WEBVTT 00:00:00.000 --> 00:00:02.140 If you're like pretty much everyone I know, you've 00:00:02.140 --> 00:00:05.540 got this constant feeling in your gut that everything 00:00:05.540 --> 00:00:08.259 just costs more. You know, the gas for your car, 00:00:08.400 --> 00:00:11.279 the groceries, your rent. It just feels like 00:00:11.279 --> 00:00:13.400 it's creeping up and sometimes jumping up from 00:00:13.400 --> 00:00:16.039 last year. That's inflation. It's that slow erosion 00:00:16.039 --> 00:00:18.359 of what your dollar can actually buy. Exactly. 00:00:18.539 --> 00:00:22.359 And for decades, the entire economic world, from 00:00:22.359 --> 00:00:25.239 policymakers to your pension fund, has relied 00:00:25.239 --> 00:00:28.640 on this one. Massive, unbelievably complex number 00:00:28.640 --> 00:00:31.699 to track that feeling. The Consumer Price Index, 00:00:31.980 --> 00:00:35.140 the CPI. The big one. Okay, let's unpack this. 00:00:35.280 --> 00:00:37.840 We always hear the CPI reported as just a single 00:00:37.840 --> 00:00:40.439 number, right? Inflation is up 4 % or 7%, whatever 00:00:40.439 --> 00:00:42.880 the headline is. But it is anything but simple. 00:00:43.000 --> 00:00:46.000 Not at all. It's the single most closely watched 00:00:46.000 --> 00:00:48.539 economic statistic on the planet, really, because 00:00:48.539 --> 00:00:50.840 it has so much leverage. It's the government's 00:00:50.840 --> 00:00:53.700 best shot at taking a huge weighted snapshot 00:00:53.700 --> 00:00:56.299 of what we all spend our money on. The so -called 00:00:56.299 --> 00:00:58.640 market basket of goods and services. And that 00:00:58.640 --> 00:01:01.039 basket is supposed to represent what you, the 00:01:01.039 --> 00:01:04.290 listener, are buying every month. We've got a 00:01:04.290 --> 00:01:06.650 fantastic stack of sources today that really 00:01:06.650 --> 00:01:08.930 get into the weeds of how this colossal number 00:01:08.930 --> 00:01:12.069 is built, why it's always so controversial and 00:01:12.069 --> 00:01:15.010 what the tiny choices in its calculation mean 00:01:15.010 --> 00:01:17.849 for your wallet. It really is the metric that 00:01:17.849 --> 00:01:21.629 tries to, I guess, arbitrate economic reality 00:01:21.629 --> 00:01:24.430 for a whole country. And its main function is, 00:01:24.530 --> 00:01:26.450 on the surface, pretty straightforward. Just 00:01:26.450 --> 00:01:28.790 tracking price changes. Right. Tracking price 00:01:28.790 --> 00:01:31.790 changes over time. If the CPI goes up by, say, 00:01:31.870 --> 00:01:35.010 5 % over a year, it means that basket of typical 00:01:35.010 --> 00:01:37.870 household stuff now costs 5 % more. But the stakes 00:01:37.870 --> 00:01:40.450 are just immense. Oh, absolutely. Because those 00:01:40.450 --> 00:01:42.670 changes, that percentage, it's not just a number 00:01:42.670 --> 00:01:45.099 on a screen. It's hardwired into the economy. 00:01:45.280 --> 00:01:48.760 It's used to index or adjust the real value of 00:01:48.760 --> 00:01:50.700 wages. It determines Social Security payment. 00:01:50.840 --> 00:01:53.319 It determines pensions. It regulates prices for 00:01:53.319 --> 00:01:56.760 utilities. And critically, it lets us deflate 00:01:56.760 --> 00:01:59.180 all these huge monetary numbers so we can actually 00:01:59.180 --> 00:02:01.859 see if the economy is growing in real inflation 00:02:01.859 --> 00:02:04.239 adjusted terms. And this is the big but we're 00:02:04.239 --> 00:02:07.099 going to keep coming back to. Our sources are 00:02:07.099 --> 00:02:09.479 crystal clear on this. While it's indispensable, 00:02:09.900 --> 00:02:13.800 the CPI is, and I'm quoting here, not a perfect 00:02:13.800 --> 00:02:16.840 measure of inflation or the cost of living. It's 00:02:16.840 --> 00:02:18.939 an approximation. It has to be. It's a necessary 00:02:18.939 --> 00:02:21.919 compromise, really, between what economists would 00:02:21.919 --> 00:02:25.620 consider a perfect theoretical ideal and the 00:02:25.620 --> 00:02:30.080 practical reality of needing a reliable number 00:02:30.080 --> 00:02:33.639 fast every single month. And it's in those compromises, 00:02:33.639 --> 00:02:35.460 those little judgment calls, that you find all 00:02:35.460 --> 00:02:38.199 the intense economic and, frankly, political 00:02:38.199 --> 00:02:40.860 fights that are always swirling around this number. 00:02:41.000 --> 00:02:42.800 That's where all the debate lives. And so today, 00:02:42.960 --> 00:02:45.360 we're going to trace the history of this market 00:02:45.360 --> 00:02:47.599 basket idea. Then we're going to dive right into 00:02:47.599 --> 00:02:49.939 the mechanics of how statistical agencies like 00:02:49.939 --> 00:02:52.319 the BLS actually handle this mountain of data. 00:02:52.460 --> 00:02:54.780 The weighting, the collection, all of it. Then 00:02:54.780 --> 00:02:56.879 we're getting into the deep water. The single 00:02:56.879 --> 00:02:59.340 biggest, most heated controversy in the entire 00:02:59.340 --> 00:03:02.099 index, which is what, over 40 % of the whole 00:03:02.099 --> 00:03:05.500 thing? Housing. How in the world do you accurately 00:03:05.500 --> 00:03:08.039 measure the cost of owning a home? The billion 00:03:08.039 --> 00:03:10.400 -dollar question. And finally, we'll look at 00:03:10.400 --> 00:03:12.620 all the different flavors of the CPI they use 00:03:12.620 --> 00:03:15.460 here in the U .S., that whole alphabet soup of 00:03:15.460 --> 00:03:18.400 acronyms, and what they mean for policy, especially 00:03:18.400 --> 00:03:20.960 for Social Security. The cost of living adjustments, 00:03:21.139 --> 00:03:24.240 or COALAs. Right. So let's start at the beginning. 00:03:24.729 --> 00:03:27.689 The anatomy of the basket itself. Where did this 00:03:27.689 --> 00:03:30.789 idea even come from? Well, the idea of trying 00:03:30.789 --> 00:03:34.409 to quantify this feeling of rising prices, it's 00:03:34.409 --> 00:03:37.310 not new at all. It actually has some surprisingly 00:03:37.310 --> 00:03:39.990 simple historical roots. We're not talking about 00:03:39.990 --> 00:03:42.569 something from the 1970s then. No, not even close. 00:03:42.729 --> 00:03:45.430 The concept of a price index was first proposed 00:03:45.430 --> 00:03:48.310 by an English economist, Joseph Lowe, way back 00:03:48.310 --> 00:03:52.139 in 1822. Wow. Okay, so what did his version look 00:03:52.139 --> 00:03:54.080 like? I'm picturing something a lot simpler than 00:03:54.080 --> 00:03:55.719 what we have today. Oh, much simpler. It was 00:03:55.719 --> 00:03:58.280 beautiful in its simplicity, really. Lowe's whole 00:03:58.280 --> 00:04:00.580 approach was just an elementary comparison. He 00:04:00.580 --> 00:04:02.539 said, let's compare the total price of a fixed 00:04:02.539 --> 00:04:05.620 basket of goods and services in one period. Call 00:04:05.620 --> 00:04:07.900 it period zero. The starting point. With the 00:04:07.900 --> 00:04:10.539 total price of the exact same fixed basket in 00:04:10.539 --> 00:04:13.840 the next period, period one. That's it. So that 00:04:13.840 --> 00:04:16.819 basic idea, the fixed basket, that's still the 00:04:16.819 --> 00:04:19.560 DNA of what we do now, even with all the supercomputers 00:04:19.560 --> 00:04:21.860 and, you know, modern economic theory. It is. 00:04:21.920 --> 00:04:24.379 That fixed basket approach remains the elementary 00:04:24.379 --> 00:04:27.379 basis of our current systems. We establish a 00:04:27.379 --> 00:04:29.740 list of what people buy and we just track its 00:04:29.740 --> 00:04:32.089 price over time. It sounds so manageable when 00:04:32.089 --> 00:04:33.730 you put it like that, but then you try to apply 00:04:33.730 --> 00:04:37.149 that to, what, 330 million people buying millions 00:04:37.149 --> 00:04:39.470 of different things? And it scales out of control 00:04:39.470 --> 00:04:41.750 very, very quickly. So let's ground this with 00:04:41.750 --> 00:04:44.430 actual math, but for just one item to make sure 00:04:44.430 --> 00:04:47.250 we get the mechanism. Good idea. So if we look 00:04:47.250 --> 00:04:49.629 at just one specific thing, let's say a gallon 00:04:49.629 --> 00:04:53.689 of milk, the single item price index is calculated 00:04:53.689 --> 00:04:57.050 by taking its current updated cost. What it costs 00:04:57.050 --> 00:04:59.420 today. Right. And you divide that by... what 00:04:59.420 --> 00:05:01.540 it costs in the base period, and then you multiply 00:05:01.540 --> 00:05:03.740 that whole thing by 100. Okay. So if the base 00:05:03.740 --> 00:05:07.839 year is, say, 1982, and milk cost $1 then, and 00:05:07.839 --> 00:05:10.879 today it costs $4, the math is 4 divided by 1 00:05:10.879 --> 00:05:14.860 times 100. The index for milk is 400. Exactly 00:05:14.860 --> 00:05:17.500 right. And that multiplication by 100, that's 00:05:17.500 --> 00:05:20.120 just a convention. It's a way to rescale the 00:05:20.120 --> 00:05:22.920 entire time series so that the base period value 00:05:22.920 --> 00:05:25.399 is always 100. It just creates a nice, clean 00:05:25.399 --> 00:05:27.139 starting point for comparison. It's a consistent 00:05:27.139 --> 00:05:29.629 benchmark. Yep. And this index is then reported 00:05:29.629 --> 00:05:33.129 monthly or quarterly, comparing prices now with 00:05:33.129 --> 00:05:35.290 those in that established price reference month. 00:05:35.430 --> 00:05:37.410 OK, now let's talk scale, because this is where 00:05:37.410 --> 00:05:39.709 the simple theory just smacks right into the 00:05:39.709 --> 00:05:42.149 wall of reality. That simple formula for milk 00:05:42.149 --> 00:05:45.350 has to be done, what, thousands and thousands 00:05:45.350 --> 00:05:48.589 of times? Tens of thousands. Our sources get 00:05:48.589 --> 00:05:51.269 into the sheer volume here. In the United States, 00:05:51.410 --> 00:05:54.149 the Bureau of Labor Statistics collects and averages 00:05:54.149 --> 00:05:57.769 prices for approximately 85 ,000 specific items 00:05:57.769 --> 00:06:01.930 sourced from 22 ,000 different stores and 35 00:06:01.930 --> 00:06:04.379 ,000 rental units. I mean, just think about that. 00:06:04.420 --> 00:06:07.060 That means there are actual people, statisticians, 00:06:07.060 --> 00:06:09.360 whose job it is to track the price of a specific 00:06:09.360 --> 00:06:11.819 brand of laundry detergent in a specific type 00:06:11.819 --> 00:06:14.240 of store in a specific city month after month 00:06:14.240 --> 00:06:16.379 after month. It's a logistical and mathematical 00:06:16.379 --> 00:06:19.699 Everest. It really is. And all of that data then 00:06:19.699 --> 00:06:23.220 has to be compiled into one single cohesive number. 00:06:23.420 --> 00:06:25.560 The weighted average? The weighted average, exactly. 00:06:25.699 --> 00:06:27.860 A weighted average of all those thousands of 00:06:27.860 --> 00:06:30.920 item area indices. And the weight that's assigned 00:06:30.920 --> 00:06:33.819 to each item just reflects how much we as consumers 00:06:33.819 --> 00:06:37.360 spend on it. Right. So the index for gasoline 00:06:37.360 --> 00:06:39.860 matters a whole lot more than the index for, 00:06:40.019 --> 00:06:43.040 I don't know, chewing gum. Precisely. Because 00:06:43.040 --> 00:06:44.860 we spend a heck of a lot more on fuel than we 00:06:44.860 --> 00:06:47.410 do on gum. So before we get into how those weights 00:06:47.410 --> 00:06:49.569 are decided, which sounds like its own can of 00:06:49.569 --> 00:06:51.990 worms, we should talk about the exclusion zone. 00:06:52.230 --> 00:06:54.810 What gets deliberately left out of the basket? 00:06:54.889 --> 00:06:56.769 Because sometimes what's missing tells you a 00:06:56.769 --> 00:06:59.290 lot about what the index is trying to do. That's 00:06:59.290 --> 00:07:01.649 a really critical point. The CPI is designed 00:07:01.649 --> 00:07:03.889 first and foremost to measure current consumption. 00:07:04.189 --> 00:07:06.889 OK, so things we use up. Right. So anything related 00:07:06.889 --> 00:07:09.850 to saving or investing or capital formation, 00:07:10.029 --> 00:07:12.550 it's always excluded. So my stock portfolio, 00:07:12.810 --> 00:07:16.420 my 401k contributions. None of that is in there. 00:07:16.519 --> 00:07:19.259 Nope. Not factored in at all. And there's a geographic 00:07:19.259 --> 00:07:22.399 limit, too. Yes. The CPI is a domestic measure. 00:07:22.699 --> 00:07:25.160 So consumer spending that happens abroad is usually 00:07:25.160 --> 00:07:28.240 excluded. The goal is to focus on the price changes 00:07:28.240 --> 00:07:30.660 that residents experience inside the country's 00:07:30.660 --> 00:07:34.120 borders. Now, the third exclusion, this one is, 00:07:34.259 --> 00:07:35.920 I think, the most politically interesting one. 00:07:37.139 --> 00:07:40.439 Yeah. Our source gives this example where taxes 00:07:40.439 --> 00:07:43.399 made up a staggering 43 percent of the final 00:07:43.399 --> 00:07:46.160 cost of a good. But those taxes are completely 00:07:46.160 --> 00:07:49.579 left out of the CPI calculation. Why? Well, the 00:07:49.579 --> 00:07:51.800 CPI is generally designed to measure the price 00:07:51.800 --> 00:07:55.000 paid for a good or service before most consumption 00:07:55.000 --> 00:07:57.540 taxes are applied. Things like income tax are 00:07:57.540 --> 00:07:59.600 definitely out. The idea is to track the price 00:07:59.600 --> 00:08:02.160 of the commodity itself, not government policy. 00:08:02.519 --> 00:08:04.500 But the implication there is huge, isn't it? 00:08:04.560 --> 00:08:07.060 It is. If a government decides to shift how it 00:08:07.060 --> 00:08:10.540 taxes, people say, it lowers income taxes, which 00:08:10.540 --> 00:08:13.399 are excluded from the CPI, but raises sales taxes, 00:08:13.660 --> 00:08:15.699 which might be included but often aren't. Then 00:08:15.699 --> 00:08:17.959 the CPI might not actually show the real increase 00:08:17.959 --> 00:08:20.000 in what you're paying at the checkout. Exactly. 00:08:20.019 --> 00:08:22.240 It might only capture the change in the product's 00:08:22.240 --> 00:08:24.720 price, completely missing the change in tax policy 00:08:24.720 --> 00:08:26.459 that's making it more expensive for you. Wait 00:08:26.459 --> 00:08:28.300 a minute. So if a government has an incentive 00:08:28.300 --> 00:08:32.179 to keep that headline CPI number low, maybe to 00:08:32.179 --> 00:08:34.460 avoid paying out higher Social Security adjustments, 00:08:34.820 --> 00:08:38.899 they could, in theory, shift the tax burden in 00:08:38.899 --> 00:08:41.419 a way that the CPI doesn't fully capture. It 00:08:41.419 --> 00:08:45.100 creates a potential statistical or political 00:08:45.100 --> 00:08:48.080 incentive. Absolutely. Because the CPI is the 00:08:48.080 --> 00:08:51.059 benchmark for so much, how you define what's 00:08:51.059 --> 00:08:53.320 in and what's out can have massive budgetary 00:08:53.320 --> 00:08:55.019 consequences. It's a very fine line to walk. 00:08:55.259 --> 00:08:57.659 And then the last exclusion is sometimes entire 00:08:57.659 --> 00:08:59.559 groups of people. That's right, especially for 00:08:59.559 --> 00:09:01.799 more specialized indices. Yeah. The goal is always 00:09:01.799 --> 00:09:04.039 to measure the experience of the average consumer. 00:09:04.320 --> 00:09:07.139 So the outliers get trimmed. Sometimes. In the 00:09:07.139 --> 00:09:10.360 UK's old retail price index, the RPI, they used 00:09:10.360 --> 00:09:13.100 to specifically exclude the very highest income 00:09:13.100 --> 00:09:15.419 households and certain low income pensioner households. 00:09:15.700 --> 00:09:17.600 The argument was that their spending patterns 00:09:17.600 --> 00:09:20.019 were just too different from the median experience 00:09:20.019 --> 00:09:23.039 and would skew the average. It just goes to show 00:09:23.039 --> 00:09:25.139 that even defining who the representative consumer 00:09:25.139 --> 00:09:28.379 is, that's a statistical and sometimes political 00:09:28.379 --> 00:09:31.210 choice in itself. So even after all this incredible 00:09:31.210 --> 00:09:33.529 data collection, all the exclusions, all the 00:09:33.529 --> 00:09:36.690 waiting, it's still just an approximation. It 00:09:36.690 --> 00:09:39.070 is. You mentioned that economists have this sort 00:09:39.070 --> 00:09:42.330 of holy grail, the true cost of living index. 00:09:42.690 --> 00:09:44.809 They do. They call it the constant utility index. 00:09:45.009 --> 00:09:47.669 Constant utility. OK. In theory, this perfect 00:09:47.669 --> 00:09:50.889 index would measure exactly how much more money 00:09:50.889 --> 00:09:53.210 you'd need to maintain the exact same standard 00:09:53.210 --> 00:09:56.509 of living, the same level of satisfaction or 00:09:56.509 --> 00:09:59.500 utility. even as prices change around you. So 00:09:59.500 --> 00:10:02.059 if steak prices double, the perfect index would 00:10:02.059 --> 00:10:04.460 track how much you'd need to spend on, I don't 00:10:04.460 --> 00:10:06.740 know, chicken or pork to be just as happy as 00:10:06.740 --> 00:10:08.879 you were before. That's the idea. It accounts 00:10:08.879 --> 00:10:11.480 for substitution. That sounds incredibly useful. 00:10:11.659 --> 00:10:13.600 So why don't we just use that? The problem is 00:10:13.600 --> 00:10:17.519 speed. A perfect constant utility index can only 00:10:17.519 --> 00:10:20.519 really be calculated retrospectively. It requires 00:10:20.519 --> 00:10:23.460 these huge, complex sets of consumer behavior 00:10:23.460 --> 00:10:27.509 data that take months or even years to... properly 00:10:27.509 --> 00:10:30.649 analyze. But the CPI has to come out now. It 00:10:30.649 --> 00:10:33.049 has to be released monthly, within weeks of the 00:10:33.049 --> 00:10:35.409 data being collected. And that need for speed 00:10:35.409 --> 00:10:38.690 forces a pragmatic approximation. And that approximation 00:10:38.690 --> 00:10:40.669 usually takes the form of a fixed weight index, 00:10:40.870 --> 00:10:43.330 what's technically known as a Lesper's index. 00:10:43.610 --> 00:10:46.070 Let's break that term down, Lesper's index, because 00:10:46.070 --> 00:10:48.090 that's really the technical name for this fixed 00:10:48.090 --> 00:10:51.299 basket problem, isn't it? It is. The easiest 00:10:51.299 --> 00:10:54.080 way to think about a true LeSpares index is to 00:10:54.080 --> 00:10:55.840 imagine you're stuck with the exact shopping 00:10:55.840 --> 00:10:58.559 list you used in the base year, let's say 2010. 00:10:58.919 --> 00:11:01.679 Even if over the next 10 years you completely 00:11:01.679 --> 00:11:03.899 stopped buying CDs and started buying streaming 00:11:03.899 --> 00:11:06.220 subscriptions, or you stopped buying beef because 00:11:06.220 --> 00:11:08.929 it got way too expensive. The index assumes you 00:11:08.929 --> 00:11:11.570 are still buying the original 2010 quantity of 00:11:11.570 --> 00:11:14.350 CDs and beef. Which means it's inherently overestimating 00:11:14.350 --> 00:11:16.690 inflation. Exactly. Because it doesn't see us 00:11:16.690 --> 00:11:18.590 doing the smart thing and switching to cheaper 00:11:18.590 --> 00:11:21.129 alternatives. It suffers from what's called substitution 00:11:21.129 --> 00:11:25.009 bias, precisely because it can't track those 00:11:25.009 --> 00:11:28.289 rapid changes in our behavior in real time. It's 00:11:28.289 --> 00:11:31.470 essentially using yesterday's economy to calculate 00:11:31.470 --> 00:11:33.509 today's prices. So we've laid the foundation, 00:11:33.649 --> 00:11:36.080 but now we get to the really tricky part. The 00:11:36.080 --> 00:11:38.659 devil is in the details. And for the CPI, the 00:11:38.659 --> 00:11:41.320 biggest detail is the structure of the weights. 00:11:42.240 --> 00:11:44.259 This is the recipe, right? It's the proportion 00:11:44.259 --> 00:11:46.240 of each ingredient in the basket. If you get 00:11:46.240 --> 00:11:48.059 the weights wrong, the final headline number 00:11:48.059 --> 00:11:50.679 is just wrong. That's absolutely correct. The 00:11:50.679 --> 00:11:52.980 weights are just the proportions assigned to 00:11:52.980 --> 00:11:54.740 all the different spending categories. They have 00:11:54.740 --> 00:11:57.820 to sum up to 100%, of course. And they are what 00:11:57.820 --> 00:12:00.279 define the statistical picture of that typical 00:12:00.279 --> 00:12:02.620 consumer. Let's use the U .S. expenditure weights 00:12:02.620 --> 00:12:05.399 from our sources to really anchor this. elephant 00:12:05.399 --> 00:12:09.460 in the room. It makes up 41 .4 % of the US CPI. 00:12:09.559 --> 00:12:12.679 A massive chunk. I mean, nearly half of all measured 00:12:12.679 --> 00:12:14.620 inflation is being driven by what happens with 00:12:14.620 --> 00:12:17.320 housing costs. And that huge percentage right 00:12:17.320 --> 00:12:19.539 there, it instantly tells you why the housing 00:12:19.539 --> 00:12:22.100 controversy we're about to discuss is so high 00:12:22.100 --> 00:12:24.659 stakes. Right. But OK, moving down the list, 00:12:24.720 --> 00:12:27.409 what are the next biggest pieces? After housing, 00:12:27.549 --> 00:12:30.769 you've got food and beverages at 17 .4%, and 00:12:30.769 --> 00:12:33.529 then transport is right on its heels at 17 .0%. 00:12:33.529 --> 00:12:35.950 So those three, housing, food, and transport, 00:12:36.090 --> 00:12:38.850 that's the vast majority of the index. It is. 00:12:38.909 --> 00:12:40.970 After that, you see medical care, which is a 00:12:40.970 --> 00:12:44.590 huge point of policy debate, at 6 .9%. Then apparel 00:12:44.590 --> 00:12:48.529 at 6 .0%, entertainment at 4 .4%, and then a 00:12:48.529 --> 00:12:51.590 catch -all other category at 6 .9%. So if some 00:12:51.590 --> 00:12:55.750 crisis makes gas prices soar by, say, 50 % overnight, 00:12:56.039 --> 00:12:58.320 That's going to have a much, much bigger impact 00:12:58.320 --> 00:13:00.720 on the final CPI number than if, I don't know, 00:13:00.759 --> 00:13:02.399 movie ticket prices did the same. Of course. 00:13:02.580 --> 00:13:04.740 The weights are the leverage points in the whole 00:13:04.740 --> 00:13:06.639 equation. Yeah. And they're constructed through 00:13:06.639 --> 00:13:09.120 this layered system, what statisticians call 00:13:09.120 --> 00:13:11.320 the hierarchy of indices. You can't just jump 00:13:11.320 --> 00:13:14.440 from those 85 ,000 individual items to the final 00:13:14.440 --> 00:13:16.559 number. It has to be built up in stages. Right. 00:13:16.580 --> 00:13:18.600 It's an aggregation process. So let's clarify 00:13:18.600 --> 00:13:20.220 that hierarchy. What's at the very bottom, the 00:13:20.220 --> 00:13:22.419 most granular level? At the very bottom, you 00:13:22.419 --> 00:13:25.559 have what are called elementary aggregates. price 00:13:25.559 --> 00:13:28.740 of a very specific type of good in a very specific 00:13:28.740 --> 00:13:31.679 place. The example from the source is something 00:13:31.679 --> 00:13:34.600 like the price of men's short sleeve cotton shirts 00:13:34.600 --> 00:13:36.919 sold in San Francisco department stores. That's 00:13:36.919 --> 00:13:40.159 specific. That's specific. It has to be. Because 00:13:40.159 --> 00:13:43.279 price movements are so localized and so dependent 00:13:43.279 --> 00:13:46.559 on the type of store, a shirt at a discount store 00:13:46.940 --> 00:13:49.779 in one city costs something very different than 00:13:49.779 --> 00:13:52.000 at a fancy boutique in another. So you take all 00:13:52.000 --> 00:13:54.259 those elementary aggregates and you average them 00:13:54.259 --> 00:13:57.299 together to form a low -level index, like outer 00:13:57.299 --> 00:14:00.200 garments, and then those combine into a higher 00:14:00.200 --> 00:14:02.480 -level subindex, like clothing, and eventually 00:14:02.480 --> 00:14:04.059 you get to the big categories we just talked 00:14:04.059 --> 00:14:06.779 about, like apparel. Precisely. And it's a standardized 00:14:06.779 --> 00:14:09.559 process, even in the European Union. Their harmonized 00:14:09.559 --> 00:14:11.940 index requires every member country to compute 00:14:11.940 --> 00:14:15.019 about 80 of these prescribed sub indices just 00:14:15.019 --> 00:14:16.600 to make sure everyone is building the number 00:14:16.600 --> 00:14:19.179 in a comparable way. OK, but here's a huge practical 00:14:19.179 --> 00:14:21.179 problem that jumps out at me. The unweighted 00:14:21.179 --> 00:14:23.340 challenge. If you're the statistician tracking 00:14:23.340 --> 00:14:25.399 the price of those shirts in San Francisco, do 00:14:25.399 --> 00:14:28.600 you actually know the sales volume of brand A 00:14:28.600 --> 00:14:31.379 versus brand B or how many were sold on sale 00:14:31.379 --> 00:14:34.960 versus at full price? Often, no. That's the problem. 00:14:35.039 --> 00:14:38.039 At that lowest, most granular level of the elementary 00:14:38.039 --> 00:14:41.059 aggregate, the detailed sales data you'd need 00:14:41.059 --> 00:14:43.919 for perfect weighting. It's just not available 00:14:43.919 --> 00:14:46.120 to the statistical office. So what do they do? 00:14:46.299 --> 00:14:49.220 They know the average price of the 10 different 00:14:49.220 --> 00:14:51.600 shirts they're tracking, but they don't know 00:14:51.600 --> 00:14:53.559 the market share of each of those 10 shirts. 00:14:54.090 --> 00:14:56.750 So they often have to compute the price using 00:14:56.750 --> 00:14:59.429 an unweighted arithmetic or geometric mean. They 00:14:59.429 --> 00:15:01.950 just average them. They just average the prices 00:15:01.950 --> 00:15:04.009 together, basically assuming that each shirt 00:15:04.009 --> 00:15:05.909 they're tracking contributes equally to what 00:15:05.909 --> 00:15:08.110 people are buying, which, of course, we know 00:15:08.110 --> 00:15:10.090 is almost never true. That feels like a bit of 00:15:10.090 --> 00:15:12.730 a weak link in the chain, you know, relying on 00:15:12.730 --> 00:15:14.789 a simple average right at the foundation. It 00:15:14.789 --> 00:15:17.549 has been for a long time, but the data landscape 00:15:17.549 --> 00:15:19.590 is changing very, very quickly. Which brings 00:15:19.590 --> 00:15:22.509 us to barcode scanner data. The game changer. 00:15:23.240 --> 00:15:26.379 Our sources really highlight this. This technology 00:15:26.379 --> 00:15:29.519 is slowly making it possible to have real weights, 00:15:29.700 --> 00:15:32.360 even at that most detailed level. So instead 00:15:32.360 --> 00:15:34.600 of a person with a clipboard writing down a price. 00:15:34.860 --> 00:15:38.200 Exactly. Instead of that, scanner data gives 00:15:38.200 --> 00:15:40.899 the statistical office the actual fails volumes 00:15:40.899 --> 00:15:44.960 for specific products. Brand X, SQ number, Y 00:15:44.960 --> 00:15:47.399 plus where it was bought, and the exact price 00:15:47.399 --> 00:15:50.279 paid, including any discounts. So scanner data 00:15:50.279 --> 00:15:52.600 is finally getting the index a little closer 00:15:52.600 --> 00:15:55.840 to that constant utility ideal by cutting down 00:15:55.840 --> 00:15:58.299 on that substitution bias at the ground level. 00:15:58.480 --> 00:16:00.399 It is, and it helps with the whole data collection 00:16:00.399 --> 00:16:03.379 process itself. It addresses a major operational 00:16:03.379 --> 00:16:05.419 problem, which is what they call respondent burden. 00:16:05.519 --> 00:16:07.360 Making it easier for the people who participate 00:16:07.360 --> 00:16:09.820 in the surveys. Right. One source mentioned Statistics 00:16:09.820 --> 00:16:11.799 Iceland found that when people in their surveys 00:16:11.799 --> 00:16:14.000 are given detailed cash register receipts from 00:16:14.000 --> 00:16:16.460 scanner purchases, they don't have to write down 00:16:16.460 --> 00:16:19.039 every single item they bought in a diary. They 00:16:19.039 --> 00:16:20.879 can just keep the receipt. They just record the 00:16:20.879 --> 00:16:23.600 total and keep the receipt. It dramatically reduces 00:16:23.600 --> 00:16:26.120 the burden on them. It increases the accuracy 00:16:26.120 --> 00:16:28.299 of what was bought. And crucially, it gives you 00:16:28.299 --> 00:16:31.659 that specific point of purchase data. So you 00:16:31.659 --> 00:16:33.639 know if it was bought at a supermarket versus 00:16:33.639 --> 00:16:36.169 a corner store. Which helps you estimate those 00:16:36.169 --> 00:16:38.029 outlet -type weights we were just talking about. 00:16:38.110 --> 00:16:40.350 It's a huge leap forward. Okay, so we've got 00:16:40.350 --> 00:16:42.789 the weights and we've got the data sources, these 00:16:42.789 --> 00:16:45.990 periodical household expenditure surveys and 00:16:45.990 --> 00:16:49.149 the big national accounts. But these two sources 00:16:49.149 --> 00:16:51.549 don't always line up perfectly, do they? No, 00:16:51.549 --> 00:16:54.289 they don't. And managing that divergence is a 00:16:54.289 --> 00:16:57.309 huge part of the job. For instance, a household 00:16:57.309 --> 00:16:59.509 survey usually excludes what foreign visitors 00:16:59.509 --> 00:17:02.190 spend. But the national accounts, they track 00:17:02.190 --> 00:17:04.450 total consumption in the whole economy, including 00:17:04.450 --> 00:17:06.490 tourists. So you have to adjust for that. You 00:17:06.490 --> 00:17:09.269 have to adjust. Or, on the flip side, the national 00:17:09.269 --> 00:17:11.410 accounts include something called imputed rent 00:17:11.410 --> 00:17:14.230 for people who own their homes, an estimate of 00:17:14.230 --> 00:17:16.970 what the rental value would be. A household survey 00:17:16.970 --> 00:17:19.029 might only look at what people actually paid 00:17:19.029 --> 00:17:21.410 out of pocket. So you have to harmonize those 00:17:21.410 --> 00:17:23.509 different definitions. What's really fascinating 00:17:23.509 --> 00:17:25.730 here and something I think most people miss is 00:17:25.730 --> 00:17:27.490 that this isn't just about counting how many 00:17:27.490 --> 00:17:30.230 dollars were spent on beef nationally. No, it's 00:17:30.230 --> 00:17:32.130 so much more complex. You have to account for 00:17:32.130 --> 00:17:34.450 regional differences and store type differences 00:17:34.450 --> 00:17:37.230 all at the same time. It's exponential complexity. 00:17:37.529 --> 00:17:40.150 I mean, the ideal would be to estimate the spending 00:17:40.150 --> 00:17:43.390 for every single detailed category, for every 00:17:43.390 --> 00:17:46.029 single type of outlet in every single region 00:17:46.029 --> 00:17:48.150 of the country. Which is impossible. It's impossible. 00:17:48.450 --> 00:17:51.440 So you have to make educated estimates. You might 00:17:51.440 --> 00:17:54.220 lack detailed regional data. So you assume that, 00:17:54.279 --> 00:17:57.220 say, the northern region, which has 24 % of the 00:17:57.220 --> 00:18:00.559 population, also consumes 24 % of all sliced 00:18:00.559 --> 00:18:03.880 bread. A rough proxy. A rough but necessary proxy. 00:18:04.019 --> 00:18:06.519 Then you apply the national outlet data, say 00:18:06.519 --> 00:18:09.400 70 % of sliced bread is sold in supermarkets, 00:18:09.420 --> 00:18:11.920 to get an estimate for sliced bread sold in supermarkets 00:18:11.920 --> 00:18:14.130 in the northern region. They use every scrap 00:18:14.130 --> 00:18:16.390 of data they have because a rough estimate is 00:18:16.390 --> 00:18:18.329 better than ignoring that spending completely. 00:18:18.569 --> 00:18:20.349 And even with all this incredibly meticulous 00:18:20.349 --> 00:18:23.349 work, the whole system still suffers from what 00:18:23.349 --> 00:18:26.150 you called the cost of infrequency. The reweighting 00:18:26.150 --> 00:18:29.349 problem. Right. This is the LaSperre's index 00:18:29.349 --> 00:18:33.369 problem in the real world. It is. Ideally. The 00:18:33.369 --> 00:18:35.609 weights in the CPI would reflect what we're spending 00:18:35.609 --> 00:18:39.630 our money on right now this year. But in reality, 00:18:39.789 --> 00:18:41.630 they're historical. They're always looking in 00:18:41.630 --> 00:18:44.269 the rearview mirror. They are. They reflect spending 00:18:44.269 --> 00:18:46.569 patterns from the last big survey, which might 00:18:46.569 --> 00:18:49.869 be a year or two old. Rewaiting is incredibly 00:18:49.869 --> 00:18:52.329 expensive and labor intensive for statistical 00:18:52.329 --> 00:18:54.829 offices. So they just can't do it constantly. 00:18:55.049 --> 00:18:57.390 And the result of that is that the index is always 00:18:57.390 --> 00:18:59.630 chasing a moving target. It creates this huge 00:18:59.630 --> 00:19:02.049 time lag before new types of products ever get 00:19:02.049 --> 00:19:05.049 into the basket. A significant delay. The sources 00:19:05.049 --> 00:19:07.950 give two perfect examples. Internet service subscriptions 00:19:07.950 --> 00:19:10.710 and digital cameras. OK, what happened there? 00:19:11.099 --> 00:19:13.339 When digital cameras first exploded in popularity, 00:19:13.619 --> 00:19:15.180 they initially had to be lumped into the same 00:19:15.180 --> 00:19:18.299 elementary aggregate as film cameras. But there's 00:19:18.299 --> 00:19:20.440 a completely different technology. Completely 00:19:20.440 --> 00:19:23.279 different. But the basket structure hadn't been 00:19:23.279 --> 00:19:25.279 updated to give them their own category yet. 00:19:25.400 --> 00:19:27.900 So if a brand new category just appears on the 00:19:27.900 --> 00:19:31.420 market like VR headsets or some new digital service, 00:19:31.640 --> 00:19:34.859 the official CPI basically pretends it doesn't 00:19:34.859 --> 00:19:36.980 exist until the next big re -weighting cycle. 00:19:37.390 --> 00:19:39.910 Which means the official inflation rate for that 00:19:39.910 --> 00:19:43.069 whole period is fundamentally incomplete. The 00:19:43.069 --> 00:19:44.829 older the weights get, the more the official 00:19:44.829 --> 00:19:47.210 number diverges from our actual lived reality. 00:19:48.069 --> 00:19:50.529 Okay. So we've established that housing is the 00:19:50.529 --> 00:19:53.890 single biggest piece of the pie. 41 .4 % in the 00:19:53.890 --> 00:19:56.190 U .S. The 800 -pound gorilla. And now we are 00:19:56.190 --> 00:19:58.450 wading into the philosophical deep end. Because 00:19:58.450 --> 00:20:00.690 measuring the cost of owner -occupied housing 00:20:00.690 --> 00:20:02.950 might be the single most divisive, controversial 00:20:02.950 --> 00:20:05.390 element in all of statistics. It's a genuine 00:20:05.390 --> 00:20:07.809 dilemma. Because a house, it's two things at 00:20:07.809 --> 00:20:10.390 once. It's this massive, long -term, durable 00:20:10.390 --> 00:20:13.549 asset, an investment. Right. And it's also something 00:20:13.549 --> 00:20:15.750 that provides a continuous consumption service. 00:20:15.930 --> 00:20:18.549 Yeah. Shelter. The whole problem is how do you 00:20:18.549 --> 00:20:21.410 surgically separate the consumption cost from 00:20:21.410 --> 00:20:23.650 the investment component? So let's break down 00:20:23.650 --> 00:20:26.430 the three competing ways to do this. Approach 00:20:26.430 --> 00:20:29.849 one, the economist's view, which is all based 00:20:29.849 --> 00:20:33.470 on opportunity cost. The core idea here is that 00:20:33.470 --> 00:20:35.990 a homeowner... is consuming the service that 00:20:35.990 --> 00:20:38.890 the house provides. And that service is basically 00:20:38.890 --> 00:20:41.849 identical to what a renter gets. Okay, so the 00:20:41.849 --> 00:20:43.549 cost isn't what you pay on the mortgage. No. 00:20:43.869 --> 00:20:46.789 The true cost in this view is the opportunity 00:20:46.789 --> 00:20:49.710 cost, is what the owner gives up by living in 00:20:49.710 --> 00:20:52.130 the house themselves instead of, say, renting 00:20:52.130 --> 00:20:54.309 it out to someone else. And the most famous version 00:20:54.309 --> 00:20:56.750 of this is the rental equivalent approach, which 00:20:56.750 --> 00:20:59.230 is what the U .S. largely uses. Right. Under 00:20:59.230 --> 00:21:01.789 this approach, the cost of owning your home is 00:21:01.789 --> 00:21:03.529 the rent you could get if you leased it out. 00:21:04.079 --> 00:21:05.960 They're trying to price the service of shelter. 00:21:06.180 --> 00:21:09.079 But the practical problem seems huge. It is. 00:21:09.119 --> 00:21:11.500 How do you estimate the rental value of a house 00:21:11.500 --> 00:21:14.200 that isn't for rent? If you live in a unique 00:21:14.200 --> 00:21:16.579 five -bedroom house and there are no other five 00:21:16.579 --> 00:21:19.539 -bedroom houses for rent anywhere near you, the 00:21:19.539 --> 00:21:21.859 statisticians have to find comparable properties. 00:21:22.220 --> 00:21:25.849 Maybe two three -bedroom apartments? Maybe. And 00:21:25.849 --> 00:21:27.349 then they have to statistically adjust for the 00:21:27.349 --> 00:21:29.950 differences. And that whole process is just full 00:21:29.950 --> 00:21:32.730 of error and subjectivity. So what's the other 00:21:32.730 --> 00:21:35.630 side of that opportunity cost coin, the alternative 00:21:35.630 --> 00:21:38.950 cost approach? This one defines the cost differently. 00:21:39.109 --> 00:21:42.089 It looks at the capital expense. It asks, what 00:21:42.089 --> 00:21:44.250 is the interest the owner would have earned if 00:21:44.250 --> 00:21:46.230 they sold the house, invested that money for 00:21:46.230 --> 00:21:48.549 a year, and then bought the house back after 00:21:48.549 --> 00:21:51.299 accounting for depreciation? It's trying to capture 00:21:51.299 --> 00:21:53.559 the cost of having all that capital tied up in 00:21:53.559 --> 00:21:55.680 a house instead of in the stock market. Exactly. 00:21:56.000 --> 00:21:58.500 But there's a huge problem. Wait, let me guess. 00:21:58.759 --> 00:22:01.720 If house prices are going up really fast, the 00:22:01.720 --> 00:22:04.019 capital gain you get from owning it for that 00:22:04.019 --> 00:22:06.359 year could be bigger than the interest and the 00:22:06.359 --> 00:22:09.059 depreciation combined. You got it. And that could 00:22:09.059 --> 00:22:11.319 make the cost of housing in the CPI negative. 00:22:11.599 --> 00:22:14.059 Which would drag the whole inflation number down 00:22:14.059 --> 00:22:16.759 even while home prices are soaring. Right. It 00:22:16.759 --> 00:22:19.849 creates extreme volatility. If prices rise fast, 00:22:20.029 --> 00:22:22.309 the cost goes negative. If they suddenly fall, 00:22:22.549 --> 00:22:25.930 the cost spikes way up. And statisticians absolutely 00:22:25.930 --> 00:22:28.329 hate that kind of volatility in a headline number. 00:22:28.490 --> 00:22:30.549 And you brought this up earlier. There's an intellectual 00:22:30.549 --> 00:22:33.309 inconsistency here, isn't there? If you use this 00:22:33.309 --> 00:22:35.589 logic for a house, why not for everything else 00:22:35.589 --> 00:22:37.670 that lasts a long time? You should if you're 00:22:37.670 --> 00:22:40.390 being logically consistent. Your refrigerator 00:22:40.390 --> 00:22:44.549 provides a cooling service for 15 years. An appendectomy 00:22:44.549 --> 00:22:47.079 provides the service of good health. for the 00:22:47.079 --> 00:22:49.940 rest of your life. My couch, my car. All durable 00:22:49.940 --> 00:22:52.440 consumer goods and services should, by this logic, 00:22:52.599 --> 00:22:55.539 be measured by their rental equivalent or their 00:22:55.539 --> 00:22:58.660 opportunity cost. The fact that economists generally 00:22:58.660 --> 00:23:02.599 only apply this incredibly complex logic to housing. 00:23:02.859 --> 00:23:05.400 And not to my sofa. It shows that it's really 00:23:05.400 --> 00:23:07.339 about the sheer size of the housing category, 00:23:07.579 --> 00:23:10.599 not about pure theoretical consistency. Okay, 00:23:10.640 --> 00:23:12.680 that brings us to the second way, approach two, 00:23:12.740 --> 00:23:15.380 the spending approach or the debt profile method. 00:23:15.880 --> 00:23:18.539 This side just says, forget all the theory. Let's 00:23:18.539 --> 00:23:20.640 just track what homeowners actually spend. This 00:23:20.640 --> 00:23:22.599 feels the most intuitive to the average person, 00:23:22.700 --> 00:23:25.359 right? It focuses on the actual outlays, repairs, 00:23:25.640 --> 00:23:28.180 maintenance, property taxes, and the big one, 00:23:28.279 --> 00:23:30.400 mortgage interest payments. Because that's the 00:23:30.400 --> 00:23:33.079 bill you see every month. Right. But the complexity 00:23:33.079 --> 00:23:36.700 of actually implementing this is kind of horrifying. 00:23:37.420 --> 00:23:39.359 Because you can't just track current spending. 00:23:39.500 --> 00:23:41.160 You have to compare it to the spending in the 00:23:41.160 --> 00:23:43.619 base period. You have to create a counterfactual. 00:23:43.700 --> 00:23:45.839 You do. If you're calculating the CPI today, 00:23:46.000 --> 00:23:48.240 you have to compare the interest someone is paying 00:23:48.240 --> 00:23:50.539 now to the interest that would have been paid 00:23:50.539 --> 00:23:53.640 on a similar house that was bought and mortgaged 00:23:53.640 --> 00:23:55.680 way back in the original reference period, like 00:23:55.680 --> 00:23:58.420 2005 or something. So you're tracking the historical 00:23:58.420 --> 00:24:01.180 debt profile of a hypothetical house bought 15 00:24:01.180 --> 00:24:03.700 years ago and then adjusting it for today's conditions. 00:24:04.039 --> 00:24:07.319 It's a staggering... Statistical exercise. And 00:24:07.319 --> 00:24:09.420 it leads directly to the great policy paradox. 00:24:09.740 --> 00:24:13.220 Which is, if the central bank raises interest 00:24:13.220 --> 00:24:17.059 rates to stop inflation, what does this method 00:24:17.059 --> 00:24:20.079 do to the CPI? It makes inflation look worse. 00:24:20.480 --> 00:24:23.420 That rise in interest rates, which is designed 00:24:23.420 --> 00:24:26.000 to cool the economy, shows up in the debt profile 00:24:26.000 --> 00:24:28.559 method as a higher cost of homeownership. So 00:24:28.559 --> 00:24:30.619 it pushes the measured inflation rate up. It's 00:24:30.619 --> 00:24:32.799 a broken feedback loop. The cure for inflation 00:24:32.799 --> 00:24:36.160 looks like more inflation. Precisely. It makes 00:24:36.160 --> 00:24:39.140 the index look unstable and counterproductive 00:24:39.140 --> 00:24:41.559 to the very policy it's supposed to be guiding. 00:24:41.740 --> 00:24:44.599 And is this method applied consistently to all 00:24:44.599 --> 00:24:47.400 our debt, car loans, credit cards? No, which 00:24:47.400 --> 00:24:49.920 is another big consistency failure. If you're 00:24:49.920 --> 00:24:52.819 going to cover mortgage interest, logic says 00:24:52.819 --> 00:24:55.279 you should cover the interest on all consumer 00:24:55.279 --> 00:24:58.099 credit instead of just the full price of the 00:24:58.099 --> 00:25:00.420 things bought on credit. But most countries that 00:25:00.420 --> 00:25:03.579 use this debt profile method don't do that. They 00:25:03.579 --> 00:25:06.140 treat mortgages as a special case. Okay, finally, 00:25:06.259 --> 00:25:08.700 let's get to the third way, approach three, the 00:25:08.700 --> 00:25:11.380 transaction prices approach. This seems like 00:25:11.380 --> 00:25:14.180 the simple, direct method. Just treat a house 00:25:14.180 --> 00:25:16.339 like you'd treat a washing machine. This is the 00:25:16.339 --> 00:25:18.329 most straightforward interpretation. You just 00:25:18.329 --> 00:25:20.769 use the agreed -upon transaction prices for houses, 00:25:21.009 --> 00:25:23.269 you ignore all the financing and borrowing stuff, 00:25:23.450 --> 00:25:26.009 and you only include the sale of new dwellings. 00:25:26.109 --> 00:25:28.710 Why only new ones? You exclude secondhand sales 00:25:28.710 --> 00:25:30.589 because that's just an asset transfer between 00:25:30.589 --> 00:25:33.029 two consumers. It's not new consumption for the 00:25:33.029 --> 00:25:34.970 economy as a whole. This sounds so much cleaner. 00:25:35.410 --> 00:25:37.630 But you said it creates a fight between investment 00:25:37.630 --> 00:25:40.630 and consumption. This is where the CPI clashes 00:25:40.630 --> 00:25:43.380 with broader economic measurement. In the big 00:25:43.380 --> 00:25:45.859 system of national accounts, the purchase of 00:25:45.859 --> 00:25:48.980 a new house is almost always treated as an investment, 00:25:49.099 --> 00:25:52.180 as capital formation, not as consumption. I see. 00:25:52.279 --> 00:25:54.619 So if the CPI, which is supposed to be a consumption 00:25:54.619 --> 00:25:57.519 index, suddenly starts including new houses as 00:25:57.519 --> 00:25:59.880 consumption, it creates a serious break with 00:25:59.880 --> 00:26:02.059 all the other major government statistics. And 00:26:02.059 --> 00:26:04.339 economists really, really want the definition 00:26:04.339 --> 00:26:07.359 of consumption to be consistent everywhere. And 00:26:07.359 --> 00:26:10.059 then there's one last almost philosophical problem 00:26:10.059 --> 00:26:13.039 with the land itself. Right. The argument is 00:26:13.039 --> 00:26:15.980 that land is this irreproducible permanent asset. 00:26:16.039 --> 00:26:19.440 You can't consume land. Therefore, the price 00:26:19.440 --> 00:26:22.279 used in the CPI should exclude the value of the 00:26:22.279 --> 00:26:24.539 land underneath the house. So the statisticians 00:26:24.539 --> 00:26:27.299 would have to somehow figure out what the empty 00:26:27.299 --> 00:26:29.480 lot would be worth and subtract that from the 00:26:29.480 --> 00:26:32.420 sale price of the house? Yes, which is incredibly 00:26:32.420 --> 00:26:34.950 difficult, if not impossible. to do reliably, 00:26:35.170 --> 00:26:37.069 especially in a dense city where the house's 00:26:37.069 --> 00:26:39.289 value is completely tied up with the land's value. 00:26:39.470 --> 00:26:41.950 So to sum up the housing confusion, it's this 00:26:41.950 --> 00:26:43.849 three way battle. And the reasons it's never 00:26:43.849 --> 00:26:46.730 been solved are feasibility, volatility and just 00:26:46.730 --> 00:26:49.410 plain old logical disagreement. That's it exactly. 00:26:50.089 --> 00:26:52.130 Statisticians struggle with the feasibility of 00:26:52.130 --> 00:26:54.349 getting good data. They worry about the volatility 00:26:54.349 --> 00:26:56.609 if they include interest rates or capital gains. 00:26:56.950 --> 00:26:59.190 And the theorists can't agree on which approach 00:26:59.190 --> 00:27:02.160 is the most logically consistent. It remains 00:27:02.160 --> 00:27:04.359 the most fundamentally challenging piece of the 00:27:04.359 --> 00:27:07.539 entire index. So these fights over data and housing, 00:27:07.759 --> 00:27:10.079 they're happening all over the world. But here 00:27:10.079 --> 00:27:12.839 in the United States, the debate gets this specific 00:27:12.839 --> 00:27:15.680 political urgency because the Bureau of Labor 00:27:15.680 --> 00:27:18.720 Statistics doesn't compute just one CPI. No, 00:27:18.779 --> 00:27:20.799 they compute several. each one aimed at a different 00:27:20.799 --> 00:27:23.079 population or using a different method. It's 00:27:23.079 --> 00:27:25.839 the alphabet soup of U .S. CPIs. It is. You have 00:27:25.839 --> 00:27:28.119 the CPI -U for all urban consumers. That's the 00:27:28.119 --> 00:27:29.900 main headline number you always hear about. Then 00:27:29.900 --> 00:27:32.799 there's the CPI -W for urban wage earners and 00:27:32.799 --> 00:27:35.880 clerical workers. There's a rarely used CPI -E 00:27:35.880 --> 00:27:38.509 for the elderly. And then there's these statistically 00:27:38.509 --> 00:27:41.809 refined CCPIU, which is the chain CPI for all 00:27:41.809 --> 00:27:44.309 urban consumers. And the really crucial thing 00:27:44.309 --> 00:27:46.170 to understand here is that all of these different 00:27:46.170 --> 00:27:49.609 indices, they're using the exact same raw price 00:27:49.609 --> 00:27:51.809 data. That's right. They all start with those 00:27:51.809 --> 00:27:55.470 8 ,018 separate item area indices we talked about. 00:27:55.589 --> 00:27:58.009 The only thing that's different between a CPIU 00:27:58.009 --> 00:28:01.029 and a CKIW is the weights that are applied to 00:28:01.029 --> 00:28:03.549 that data. Exactly. The weights are adjusted 00:28:03.549 --> 00:28:06.069 to reflect the spending patterns of their target 00:28:06.069 --> 00:28:08.710 population. So the standard ones, the CPI -U 00:28:08.710 --> 00:28:12.089 and CPI -W, they use weights that are held constant 00:28:12.089 --> 00:28:15.029 for 24 months. Right. They're only updated in 00:28:15.029 --> 00:28:17.630 January of even -numbered years, which means 00:28:17.630 --> 00:28:21.430 for up to two full years, the index is assuming 00:28:21.430 --> 00:28:23.210 that you're buying the exact same proportional 00:28:23.210 --> 00:28:25.670 basket of stuff, no matter what happens to prices. 00:28:25.910 --> 00:28:28.210 And that fixed weight structure is exactly why 00:28:28.210 --> 00:28:29.970 we have the substitution effect problem we talked 00:28:29.970 --> 00:28:34.049 about earlier. of beef skyrockets. The normal 00:28:34.049 --> 00:28:36.710 CPIU just assumes I'm still buying the same amount 00:28:36.710 --> 00:28:38.190 of beef, even though I've obviously switched 00:28:38.190 --> 00:28:41.049 to cheaper chicken. That is the classic substitution 00:28:41.049 --> 00:28:44.450 bias. And that is exactly what the CCPIU, the 00:28:44.450 --> 00:28:46.970 chained CPI, was designed to fix. How does it 00:28:46.970 --> 00:28:50.170 do it? The chained CPI updates its weights every 00:28:50.170 --> 00:28:52.910 month. Oh, wow. It immediately reflects changes 00:28:52.910 --> 00:28:56.589 in our spending patterns. If millions of us substitute 00:28:56.589 --> 00:28:59.650 chicken for beef, That shift shows up in the 00:28:59.650 --> 00:29:02.880 very next month's CCPIU calculation. So it's 00:29:02.880 --> 00:29:05.759 a much more dynamic, much more accurate measure 00:29:05.759 --> 00:29:07.779 of what's actually happening. Statistically, 00:29:07.859 --> 00:29:10.279 yes. It's a superior measure of inflation because 00:29:10.279 --> 00:29:13.160 it tracks consumer behavior in almost real time. 00:29:13.299 --> 00:29:15.839 And because we consumers always tend to buy the 00:29:15.839 --> 00:29:18.559 cheaper thing when prices change, the chain CPI 00:29:18.559 --> 00:29:21.859 has to, by its very nature, run a little lower 00:29:21.859 --> 00:29:24.079 than the fixed weight CPIU, right? That's right. 00:29:24.180 --> 00:29:25.980 Because the fixed weight one is artificially 00:29:25.980 --> 00:29:28.160 inflated by pretending we're still stuck buying 00:29:28.160 --> 00:29:30.839 the expensive items. The chain CPI accounts for 00:29:30.839 --> 00:29:33.259 our adaptation. OK, now let's get to the biggest 00:29:33.259 --> 00:29:35.019 political pressure point in this whole system, 00:29:35.140 --> 00:29:38.220 the COLA controversy. The cost of living adjustment 00:29:38.220 --> 00:29:41.140 for, what, 70 million Social Security recipients 00:29:41.140 --> 00:29:44.160 is based on the annual increase in the CPIW. 00:29:44.200 --> 00:29:46.599 The Index for Urban Wage Earners and Clerical 00:29:46.599 --> 00:29:49.400 Workers. Yes. And this is just a classic case 00:29:49.400 --> 00:29:52.599 of a statistical mismatch with huge social consequences. 00:29:53.019 --> 00:29:55.980 It is. You're using an index designed for the 00:29:55.980 --> 00:29:58.940 working population to adjust the incomes of the 00:29:58.940 --> 00:30:01.349 retired population. And Social Security recipients 00:30:01.349 --> 00:30:04.599 are primarily the elderly. And the elderly spend 00:30:04.599 --> 00:30:06.960 a substantially larger portion of their income 00:30:06.960 --> 00:30:09.700 on health care. Some estimates say between 20 00:30:09.700 --> 00:30:12.640 and 40 percent. Then younger working people do. 00:30:12.839 --> 00:30:15.599 And we all know that for decades, inflation in 00:30:15.599 --> 00:30:17.900 health care has just massively outpaced inflation 00:30:17.900 --> 00:30:21.019 in the rest of the economy. So the CPIW, the 00:30:21.019 --> 00:30:23.980 index being used for the CLA, severely underweights 00:30:23.980 --> 00:30:26.480 the one category where seniors are spending most 00:30:26.480 --> 00:30:29.339 of their money and facing the highest price increases. 00:30:29.599 --> 00:30:31.880 Which means the CLA they get may not be adequately 00:30:31.880 --> 00:30:35.099 compensating them. for the real increase in their 00:30:35.099 --> 00:30:37.640 personal cost of living. Does the BLS know this 00:30:37.640 --> 00:30:40.279 is a problem? Oh, they do. They recognize this, 00:30:40.359 --> 00:30:43.720 and they actually track the CPIE and index specifically 00:30:43.720 --> 00:30:46.059 for the elderly. But it's not used for policy. 00:30:46.500 --> 00:30:49.440 It is not used. And the reason is purely about 00:30:49.440 --> 00:30:52.500 the budget. Our sources lay out the brutal economics 00:30:52.500 --> 00:30:54.799 of it. The Social Security Trust Fund currently 00:30:54.799 --> 00:30:58.599 has a projected solvency of about 40 years. If 00:30:58.599 --> 00:31:01.829 we were to switch to using the CPIE, which is 00:31:01.829 --> 00:31:04.490 more accurate for seniors and would provide higher 00:31:04.490 --> 00:31:07.329 annual COLA payments. It would drain the fund 00:31:07.329 --> 00:31:10.009 faster. It's estimated to shorten that solvency 00:31:10.009 --> 00:31:13.509 forecast by about five years. Wow. So a seemingly 00:31:13.509 --> 00:31:16.369 neutral statistical choice, which indexed to 00:31:16.369 --> 00:31:18.869 use it, has consequences that ripple through 00:31:18.869 --> 00:31:20.930 the entire federal budget. It literally changes 00:31:20.930 --> 00:31:23.069 the retirement age and tax burdens for future 00:31:23.069 --> 00:31:25.829 generations. The policy implications are just 00:31:25.829 --> 00:31:28.599 staggering. And this all came to a head and became 00:31:28.599 --> 00:31:31.619 a huge political flashpoint around 2010 during 00:31:31.619 --> 00:31:34.140 budget negotiations. And the debate was all about 00:31:34.140 --> 00:31:37.819 that chain CPI, the CCPIU. Right. And the proponents, 00:31:37.819 --> 00:31:39.779 people like former White House chief of staff 00:31:39.779 --> 00:31:42.339 Erskine Bowles, they argued really passionately 00:31:42.339 --> 00:31:45.119 that we should switch to the chain CPI for COLA 00:31:45.119 --> 00:31:47.390 adjustments. This argument was twofold. It's 00:31:47.390 --> 00:31:49.730 more accurate and it saves a ton of money. They 00:31:49.730 --> 00:31:52.509 argue the CCPIU is a more accurate measure of 00:31:52.509 --> 00:31:54.549 true inflation because it captures substitution. 00:31:54.970 --> 00:31:57.670 And switching to it was estimated to save the 00:31:57.670 --> 00:32:01.009 government over $290 billion over a decade. And 00:32:01.009 --> 00:32:04.289 that's because, as we said, the chain's CPI always 00:32:04.289 --> 00:32:07.029 runs a little bit lower than the standard CPI. 00:32:07.150 --> 00:32:09.890 Right. Historically, about a quarter to a third 00:32:09.890 --> 00:32:12.710 of a percentage point lower every year. So by 00:32:12.710 --> 00:32:15.769 adopting the more accurate index. The government 00:32:15.769 --> 00:32:18.029 would save hundreds of billions of dollars by 00:32:18.029 --> 00:32:20.670 automatically giving out slightly smaller annual 00:32:20.670 --> 00:32:23.809 increases to federal pensions and Social Security. 00:32:23.950 --> 00:32:26.450 That's the dynamic. Exactly. It was framed as 00:32:26.450 --> 00:32:29.589 a marriage of fiscal responsibility and statistical 00:32:29.589 --> 00:32:32.910 accuracy. But the critics, people like former 00:32:32.910 --> 00:32:35.269 Labor Secretary Robert Reich and groups representing 00:32:35.269 --> 00:32:38.470 retired federal employees, they pushed back hard. 00:32:38.670 --> 00:32:41.170 Very hard. What was their argument? Their argument 00:32:41.170 --> 00:32:45.180 was that. The CCPIU, while maybe it's better 00:32:45.180 --> 00:32:47.220 at capturing the substitution effect for things 00:32:47.220 --> 00:32:50.579 like groceries, still completely fails the elderly 00:32:50.579 --> 00:32:53.240 demographic because of that huge weight of health 00:32:53.240 --> 00:32:55.299 care. They said it doesn't solve the core problem. 00:32:55.500 --> 00:32:57.920 Exactly. They argued it still doesn't adequately 00:32:57.920 --> 00:33:00.339 address the fact that seniors spend 20 to 40 00:33:00.339 --> 00:33:02.099 percent of their income on these high inflation 00:33:02.099 --> 00:33:04.680 health care costs. They saw it as a backdoor 00:33:04.680 --> 00:33:06.940 way to cut benefits for the most vulnerable people. 00:33:07.119 --> 00:33:10.670 And in the end? In the end? While it was seriously 00:33:10.670 --> 00:33:13.089 considered during major deficit talks in 2013, 00:33:13.509 --> 00:33:16.410 the political blowback was just too intense and 00:33:16.410 --> 00:33:19.349 is not adopted. So we've got the CPIU, the W, 00:33:19.549 --> 00:33:23.009 the E, the CCPRU. With all this variation and 00:33:23.009 --> 00:33:25.589 controversy, where does the most powerful economic 00:33:25.589 --> 00:33:28.529 player of all, the Federal Reserve, look for 00:33:28.529 --> 00:33:31.490 its main signal on inflation? They use yet another 00:33:31.490 --> 00:33:34.049 index, the Personal Consumption Expenditures 00:33:34.049 --> 00:33:37.549 Price Index, or PCEPI. So why do they prefer 00:33:37.549 --> 00:33:41.190 that one? The Fed favors the PCEPI because it 00:33:41.190 --> 00:33:43.309 believes it addresses some of the CPI's biggest 00:33:43.309 --> 00:33:45.809 structural flaws. It's a broader measure that 00:33:45.809 --> 00:33:47.910 comes from the national accounts, not just household 00:33:47.910 --> 00:33:50.890 surveys. And crucially, it updates its weights 00:33:50.890 --> 00:33:53.490 more dynamically than the standard CPI -U. So 00:33:53.490 --> 00:33:55.329 it's better at handling that substitution effect. 00:33:55.630 --> 00:33:57.789 Much better. It's also seen as having a more 00:33:57.789 --> 00:33:59.970 comprehensive scope across the whole economy. 00:34:00.130 --> 00:34:02.049 And the statistical difference between the two, 00:34:02.109 --> 00:34:03.869 it might seem small, right? We're talking about 00:34:03.869 --> 00:34:06.349 maybe half a percentage point difference in the 00:34:06.349 --> 00:34:08.210 historical average. It seems small. but over 00:34:08.210 --> 00:34:10.369 the span of a generation, that difference is 00:34:10.369 --> 00:34:13.309 enormous. How so? Historically, inflation, as 00:34:13.309 --> 00:34:16.469 measured by the PCEPI, has averaged about 3 .3 00:34:16.469 --> 00:34:19.289 % a year. But CPI measured inflation has averaged 00:34:19.289 --> 00:34:22.269 3 .8%. So that 0 .5 percentage point difference, 00:34:22.409 --> 00:34:24.489 it just compounds year after year after year. 00:34:24.650 --> 00:34:27.389 It does. Imagine calculating a worker's real 00:34:27.389 --> 00:34:31.730 wage growth over 30 years using the 3 .8 % CPI 00:34:31.730 --> 00:34:35.449 benchmark versus the 3 .3 % PCEPI benchmark. 00:34:35.730 --> 00:34:38.750 The real value... of your wages, of your savings, 00:34:38.829 --> 00:34:41.579 of your debt. It all shifts massively. And the 00:34:41.579 --> 00:34:44.360 PCAPI showing lower inflation, that's a key reason 00:34:44.360 --> 00:34:47.000 why the Fed uses it for its 2 % long -term inflation 00:34:47.000 --> 00:34:49.820 target. It is. OK, so let's bring this complex 00:34:49.820 --> 00:34:52.039 statistical world crashing right into the real 00:34:52.039 --> 00:34:54.079 world with some recent headlines. Our sources 00:34:54.079 --> 00:34:57.300 track the US CPI data using 1982 as the base 00:34:57.300 --> 00:34:59.900 year, where the index is 100. By December of 00:34:59.900 --> 00:35:03.880 2021, that CPI reading hit 7%. And that 7 % figure, 00:35:04.000 --> 00:35:05.960 that was the highest level in over 40 years. 00:35:06.099 --> 00:35:08.179 It was the headline shockwave that really marked 00:35:08.179 --> 00:35:10.360 the beginning of a major shift in monetary policy. 00:35:10.639 --> 00:35:13.039 So even though the Fed prefers the PCEPI for 00:35:13.039 --> 00:35:15.800 its long -term thinking. The CPI is the metric 00:35:15.800 --> 00:35:18.820 the public sees and feels. And it was that really 00:35:18.820 --> 00:35:21.639 dramatic CPI signal that prompted the Federal 00:35:21.639 --> 00:35:24.780 Reserve under Jerome Powell to begin aggressive 00:35:24.780 --> 00:35:27.719 quantitative tightening and start signaling the 00:35:27.719 --> 00:35:29.820 rate hikes that were expected to kick off in 00:35:29.820 --> 00:35:33.889 March of 2022. So the CPI. For all of its flaws, 00:35:34.090 --> 00:35:35.969 the housing problems, the substitution bias, 00:35:36.170 --> 00:35:39.130 all of it, it's still the immediate vital signal 00:35:39.130 --> 00:35:41.929 that forces the central bank to act. It's the 00:35:41.929 --> 00:35:43.969 number that makes people feel the heat, even 00:35:43.969 --> 00:35:46.090 if the Fed is looking at its sibling index for 00:35:46.090 --> 00:35:49.170 the long term plan. Hashtag tag you tag outro. 00:35:49.719 --> 00:35:51.679 So if we try and connect this entire deep dive 00:35:51.679 --> 00:35:53.880 to the bigger picture, I think what we've really 00:35:53.880 --> 00:35:56.900 uncovered is that the Consumer Price Index, it's 00:35:56.900 --> 00:35:58.739 a statistic that's built on a whole series of 00:35:58.739 --> 00:36:01.619 really difficult philosophical and, frankly, 00:36:01.699 --> 00:36:04.280 political choices. We've really seen two main 00:36:04.280 --> 00:36:06.920 battlegrounds that define the whole number. The 00:36:06.920 --> 00:36:09.079 first one is that fundamental philosophical problem 00:36:09.079 --> 00:36:11.179 of how to measure owner -occupied housing. Right. 00:36:11.260 --> 00:36:13.440 Do you choose the subjective but maybe theoretically 00:36:13.440 --> 00:36:16.099 better opportunity cost method? Or do you choose 00:36:16.099 --> 00:36:18.599 the debt payments method that's unstable and 00:36:18.599 --> 00:36:21.710 creates that policy paradox? Or do you just choose 00:36:21.710 --> 00:36:24.150 a transaction price approach that messes up all 00:36:24.150 --> 00:36:27.150 your other economic statistics? Each path gives 00:36:27.150 --> 00:36:29.269 you a fundamentally different inflation number. 00:36:29.429 --> 00:36:32.210 And the second battleground is that political 00:36:32.210 --> 00:36:35.590 and social conflict over what's fair versus what's 00:36:35.590 --> 00:36:37.849 statistically rigorous. Right. You have the policy 00:36:37.849 --> 00:36:41.050 choice of which index is the fairest one to use 00:36:41.050 --> 00:36:44.389 for Coale. Is it the CPI -W, which is designed 00:36:44.389 --> 00:36:48.170 for workers, not retirees? Is it the CPI -E? 00:36:48.619 --> 00:36:50.599 which is demographically perfect for seniors 00:36:50.599 --> 00:36:53.079 but creates huge fiscal problems down the road. 00:36:53.239 --> 00:36:56.340 Or is it the chain CPI, which is statistically 00:36:56.340 --> 00:36:59.400 more robust but ends up resulting in lower benefit 00:36:59.400 --> 00:37:01.800 increases for people? The statistician is in 00:37:01.800 --> 00:37:04.659 this impossible position of being tasked with 00:37:04.659 --> 00:37:07.219 both scientific accuracy and managing political 00:37:07.219 --> 00:37:10.219 consequences. The CPI is never just about the 00:37:10.219 --> 00:37:13.760 math. It's a series of policy compromises disguised 00:37:13.760 --> 00:37:16.360 as a single number. So what does this all mean 00:37:16.360 --> 00:37:18.849 for you, the listener? Well, it means that the 00:37:18.849 --> 00:37:21.050 headline inflation figure you hear, whether it's 00:37:21.050 --> 00:37:23.769 2 % or 7%, it's never a perfect reflection of 00:37:23.769 --> 00:37:25.650 your personal experience. Right. If you're an 00:37:25.650 --> 00:37:27.750 elderly person who relies heavily on prescription 00:37:27.750 --> 00:37:30.530 drugs and hospital visits, your personal inflation 00:37:30.530 --> 00:37:33.469 rate is almost certainly much higher than the 00:37:33.469 --> 00:37:37.309 CPI -W suggests. But if you're a young, adaptable 00:37:37.309 --> 00:37:39.710 consumer who's really quick to substitute from 00:37:39.710 --> 00:37:42.369 one product to another when prices change? Your 00:37:42.369 --> 00:37:44.369 personal inflation rate might actually track 00:37:44.369 --> 00:37:47.679 much closer to that lower. chained CPI figure. 00:37:47.980 --> 00:37:50.639 Understanding the weights helps you figure out 00:37:50.639 --> 00:37:52.909 where you sit on that spectrum. And it really 00:37:52.909 --> 00:37:55.710 encourages you to see the power in these small 00:37:55.710 --> 00:37:58.409 statistical choices. That difference between 00:37:58.409 --> 00:38:00.989 the historical average inflation of the CPI at 00:38:00.989 --> 00:38:05.909 3 .8 % and the PCEPI at 3 .3%, that has massive 00:38:05.909 --> 00:38:07.969 cumulative implications for the real value of 00:38:07.969 --> 00:38:10.010 your savings, for the future of entitlement programs, 00:38:10.210 --> 00:38:12.489 and for the purchasing power of your wages over 00:38:12.489 --> 00:38:15.269 your entire lifetime. That half a percent can 00:38:15.269 --> 00:38:17.210 absolutely be the difference between a comfortable 00:38:17.210 --> 00:38:19.710 retirement and a really strained one. And the 00:38:19.710 --> 00:38:21.769 whole process is always one of catch -up. The 00:38:21.769 --> 00:38:23.309 statisticians are always looking at the past 00:38:23.309 --> 00:38:25.050 to try and calibrate the present. Which brings 00:38:25.050 --> 00:38:27.469 us to our final provocative thought for you to 00:38:27.469 --> 00:38:30.849 chew on. If the CPI basket is always being updated 00:38:30.849 --> 00:38:33.909 periodically every two years, every decade to 00:38:33.909 --> 00:38:36.869 reflect our current habits, what happens during 00:38:36.869 --> 00:38:39.809 periods of genuinely rapid, massive behavioral 00:38:39.809 --> 00:38:42.969 change? Think about the sudden explosion of digital 00:38:42.969 --> 00:38:45.710 services, the shift from buying CDs and DVDs 00:38:45.710 --> 00:38:47.789 to streaming, or the huge reliance on virtual 00:38:47.789 --> 00:38:50.409 everything during the pandemic. If the index 00:38:50.409 --> 00:38:52.599 is always - lagging behind measuring the prices 00:38:52.599 --> 00:38:55.059 of yesterday's consumption mix? Is it inherently 00:38:55.059 --> 00:38:57.380 incapable of capturing the true speed of modern 00:38:57.380 --> 00:39:00.039 inflation? Is the one number we rely on for our 00:39:00.039 --> 00:39:02.599 most crucial decisions always by its very nature 00:39:02.599 --> 00:39:04.519 a step behind the curve? We'll leave that with 00:39:04.519 --> 00:39:04.659 you.