up a little bit my my my talk about these topics is just that we have seen, for example, we do not really agree even on what is really for example, about quality. As this one, you just say something that it was different in my presentation. So we clearly saw that, yes, there is also a lot of debate about also this topic. So we could construct something It's not open? It's okay. It's okay? Okay, sorry. No, but for example, that we should take the example of the US, or the Swiss example, or in the UK, it depends. But I think what is really important, and for example, India or the ILO say something that it's at the end, the core of the question, it's if we wanted to have a specific index or if we wanted to have something for all the purposes. And I mean, just when I started with the presentation about GDP and well-being, it's really related to that. It's the fact that, yes, of course, sometimes it's more complicated than just one indicator. But I guess if we look at the GDP problem, for example, we could not say that the growth could measure the well-being because it's, of course, a little bit limited and we should also take into account of inequality measurements and so on. So I guess for the inflation, it could be the same now. It's to have not just one indicator, but if we have many indicators, this could be useful and I think more understandable for users and also the public. So I think that it's really something not really new actually in the lot of statistics to have many measurements, so I don't really understand the problem to have maybe one new measurement of inflation. So that will be maybe my last thought. Thank you very much.

Okay. Well, thank you very much to our speakers and to those who gave the questions. And now it's coffee time. Chris, what time do you think we should aim to be back?

So, yeah, so lots of interesting discussion, which is excellent, but can we try and have 15 minutes and be back in 15 minutes to try and start by 20 past at the latest? Is that okay? Thank you. Okay, we're going to make a start back shortly. Can we please start taking our seats again? Okay, welcome back everyone to the final session of the day. Given the amount of interest in the household cost indices before coffee, we've got a bit of room at the end of the day tomorrow because unfortunately Jens Mehlhoff can't make it. So, Rowan has kindly agreed to take his presentation at 5 o'clock tomorrow, which means we're going to cover household cost indices for the rest of today's session, okay? So, any questions that you might not have got around to earlier, we'll cover them after the presentations, if there are additional questions. So, I'll hand back over to Jill, who's going to take us through the rest of the day.

Thank you very much, Chris. I think what we will do though for this half of the session is to go back to the standard way of having a presentation followed by questions because these two presentations are rather different, whereas the two previous ones in a sense were discussing overlapping items. So we will start with Indonesia and Windy Putanto is going to talk to us about the work that they have been doing. Windy.

Okay, good afternoon, everyone. My name is Windy Arso Putranto. I'm from BPS Statistics Indonesia. Today I'm going to present a case study titled Creating Accessible Indicator for Public Understanding of Price Change, a case study of the household cost index in Jakarta, Indonesia. Yeah, the motivation of this study, since there has been growing public concern that the official inflation rates does not reflect everyday living experiences. The CPI, or Consumer Price Index, captures the general picture, but often miss the specific inflation pressure faced by different households. To address this, we developed the Household Cost Index, offering deeper insight into inflation disparities among various socioeconomic groups. For this initial step, we focus specifically on Jakarta, the capital city of Indonesia, considering its important economic role, unique consumption patterns, and significant income disparities. For the methodology, we use data from 2022 Household Expenditure Survey, known as SBH for Jakarta, household are grouped by expenditure quantiles from the lowest spenders, Q1, to highest, Q5. We compile community basket and spending weight for each group using 479 items from the Jakarta CPI list. Finally, we calculate the HGI for each group using 2022 as the weight reference year and monthly price change from 2020 to 2024. This slide shows the expenditure range for each quantile group. It highlights how different household groups consume different types and quantile qualities of goods and services, face different prices, and have varying preferences and demand elasticity. Here we can see the CPI compared to HCEI across expenditure groups. If we make three groups, during the first period, Q5 HCEI was higher than the CPI by 0.91 points, while Q1 was slightly lower. However, in the third period, Q5 HCEI fell below the CPI, showing a shift Meanwhile, the gap between Q1 and Q5 widened over time. This slide illustrates how household groups prioritize spending differently across categories like food, housing, transportation, and more. It shows how consumption patterns shift according to income levels and living standards. For each category, For example, in this slide, lower expenditure groups allocate a larger portion of their budgets to food, beverages, and tobacco products. This is typically pattern, lower income household dedicate a greater share to spending to essentials. For the transportation, transport spending also varies sharply. Q5 household prioritizes private vehicle purchases and maintenance. reflecting a strong preference for private transport. Meanwhile, middle group like Q3 and Q4 spend more on public transportation service with minimal private vehicle spending. For the other example, for the comparison in information, communication, and financial service, in this category, Q5 households spend significantly more on communication equipment like devices, while spending on communication services is relatively balanced across all groups. Q5 households also lead in insurance spending, showing higher financial flexibility and planning. For the education services, education spending shows stark contrasts. While public secondary education is free, costs related to early childhood and higher education are significant. Q5 households have the highest weight in higher education, indicating greater capacity to fund private schooling and higher education costs. For the conclusion, inflation pressures are not experienced equally across households. The Household Cost Index provides a more relatable and accessible way to understand real household experiences. It's also offered the government a better tool for designing targeted subsidies and stimulus program. However, the study does not yet address factor like differences in quality, price variation, and demand elasticity, important areas for future research. And finally, looking ahead, we plan to broaden the scope of the study by including more household classification and region. We aim to improve future survey design for even more detailed analysis, ensure that stakeholder and public understand their roles and relevance. I will stop here and thank you very much for your attention. Back to Jill.

Thank you very much, Windy, for a sort of concise presentation. Very clear. Thank you. Right. Questions. Right. Well, while people are bringing sort of contemplating what they might ask, that gives me the chance to get my own questions in. got two actually. First, I think, what would you see as a more general point, what would you see as the next priority in developing these in Indonesia? I mean, you've mentioned several things on the last slide. And also, I don't know if you can go back to the housing one, because that raised questions in my mind as to how you measure housing. in the indices. I mean, it's a very controversial topic and there are various different methods. So I wondered how you measure housing costs and how that sort of impacts on the breakdown that you showed.

About measuring the housing, right? In Indonesia, mostly in rural area, they have their own house. But in urban area, some have for sure their own house and rental house and other type of houses. So for the survey, we collect all, but not included the occupied own housing. So for this case, sorry, we not give the presentation here. Yeah, the inflation pressure in Indonesia, mostly for housing, is in the urban area, yeah, urban area.

And that would be where you tend to get higher income people living, whereas the rural areas would tend to be, I mean you get both high and low income in the urban areas, whereas rural areas it's more.

In the next study we plan to study on the other region, for example rural area, we want to see the way or the difference group is different or not in the rural area.

Okay.

Thank you. Right. Other questions?

Thanks for the presentation. So just from a practical perspective, and we spoke about it already and we'll probably come on to it again later, how do you manage the you mentioned how you explain it to users, but how do you manage any confusion between users questioning why the CPI is showing a different inflation rate to the household cost indices? Or do you even get that confusion?

Yeah, I think we already talked about this in the previous presentation, that it depends on the purpose or the objective of using the indicators. Of course, after this, the socialization for the government or other stakeholders must be done because we don't make any confusion in the stakeholder side, right? For sure, maybe they want the lower inflation rather than the higher inflation, but for any policies, I think that we need other another complementary indicator besides CPI. So socialization is needed.

Okay, thank you. Right, final chance for any questions. Right, well, we'll have probably a more general question and answer session at the end, so don't go away, Wendy. stay there. But meanwhile, we will go on to the presentation from the ONS and the Fraser of Allander Institute. And this is a dual presentation. Luke, Michael Idies, yes, from the ONS will start, and then Hannah Randall from the Fraser of Allander Institute at the University of Strathclyde will carry on. Now, we had in this morning a fascinating presentation from Italy showing exactly the differences between the problem of cheapflation and how lower cost brands seemed at times to have much different in higher inflation rates than higher cost ones and this is exactly the sort of problem that the University of Strathclyde was asked to investigate. So I'll now give the floor to Luke who's going to give the background and then Hannah will carry on with what they found.

Hello. Okay, hello everybody. Appreciate this is the last, one of the last talks that we're going to have today. So thank you for all the speakers and everything they've heard so far. This is my first conference, so really good to meet everybody. So as Jill said, I'm Luke Michaelides. I work at the Office for National Statistics. and I lead the production of the household cost indices for a grand total of nine months so far. So I'm fairly new to it all, but what I'm going to do is, well, somewhat repeat some of the things that have been said, which happens when there's been a few presentations before, and then hand it over to Hannah, who will then talk a bit more about her research. So first of all, I'm going to compare some of the different measures of inflation that we use at the Office for National Statistics. What better place to start than the CPI itself, the Consumer Price Index. Well, the CPI measures the average change of prices in goods and services purchased by households over time. and the main measure of inflation in the UK, used mainly by the government and the Bank of England. So, the key features of the CPI is that it excludes household costs, mortgage payments, for example, as you've discussed a bit about. It also follows international standards, and it's based on a basket of goods and services reflecting typical consumer spending. We then have the CPIH, which is like the CPI but includes owner-occupied household costs, or OOH. This is recommended as one of the most comprehensive measures of UK inflation. Again, similar to the CPI, but includes costs associated with maintaining, owning and living in a home, and uses rental equivalents to estimate household costs, which is what homeowners would pay if they rented their own home. Then we go on to the household cost indices, which is, as we have discussed, measures inflation based on how prices change and impact on different household types. So it focuses on a cost of living based, of cost of living pressures for different income groups, pensioners and working age households. And the key features are weight spending differently across households. So for example, low income household groups spend higher percentage on essentials as an example, but unlike the Consumer Price Index and CBIH, it includes mortgage interest payments and student loan repayments. And really, the aim of this is to experience felt inflation rather than just a price change. And finally, we have the Retail Price Index, which is an old measure of, well, a measure we still use, but used in inflation. It includes housing costs. There's no longer an official national statistic, which in the UK is a procedure it goes through due to methodological reasons. It's still used heavily by some contracts, for example, rail fares, student loans, pensions, and it includes mortgage interests on payments, but based on the arithmetic mean formula, which it tends to have an overestimation of inflation compared to CPI and CPIH. So we are encouraged But it's not as encouraged as some of the other measures, CPI and CPIH. So again, um, I'll quickly race through these slides because, um, it's kind of repeating what I've just said, but again. is a summary of CPI, CPIH, different types of weights we've heard of, plutocratic and demographic by previous presenters. There's large measurement biases in the change price acquiring goods and services for consumption and again owner occupied housing is omitted from the CPI but is in the CPIH. And then the household cost indices, sorry I don't have the slides in front of me so I've got this paper, is The weighting is demographic based on the average household spending shares and again does have owner occupied housing and captures payments for houses directly or rather than indirectly through mortgage interest, things like stamp duty, etc. Skip through this. may have seen this slide similar to Sebastian's. We promised we didn't copy each other. This is just a complete coincidence. But this is a nice chart to compare the differences between CPIH and CPI. So as we see, we've got plutocratic weighting versus democratic. Timing of prices changes. For example, the acquisition of some things to CPI and CPIH is where is at the point of acquisition versus payments that like student loans that are paid over time. which can form part of people's lived experience. Right, I will hand over to Hannah to continue the presentation.

Thanks, Luke. So, Luke's presented what the HCIs currently are, and they commissioned myself and a team of researchers at the University of Strathclyde to consider where the HCIs could go next. And so there are three main differences that we were considering that might drive differences in experienced inflation across different types of households. And so the first is that they buy different combinations of goods and services. So they have non-homotheticity. And that's what the HCIs currently consider. But the ONS was really interested in this question of what about households paying different prices for the same or similar goods and services. And so that's what we focused on in this paper and what we'll focus on in this presentation. There is a third reason. So it could be that different types of households substitute differently between goods and services when the relative prices change. And we leave that for future research. So we're looking specifically at how you might be able to incorporate different prices paid by different types of households for the same or similar goods and services. And we're primarily focused on goods prices because what we found when we looked into the literature was that that's what's previously been done. Not a lot of consideration has been given to services in this area so far. So we also leave that for future consideration. So we started off by looking at what other countries are already doing. And so as has been alluded to, there are several countries that do have measures that are similar to the HCIs. So for example, New Zealand and Australia have these, but none that we found that account for household groups paying different prices for the same or similar goods. We did come across some countries that are using very innovative methods, and this came up a lot in the poster session as well, people showcasing what methods their countries are using to collect household surveys and expenditure survey data. So for example, Finland and Norway working on receipt scanning and survey collection for spending information. And an interesting project that we came across that I'll mention again at the end is how it fits into what we were considering. But we came across this project called the HBS project, which as far as we understand, is a Eurostat initiative that potentially was being piloted or used in the Netherlands. but it's an app that combines receipt scanning with some personal characteristics of the people using it and some voluntary data linkage, both to data that's already held by national statistics organizations, like barcode information, as well as secondary data from people's banks and loyalty card information and so on. So it kind of allows this mapping of different spending behaviors that's tied to personal characteristics. So We thought that these data had the potential to form a basis for calculation of HCIs that account for different prices paid, but as far as we're aware, they're not currently being used that way. We also looked at the academic literature, so I'm sure most of you will be aware of large body of literature on non-homotheticity across different household groups. There is also a smaller group of papers that considers households that are paying different prices for the same or similar goods. And some of that was alluded to in the presentation from Italy this morning. So I'll spoil this slide a little bit and say all of these are using commercial household scanner data. And so most of these papers are playing around with what level do you aggregate that at, when do you calculate the change in prices, how do you do the weighting, and so on. So one approach is to calculate a household level price indices with the actual prices that are paid at the barcode level. But then that restricted the sample to goods that had been, you know, barcode identical goods that had been bought by the same household over a one year period at least twice. And so another approach to this is to aggregate it up to the goods module level and calculate the price index there and look at what different types of households are purchasing. third approach was to define strata at the barcode level, but then calculate the expenditure weights by group and apply the inflation rate of each strata. And a final approach is to calculate group-specific inflation rates for each strata at the barcode level and then apply those to baskets at the household level and then look at demographic correlates. And so again, all of these are using commercial household scanner data. Most of them are focused in the US with the exception of two which look at the UK. And so we reviewed different types of data that we thought could be used to connect up, to answer this fundamental question of how do you connect up expenditure behavior at a pretty detailed level with household characteristics. And so the first is you could get this commercial household scanner data from a source like Kantar or Nielsen. And so that links household characteristics directly to prices and quantities at the barcode level. So it's a very granular data set, but it only captures retail purchases and it only captures about, I think, 30 to 40% of goods in the CPI basket. And it will miss out on a lot of things, especially infrequently purchased goods or services. It's also, importantly for the ONS, very expensive data. And so this seemed like probably unlikely to be adopted. So we did look at other types of data that are available and particularly types that ONS is already using. So one is supermarket scanner data. And so that captures prices and quantities, but it doesn't directly link them to household characteristics without some background work. And so it could have some similar limitations to household scanner data in terms of capturing only frequently purchased goods. But we thought using some kind of approach, for example, taking loyalty membership cards and linking household characteristics by that. Another option is the household expenditure survey data. And so this underlies other aspects of CPI. And so the This is the living cost and food survey in the UK. And so that can currently link total spending on certain goods categories, including infrequently purchased goods to household characteristics. And so there's a possibility that if you updated the collection method for it, that you could give scanners to households and ask them to scan in purchases or report infrequently purchased goods. and get the information that you need that way. Some of the limitations here are just a limited sample size, a declining response rate, and potential underreporting of purchases. We did also consider so direct receipt scanning. So this is using maybe something like the HBS app. So taking that would give you the quantities and total cost per item type, and you could calculate an average unit price. If you did scan it through an app or some other -- or linked it to other -- to some other data, you could link it to household characteristics. But it might lack the granularity of other data, and it might face similar limitations to household scanner data. And there was also a concern potentially that, depending on how you recruited your sample, that it could be biased. And finally, we did consider banking data, which does have some individual characteristic information, but it's difficult to link that up into information about households. And it doesn't really have granular enough information about the specific goods that were purchased and their prices to be of particular use here. And so the options that we narrowed down to were that ONS could purchase commercial household scanner data and take a similar approach to what's been taken in academic literature. They could link supermarket scanner data to household characteristics in some way, so potentially using store membership cards. So there's a very interesting paper from Japan from a few years ago that did this and was able to look at spending across different household types. Or potentially, it could use some information about the consumers or the stores post code, which in the UK is quite specific. And we have information at the post code level about the types of people that live there. The third option is to update the living costs and food survey collection methodology to get more specific information about the types of goods that are purchased, the quantities and average unit prices. And then the last option that we considered, which is a bit more experimental, would be to use a periodic or a one-off study. that looks at dispersion in the HCI rates relative to the all household HCI rate and to calculate some kind of household group adjustment factor for class level price indices. So there's a lot of considerations here for ONS, but some of the main ones were that all of these options would take significant time to implement and be costly in terms of the resources that you would require, the time that you would need, and the amount of money that you would have to put into it. So we didn't really find any particularly low-hanging fruit in this area. We were also really conscious that the optimal data and approach is likely to be different across different goods and services. So we really considered focused in on goods because that's what we had the most information about. But you might need a completely different approach for different services. And the last consideration is that some of these approaches, and particularly that last one that I mentioned, kind of a benchmarking study, are not well established. And so you would need to pilot it and provide some proof of concept before you could actually use it reliably.

That's it. Well, thank you both very much for this fascinating paper. Right, questions? Well, I'm getting spoiled because I'm getting a chance to ask my own questions. Oh, no, right, United States, Rob.

Thank you.

Just a real quick question. Could you clarify, I think I heard you say, is this Luke? that the RPI, the retail price index is the one currently used for pension escalation? Could you clarify that? And then B, are you designing the HCIs to replace the RPI for this purpose? Is that the main objective behind your production of these indexes? Or is it just simply mechanism to provide your users with information about how inflation does vary across different households.

Thanks, Rob, for the questions. The first question about pensions, I'm going to be honest and say I don't know a lot about the RPI, so I might hand that over to someone that probably, Chris, he probably does, but I can answer the second one in the sense that at the moment we've spent a lot of time getting our HCIs into a production system. So as I've seen previous, we've been working on the HCIs for a while, but the actual quarterly production of HCIs started December 2023. And I've been focusing on getting that into essentially trying to go for accreditation, which in the UK is a statistic that is It's like a badge for national assist. So the aim that we're trying to achieve is to get that to the national badge and before we think about replacing anything like the RPI. But in theory, if we achieve that, we can use this in a lot of different ways, i.e. policy making and et cetera. But at the moment, it's fairly new, but we are going in the right direction. And tomorrow, I'll say this now, I have a poster tomorrow to talk about specifically about the HCIs and where we've come to in the ONS. So do come along and have a chat to me about anything more specific about the HGIS in the ONS.

Is the RPI the index currently used for the pension escalation in the UK?

So no, it's not. So I think we use CPI for benefits. and pensions, if I am right. I think Luke touched on it. So we have got what we call the landscape of consumer prices in the UK. We have got the macroeconomic measures, which are CPI and CPIH. We have got the household measure, which is the HCIs, which we are developing and iteratively improving. And that is going to tell the story of different inflation rates within households. And then we have got this legacy measure, which is RPI. We could spend a day discussing the issues on RPI and the landscape in the UK. It's a measure that we say is flawed because it's heavily reliant on the Carli, but it's tied up into a lot of government debt in terms of gilts and bonds, and that's one of the reasons why we can't stop producing the RPI until the reliance on the government use of it falls away. But in terms of answering your question, benefits are uprated by CPI.

I'd like to just add to that, if I may, Chris. The RPI, as you said, is officially frowned on. Government, though, uses it not just for government bonds, which is obviously baked into the cake, but it does use it for some other methods. In fact, the government is really quite guilty of inflation rate shopping, as it were. But it is also used in the private sector, and in particular with a lot of private sector pensions. some of whom have tried to change to CPI. I actually found myself being an expert witness in three legal cases as to whether these pension funds could change or not. Some have been able to and some not, it depends on their particular characteristics. So, yes, I think there was Patrick first and then we'll come back to Tanya.

So I heard, I think both presenters say we're thinking about these issues because people say they don't believe the official rates of inflation or something along those lines. And I mean, in the 20 odd years I've been in this game, I think, you know, probably heard that every few months, someone or other says, no, the CPI that you publish doesn't represent my rates of inflation. And I think that, I mean, there's a couple of issues One, I think, so this is probably not a question, it's more comment, that probably that's a challenge to us on how we communicate to the public and how we help people to understand how CPIs are calculated. But I think that there's a, it's not just about people buy different things or I think there's a bunch of kind of behavioral, if we go into kind of behavioral psychology or behavioral economics, you can find a set of kind of concepts there that will help us understand, for example, that people are pretty much fixed on things they've purchased frequently or something they've purchased recently that might have been a very big purchase and this skews or colors their sense of their own inflation rate. So I think, you know, these might be some things that we want to look at to try and understand how to get the message out there rather than, I mean, you know, the end point for me of some of these kind of conversations is kind of the personal inflation calculator model where we each have our own personal inflation rate, which I think abrogates the reason why we do what we do. You know, we're there to provide a message of this is the inflation experience that either groups or the country as a whole is experiencing. So, yeah, I think we could probably look a bit more at those.

Thank you. Tanya?

Thank you. Yeah, mine was a bit more of a comment as well, I guess, really, it's the discussion. The When we originally kind of went live with this at ONS, there was a hell of a lot of discussion around the naming of it. And I think it was really important. Yes, lots of hours spent discussing this. And I was just looking back at the Indonesian slides, actually. And I think you guys call it the household cost index, whereas we make a very specific point to call it household cost indices. to represent the fact that they are groups and they are a set of indices to be taken together to reflect those specific groups inflation. And that the reason I wanted to raise it was because about the RPI, we don't produce a household cost index that represents the whole group of the population, which is what RPI kind of does. So I think that's an important distinction when we're talking about these, that we need to be, they're groups of indices that can be taken together to reflect specific population groups rather than a overarching index, which I think we're all agreed is still the CPI in a sense.

Yeah, so I've got a general question for all the presenters as well. Possibly playing devil's advocate a little bit. So this is important discussion. I really like Rob's analogy about the pump and the gauge. But like Tanya's mentioned naming, I'm going to ask a question about frequency. So if I'm gauging my tire, I'm regularly gauging my tire to make sure it's at the right pressure. I'm not necessarily using that pump unless I need to pump the tyre up, so I'm using that once a year, possibly, because my tyre only goes down. To take the analogy on, what are your thoughts on the frequency of household cost indices? Do we need them monthly? Do we need them quarterly? Or if they're used for benefit uprating, which is typically annual, is there just a need for annual? Like I said, I don't have a strong opinion. I thought that was just a really good analogy, and it's come into my mind. I'd welcome Angela's thoughts as well on what you think about frequency.

All right, well, thank you, Chris, for directing that question at me. I think the answer is frequency is determined by the ability of how far you can stretch the household survey. But at a minimum, it should be annually, I believe. It would be ideal if you could have -- so in the U.S., our household spending survey is continuous. So we are able to produce monthly expenditure estimates from that survey and use them in a superlative formula.

So that's ideal.

for a democratically aggregated index as well. The problem is households have to also be in that survey for a longitudinal period of time and they're not. They're only in for one year. So what do you do about calculating a superlative index which to aggregate across items at the household level? So that's a component that I think is important. Two ways around that. You could do well, spares for a base period, and then in the next year, you have a posh index, and then take the geometric, get your Fisher. And of course, these two indexes would be on completely different samples of households. Or you could do what I think a lot of people are doing, or a lot of some of the literature does, is you can group the households into quote unquote, "like kind homothetic classes." As long as those classes persist over time, now you'll have data for each group and that will enable a Tornquist or a Fisher aggregation at the individual grouping level. So I say annual and at a minimum quarterly or monthly if you have the household spending data to do it.

Thank you, Rob. Sebastian, do you want to come in on that one? Or Wendy, do you want to come in on that one?

Because it's also determined for which group you want to measure. For example, in Indonesia, we need index for escalating the minimum wage, so I think measuring the HCI annually is a complement beside the formula now using the CPI. So annually, I think it's good. Thank you.

Okay. Anybody else either want to come in on that or raise a different question? Okay, Carsten, I think you want to.

Yes, thank you, Jill. Yeah, extremely interesting discussions and presentations here in this partly new area, I mean. looking forward, will this be the development that we need to walk on to, follow two paths in developing CPIs and HCIs, household cost indices. I think in this kind of, maybe you can imagine how a framework for this, for house cost indices could look like. Maybe there are still some issues, I mean, something that has been brought up is about the issues about the weights and so on and so forth, fair enough. I think another issue to take into account when we talk about indices for compensation purposes is to look at it from the user's perspective, I mean the government, whoever will use these. And I mean today, I mean a lot of, for example, where here in Switzerland, our rents is adjusted with the CPI and it has a huge influence, also macroeconomic. In other countries, you adjust households expenditures with a net price index. To avoid this effect you can have with For example, if the government, they introduce new taxes on whatever, alcohol or tobacco, it makes the CPI goes up. And if the household expenditures are, they are reimbursed or compensated by the government, people are compensated when they receive their payment from, or their reimbursement from the government sector. And This is the reason why in some countries you choose to adjust rents not by the CPI but by a net price index. But I think just that you need, if you want to develop a kind of a framework for household cost index, it's also worth to look into this, how what are the perspectives of the users of these indices, what are the needs, and what could be the impact in this circle of, you know, payments going around that otherwise you could have a vicious circle. I mean, the government put taxes on tobacco, you know, people are compensated and get more money and they continue to spent money on tobacco and they are next month or next year they are compensated even more. So I just think this aspect is also worth to -- shouldn't be forgotten. Thanks.

Yes, thank you, Carsten. I think you're absolutely right that the user perspective needs to be brought into this and indeed be very important. I was, just reminds me of what the lady from the ILO said, and she reminded us that the main rationale for these indices has changed over time. Originally, they were for primarily for wage compensation and for use in wage bargaining. Then I think it all changed a bit in the 1990s and it became much more macroeconomic. It was in the 1990s that inflation targeting became a key method of economic management in many countries. Now, perhaps we're seeing again, particularly after the last few years, more interest in going back to the household perspective. And I do like to remind some of my colleagues who've been on the Bank of England Monetary Policy Committee that no method of economic management has survived forever. They all change. Even the international gold standard only lasted for about 40 years, although we had it for longer in the UK. So one day, there'll be inflation targeting may not-- may be superseded by something else or may evolve into something else. Anyway, that's by the by. Right. We've got time for some more questions or more discussion if people want to either ask a question or perhaps please say what you think about some of the questions that have already been raised, because there's clearly a lot to think about. I know we're at the end of the day and we're all getting a bit tired perhaps.

If there's no more questions, Rowan's.

Going to deliver his session again. Are there more questions or comments? So in that case, I think we'll revert to the original program and ask Rohan to take us through the guidelines on using CPIs in producer price indices. But perhaps I can just thank all the speakers here for giving us some fascinating perspectives. And thank you also to those who asked questions and made this such a lively discussion. Thank you.

Okay, and thanks to Jill. Like I said, Jill has been a driving force of the HCIs in the UK, and it's great to have your input, Jill. Rowan, apologies, we pulled you, and now we're bringing you back. So if you want to come up and, I mean, this is an interesting topic in itself about the potential use of CPIs in produce prices.

So the floor is yours if you want to come and join us.

Okay, are we on?

Is this on? Yeah? Okay, wonderful. I always love a last session, I don't mind it at all. So I'm here today, I'm representing the Vorberg Group on service statistics. And some of the work that we've been doing, it's a two-year update. I was here with some other colleagues a couple of years back and we had some wonderful feedback from the room and some key individuals in the IWGPS and really led to some good thoughts and ideas. And what we're looking at here is PPI as an official user of CPI, yes, that's one way to look at it, and on the flip side, CPI as a key contributor. So when we're looking at PPI as an official user, We have that PPI perspective, how are we going to mine value out of a data source that's already being produced, and how do we ratify those differences or leverage the data that's being brought into our statistical domain, but also solve these conceptual differences that naturally exist. So what we'll take a look at is the work that we've done. We'll do a quick, quick, quick recap about the previous work done, and we'll give you an update on where we are today, where we're heading to. And that'll be a run through an additional case study, not in detail, just to mention the feedback we received from the IWGPS. The work so far on merging the theoretical and practical papers, and A little look at SU explanations, advice, program development, how that impacts. And finally, some new insights, but similar story coming out of the most recent alternative data survey that we run every year through the Forberg Group. So the history of it is 2014, we had our first paper that broached the idea of tapping into and using the CPI. And of course, the moment that you unblock that thought, it spreads like wildfire. And we quickly understood that people were pressing that button, but perhaps we didn't have a button designed correctly back in 2014. So we needed to set principles, we needed to develop guidelines to ensure that harmonisation across how different organisations, how different statistical institutions were utilising and tapping into CPIs, whether directly or through data sources that were already being collected, whilst also thinking about how to harmonise across the CPI and PPI program. and harmonizing between the production and expenditure approach to national accounts. And so we set up and had a number of task forces which were wonderful to be a part of and some great people got involved on those fronts. And we came up with two papers. One was a theoretical look at what to think about and the major conceptual differences, and we'll touch on that a little bit. And the second was, well, we are already applying this, so what have we learned from that practical application and how can we share that knowledge, both for PPI producers but also yourselves who are producing CPIs? How can you support this and how can you leverage off PPIs? So the buzzwords that come around that naturally, harmonisation, coherence, institutionalising, ideals, and tapping into and having good metadata, thinking about price statistics as a system rather than two individual products or many individual products. And how can we get right to the core of finding CPIs that are fit for purpose, and what does fit for purpose mean? So, of course, we started at the price statistics fundamentals, which all our price statistics share, whether you're in CPI, PPI, wage price indices, house price indices, we all have these wonderful fundamentals which we can draw from. And we came up with this core set of aspects which we felt were the most important in in getting at those differences and reducing the risks that those differences would introduce bias into PPI programs. And sometimes we're fortunate, sometimes the CPI will almost be an exact replacement or equivalent to the PPI and sometimes not. And sometimes through investigations with our supply use data, we can see where portions of coverage can be solved through the CPI, which means that we can then target our collection resources to those coverage areas where we need to, rather than trying to cover everything multiple times. So reducing that double work and aligning to that concept of collect once, use many times. And that was drilled into me from the Australian Bureau of Statistics in my time there. So the valuation basis is a classic difference, basic price and purchaser price. Classification concordance, as you can imagine. There we can draw opportunities, both draw on opportunities but reflect on risk. And a very pertinent one now that we've gone through this COICOP update and equally on the PPI front, we've had product and industry classifications that have gone through an update. Imputations and quality adjustment, you can reflect on what happened in the pandemic period where our CPI experts may have decided to go down the approach of using the annual movement to plug gaps At the total level, a PPI would conversely, they would utilise an industry level movement and potentially not use the annual movement but utilise a quarterly movement. So those sorts of small aspects don't have to have an impact but do need to be communicated in a reasonable way so that the user, in this case the PPI team, can make a reasonable decision about what they are willing to accept and what they would rather put into practice themselves. So we, the work that was to date was reviewed by the IWGSP and a big thank you to Karsten and Brian and Federico and a colleague of Karsten's who was an expert, or he's an expert in supply use tables. We wanted to have the supply use aspect peer reviewed by an expert in that rather than us as a user. So thank you for that. And that came back with some great feedback and ideas. So one was we had two papers and we want to meld those together, we want to merge those together into one document or one guidance, set of guidance material that either is a standalone document or that can be incorporated into the manuals in their own respective ways. And here we found that we wanted to have some more detail around the valuation basis. As much as that seems straightforward, it's not, and we understand that another part of the feedback was that we want to have written guidelines that are appropriate for advanced producers of price statistics and also our friends that are developing price statistics, PPIs, CPIs, programs, the price statistics programs from the ground up. We want to have guidance that meets both and needs, both institutions' needs. Yeah, and then we have the review of the COICOP, and we've mentioned that. And of course, the big chapter was supply use and effectively using supply use. So a classic example, you might think about veterinary services in some countries that would be almost entirely business to household, and we would say that apart from the valuation difference, We should therefore be able to tap into the CPI and we don't have to collect separately. And in another country, that could be a very different kettle of fish. But it's where we have those high level areas of business to household transactions occurring where we have some very good opportunities. Keeping in mind that import and export play a very important role here. So we have some focus areas which we're working on and we're looking to be finalised with this consolidated paper by the close of the year in the end of September when we will be having the next Vorberg meeting. So the next aspect which I wanted to show you just to give you some insight in case you didn't know or you're not aware of how your own organisations are tapping into the CPI. This is a snapshot of the alternative data survey that we run in the Wahlberg Group. It doesn't just look at the use of CPI, but it looks at a number of different categories of alternative data. But these categories are those which we have found a predominant use of the CPI. And for those of you that might be thinking, we'd really like to start up a PPI program, how do we tap into this when we don't have a supply use table to guide us, then this would be a good place to turn. We have 22 countries which have contributed to this data. And it gives us a good sense of where CPIs could be possible for yourselves as well, although we reflect on the fact that each economy has its own setup and makeup, and it may not be correct. for your purposes. But transportation and storage is a dominant one. Accommodation and food services is another. And I think accommodation services is a very good example of where I've seen data that's been taken into the PPI from administrative data sources, then making improvements in the CPI program. So it doesn't have to be a one-way street. It can be working on both sides of the equation, that we make improvements across both. Another area where we've seen some good improvements is through the quality adjustments, whether it's consumer electronics or whatnot. It's certainly woken up questions of why are we adjusting differently for the different CPI and PPI indicators when we're measuring things which are quite similar. So services obviously stand out. They stand out because there is a direct transaction from the business to the consumer often. But there can be other areas like utilities as well where you might find that a data set which you've successfully hunted down for the CPI also includes data for the PPI. And I would encourage you to be looking for those opportunities as well if you're setting up CPIs or you're reviewing your CPIs. think about your PPI peers and ask yourselves if those administrative data sources that you've managed to attain hold value for your PPI colleagues as well. This next table is just a point of interest. I believe that you've been also looking into some alternative data surveys. This is an example of where PPIs are using multiple sources to fill coverage for the PPIs. And what sticks out for me here is where we're tapping into CPIs and another data source. So we're covering the business to household portion of the output, utilising the CPI, converted, of course, from a purchase of price to a basic price, and that's where you can come in and support that and help that. You might be collecting your prices in, or you are collecting your prices in purchaser prices, but you might add a little bit more data to that, a little bit more questions to that from your providers or from your data collection activities to ensure that you can make that conversion. Now, often it's an easy one. Often it's just the value added tax, but there can be other differences like the subsidies and whatnot. I'm not going to go into these graphs, but we have a few nice visualisations of how CPI is being used and the proportion of it being utilized as the main source. And these can be found on the Vorberg website if it's of interest to you and if you're exploring this as an option. The X is where there's a frequency of multi-source data usage. Okay, so one of the main things that we're considering when we're bringing in a CPI, beyond valuation basis, beyond data sources is, or connected to valuation basis, is the extent to which there's import and export. If there's a large amount of export, obviously that's not going to the CPI and the CPI would only be fit for purpose if the price development could be established without a doubt to be of a similar nature. And vice versa, there's imports coming into CPI that wouldn't necessarily exist in PPI, unless we could establish that there was a similar price development, then we would be suspicious in our want to utilise that. But beyond that, we have very reasonable ways of adjusting the data or segmenting the data for practical use. So first we'll look at the opportunity that the data exists, and then we'll look at the opportunities to be able to reduce the risks of that data not being fit for purpose and causing issues. Okay. So we've done the quick presentation today. I knew that we were a little short on time. I think the big message is that we are putting this paper together, we are writing this paper in a way, or this guidance in a way that is for you as an audience, for CPI practitioners as an audience, and for PPI practitioners as an audience. And our hope is that we can unify both those perspectives And that we're keeping in mind that price statistics is a system in its own right, the system of price statistics, just as we have system of national accounts. We have a production and expenditure side to be mindful of. And we have opportunities to tap into value that already exists and then perhaps awaken opportunities to invest resources into other industries like our retail and wholesale trade. where we would ideally like to move into. If you have any questions or queries, I will be here for the next few days and/or this afternoon. And just as a little opportunity to market that the Vorberg Group meeting has found a home, it will be being held in Copenhagen in Denmark. We're very happy to welcome you or your PPI colleagues. and that will be occurring in late September. Thank you.

Thanks to Rowan, a fascinating presentation again. Having been part of the Volberg group previously, it's really pleasing to see the progress being made and the interaction with the CPI expert group and I'm assuming elsewhere as well. So it's really good that we start to join up. I think we should give Rowan a few minutes of questions. I can see there's a few hands going up. So we'll go to Mexico, Switzerland, and then Tunisia.

PPA, welcome. Are you using COICOP 2018 for both index, COICOP 2018? And did you--do you like the experience of joining both? I mean, do you have any trouble with PPI?

Yeah, okay, excellent. Yeah, so the shift to COICOP 2018 is very much of interest to us. So we did a mapping, we did a concordance using the previous COICOP and our own previous product and industry classifications. And we've also gone through a change as well. So over this coming year, that will be a very important process that we get done is to ensure that we have a good mapping of the new COICOP with the new product and industry codes. Absolutely, yeah. Thank you, thank you.

Thank you very much, Ron, for this excellent presentation. I think this is a very important topic, and this is a very important collaboration. And we are talking about services here in the first place, because that's for services that the corporation is the greatest, the biggest, entirely or partially, or to cooperate in the sample design or for the survey, et cetera. We do know that the new COI COP will have a new division with financial services and insurances. It's a new division, so it's going to be well visible. And we all know that we are, for the CPI, measuring price movements for services is very difficult, challenging. So I think this is a very important collaboration, and we should put on the agenda for both Ottawa Group, UNECE, or VOB, a topic, services and cooperation for SPP and CPI. And it will be very interesting and both sides could look and search for new ideas of collecting are very difficult topics and put our strengths together in order to find a good solution. But this collaboration is very important and this is necessary because our user will compare the price movement for both of them as well. So if there is no differences in the definition of the services, so we should closely collaborate together. And I And I strongly suggest to put that on the agenda for all meetings. Thank you.

Thank you very much. I'll be happy to raise that point with the Volberg group board. That's wonderful. Thank you.

And I think we had Tunisia.

Thank you. Just to say, for me, we speak about three different index. Of course, the methodology, I think it's the same for calculation, but the source of data and to collect data, it's completely different. I don't know how Of course, for analyzing the CPI, you can use PPI and use the export and import index. But I don't know. There is how you can use CPI for PPI. For me, it's completely different. Can you explain how the definition of the price is not the same? For CPI, we speak about the price for consumption, household consumption. For the PPI, it's the first transaction or the price of the unit production. And for export and import, it's completely-- it's the importation, exportation. Do you can explain more how these three index can be used together? Thank you.

Yeah, I should have prepared a poster session. So in both the papers, we do run through this in quite a lot of detail. And you're absolutely right, there are some fundamental differences. But fortunately, there's also a lot of fundamental similarities. If we can say an example, let's think about Yeah, we could go back to veterinary services. So if a business, a veterinary service business, provides a service to a household, and fortunately there's a subsidy as a part of this, and that subsidy will artificially reduce the price for the household, and it'll artificially increase the price or the revenue intake for the veterinarian. And then we once we travel down that path of the transactions, in the case of a service, we don't have the transportation of the goods, so we can remove that from the difference between the basic price and the purchaser price. And Then we might look at insurance around that as well. And we could say, well, no, we don't have that insurance aspect either. So that's great. That would be a risk. So we're removing that risk. And we unpack all of these different valuation risks which could cause a price change, which should cause a price change for the CPI, which absolutely should not cause a price change for the PPI. Then we might look at the value added tax. Of course, a change in value added tax should cause a difference for the household, but it should not cause a difference for a business who's passing on that tax to the household. So we go through all of those differences, and if we can, if we're able to, we adjust those differences out of the CPI if needs be, or we talk with our CPI friends and PPI friends, and we collect the data in a way that we're collecting the data so that it's broken up into those main pieces. to be able to collect once and use many times. So, yeah, I really thank you for that observation because on the outset it can look quite daunting to break these things down and naturally there are product categories where it's simply not possible, it's not feasible and we designate those as not fit for purpose so we do not seek to cover PPI output using a CPI in its place. Another example would be simply in the product definition as per the classification. Sometimes we have an exact match. Great, happy days. We can go ahead and ask ourselves if there's any other risks that we need to mitigate. And other times, the definition of the product is so extremely different. Honey is a classic. I love the definition of honey across different classifications, whether it's in beekeeping or whether it's together with sweets. It can be very differently placed. And therefore, we have to keep those differences in mind. So yeah, I take on board your comments. And I encourage you to read. If you have the time and the energy, read through the documents that we've already put out there, the theoretical documentation, and/or wait until September comes when we pull together the case studies and the theoretical aspects and we'll do a new round of feedback. And yeah, we really appreciate any thoughts and ideas that you have on this. Absolutely, thank you.

One last call for any questions. No? I think, thank you, Rowan. I think this is the start of something pretty good, and yeah, we'll continue to work together, I'm sure. So that brings us to the end of day one. As always, a big thank you to the session chairs, the presenters, the poster session presenters, and of course, everyone for being involved in what's been very fruitful, enjoyable, and quite challenging discussion at times. So we'll be back tomorrow at 9:30, where we're going to be going back to new data sources and collection methods for the second part of that topic. But I bid you all a good evening. Enjoy the weather, and we'll see you all tomorrow. Thank you very much.