'These Inflationary Impacts Were Persistent': Discussing Geographic Inequality in Food Inflation

Tom Heintjes: Hello, and welcome to another episode of the Economy Matters podcast. I'm Tom Heintjes, managing editor in the Atlanta Fed's Public Affairs Department, and today we're sitting down with Michael Navarrete, an assistant policy adviser and economist in the Atlanta Fed's Research Department. Today, we're going to discuss recent research by Michael and his coauthor, Seula Kim, on something we can all relate to: food inflation. Michael and Seula cowrote a paper in our Policy Hub series called "Geographic Inequality in Food Inflation." So, without further ado, let's say hi to Michael-thanks for being with us today, Michael. I've been looking forward to having you on the podcast and talking about your work.

Michael Navarrete: Hi, Tom. It's great to be here today on the Economy Matters podcast. Thank you so much for having me.

Heintjes: Michael, the cost of food is obviously something everybody can relate to-we all have to buy groceries. I'm always curious about what puts something on a researcher's radar. How did you come to explore this topic originally?

Navarrete: Research is a long process. My coauthor and I, Seula Kim, started this project back in 2021 when we were both graduate students at the University of Maryland. We were working together on another project with our professor using this data measuring national inflation, national sales, using this scanner-level data-which I'll go into more detail about later-and then we were wondering about all of this variation at a more disaggregated level. Instead of looking at national food inflation, national sales for food, what if we looked across different areas, specifically metropolitan statistical areas [MSAs], and was there a systematic variation? Specifically, we were curious: Do we see differences in food inflation between rich and poor areas?

Heintjes: So, literally comparing apples to apples.

Navarrete: Yes, we were very interested in food inflation, so yes-apples to apples, oranges to oranges.

Heintjes: Your work notes that food prices rose 1.8 percent a year overall during the period you studied-which was, I should note, from 2006 to 2020. But the increase is about 0.5 percent higher in poorer MSAs, amounting to an 8.8 percentage point increase. That's pretty significant.

Navarrete: Yes, it's a pretty large increase. Poor MSAs are facing much larger increases in food inflation. I'll go into detail, but part of this could be just differences in basket composition, which would be differences in the goods. Even though we're comparing food items across rich and poor MSAs, the actual items being purchased could be different between the rich MSAs and poor MSAs, and that explains some of the difference. However, it doesn't explain all the difference in the food inflation between the rich and the poor MSAs.

Heintjes: Michael, before we get much further: You're mentioning poorer and richer MSAs, but how do you determine which MSAs to examine? Also, how does your work define "poorer" and "richer"?

Navarrete: In order to determine richer and poorer MSAs, we're going to rely on BEA [US Bureau of Economic Analysis] income per capita data-so, data from the BEA-and then we're going to define the average income per capita of each of the MSAs in our sample, which is about 180, and then just sort them based on income per capita. So then, given that we're interested in making comparisons across deciles, the bottom decile of MSAs sorted by this income per capita are going to be the poorest MSAs, and then the MSAs with the highest average income per capita, according to BEA, are going to be the richest decile of MSAs.

Heintjes: Right-OK, great. Well, people tend to look at inflation, as you noted, on a national level-but there's a good bit of variance once you zoom in to the more granular level, isn't there?

Navarrete: There's a lot of variation once you go more granular, so exactly-this is how the project started off. We were just focusing on the national level, and then we were just curious: How much heterogeneity is there once we go more disaggregate, once we go from US food inflation to food inflation in MSAs such as Atlanta, New York, and obviously also much, much poorer MSAs as well?

Heintjes: I'm not an economist, so I try to stay in the shallow end as much as I can on these matters, but your paper discusses something that was interesting to me, the Herfindahl-Hirschman index, which I confess I was not familiar with before. Can you briefly discuss what the-we'll call it the "HHI"-measures, and the HHI's application to your research in this paper?

Navarrete: Yes. First, let me say why we're interested in this HHI measure.

Heintjes: Yes, thank you.

Navarrete: We find this stylized fact that you previously mentioned of higher food inflation in these poorer MSAs relative to these richer MSAs. One of the reasons that we conjecture why there is higher food inflation, beyond just the differences in basket composition, is due to differences in market concentration. Specifically, these poorer MSAs-they have higher market concentration. So potentially, when there are these cost shocks, these retailers are able to pass through a higher share of that cost to consumers, resulting in higher food inflation. What is this HHI measure? It's just going to be a measure going from 0 to 1, where you're going to sum the share of sales by the retailer. For example, let's just do the extreme cases, to illustrate. Let's say this MSA just has one retailer, so then their share is going to be 1. They have the entire share of the entire MSA, so then they're going to get a value of 1, which is going to correspond to a monopoly. But then if you had infinitely many retailers, and they all had an infinitesimal amount of market share, then their HHI would be very close to zero. In reality, you're going to have somewhere between 0 and 1, where a value closer to 1 is going to represent higher market concentration. So that's why we were interested in HHI-we wanted a precise measure of market concentration.

Heintjes: That's really interesting. Inflation is often measured in terms of what we call a "basket of goods." Can you describe this theoretical basket? Is it staples like bread, milk, eggs, cooking oil-items that everyone, rich or poor, purchases regularly? I guess I'm asking if this basket of goods in poor MSAs is the same as, or similar to, the basket in more affluent MSAs.

Navarrete: There are very large differences in the basket of goods between these rich and poor MSAs. In our baseline scenario, we don't want to put any additional restrictions. We want the basket of goods to best reflect what's being purchased by consumers. But if we use this basket of goods, it is going to result in larger differences in the baskets between the rich and the poor MSAs. And there's some systematic difference, where specifically the richer MSAs-not only do they have more retailers, they have more goods, so then the consumers have a larger variety of options to choose from. Then in the poorer MSAs, consumers have fewer retailers and also fewer goods to choose from. Typically, they're consuming a subset of the goods, relative to the richer MSAs. So, in our baseline scenario, we don't put any restriction on what the goods can be. The only restriction that we have is the "good" in our terminology has to be a continuing good-so, that good has to be purchased in the current quarter and the previous quarter. And just to give you an idea of how disaggregated we mean by "good," when you go to a grocery store and you buy a can of Campbell's soup, that has a barcode. So, when we say "good" we mean basically anything with a barcode-that's how disaggregated we mean. If you buy a different brand of Campbell's soup that's slightly larger, that's a different good, just to give you some idea of the disaggregated nature. We then also do a version that we like to call the "common goods rule." Then we have ten deciles of MSAs, sorted by income per capita, as we discussed previously, and then in order to make it more of an apples-to-apples comparison, or orange-to-orange comparison, we restricted the exact same goods. So then, the basket of goods is going to be the same in the richest decile of MSAs relative to the poorest decile, and we still see this pattern of higher food inflation in the poor MSAs relative to the richer MSAs. However, it's worth noting that once we apply this common goods rule, this attenuates the difference. So that 8.8 percentage point increase in the poorest MSA relative to the richest MSA after our entire sample, dramatically falls, but it's still there.

Heintjes: Interesting. But the tools you used allowed you to look at different types of food beyond just staples-right?

Navarrete: Exactly right. We have a very rich data set. We were using the NielsenIQ Retail Scanner data set, which covers millions of UPCs [universal product codes]. So again, those things that you scan at the grocery store, that's the granularity of data that we have. We have everything from the Campbell's canned soup, to beef-different types of beef. We have very, very granular data.

Heintjes: Were you able to know anything about the buying habits of richer MSAs versus poorer ones?

Navarrete: Yes. I would say the for sure difference is just the sheer variety of products available between rich and poor MSAs. The poorer MSAs are consuming fewer products, but part of that is mechanical-because the poor MSAs offer fewer products to consumers. There are fewer retailers, and even in the same retailers there are fewer products-even within the same retailer across rich and poor MSAs. And then what the actual households in these different MSAs are purchasing are different. So then in the richer MSAs, these households may be purchasing higher-priced items as well.

Heintjes: Michael, you know in your paper that correlation does not mean causality-which is always a good thing to keep in mind, and economists often make that noted-and you note that poorer MSAs might have different preferences or price sensitivities. I wonder how you've gathered the data that you use in your analysis, because there are so many MSAs.

Navarrete: Yes, there are so many MSAs. We end up having a sample of about 180 MSAs. However, this isn't all of the MSAs in the US. We ended up choosing all of the MSAs available to us, given the data that we had-we relied on the NielsenIQ Retail Scanner data set. So, every MSA that was in that data set, we used-but this doesn't include all the MSAs. Actually, if anything, the MSAs that are least likely to be in the sample are going to be the smaller, especially poorer, MSAs. So, this is just a subsample of all the MSAs in the US, but a very large subsample.

Heintjes: Well, you noted you had the avian flu outbreak, which caused a spike in egg prices. What did you observe among richer and poorer MSAs when it came to egg prices?

Navarrete: Yes, we saw that. So, the richer MSAs, again, have this pattern of having a lower market concentration, and these poor MSAs have this pattern of having higher market concentration. So then, among these poor and rich MSAs that are affected by the bird flu, we saw that these poor MSAs with higher market concentration experienced higher inflation in the inflationary period of the bird flu-but actually during the deflationary period, they weren't experiencing higher deflation. So even though the bird flu was a temporary shock, where the egg prices surged up and then went back down, these inflationary impacts were persistent, given that we didn't see a higher deflationary period for these poor MSAs with higher market concentration.

Heintjes: That's really interesting. You know, it was my privilege to be your editor on this paper, and as I worked on it, I kept expecting to encounter the phrase "food desert"-which, you know, we hear commonly. People talk about food deserts. In fact, I was even tempted to suggest that you use it, since it's a concept everyone is familiar with, but it didn't occur once in your paper. You talk about, as you've mentioned, retailer concentration. In the context of your analysis, is "food desert" a misleading or inaccurate term?

Navarrete: I would say that we, as you noted, prefer "market concentration," given that it allows us this precise measure where we can directly quantify what would be more concentrated. We can go from anywhere between 0 and 1, in its entire gradient.

Heintjes: I guess that gets back to the HHI you were talking about.

Navarrete: Exactly. That's why we really prefer talking about market concentration with HHI. And then "food desert"-I think a lot of these poor MSAs could potentially fall into the definition of food desert. However, food deserts might also bring to mind rural areas where you have potentially just one retailer for a very wide range. I would like to note, though, that our paper exclusively focuses on MSAs-so by definition, these are all urban areas.

Heintjes: And that's a good thing to keep in mind. As we noted earlier, this paper covers the period of 2006 to 2020, and of course we all recall a certain event from 2020 that caused a great deal of disruption and high inflation in its wake. And I'm not asking you to speculate-well, I guess I am. But do you have any thoughts on what the differences in food inflation between poor and richer MSAs might look like since 2020-maybe "more of the same but on steroids," or anything like that?

Navarrete: So, I'm not one to speculate.

Heintjes: I know that-hypothetical.

Navarrete: Yes. As you noted, the paper ends in 2020, but if you were to twist my arm, I would guess probably more of the same-where you see we had larger cost shocks that happened after 2020 with the surge in food inflation, I would expect to see higher food inflation during these cost shocks in these poor MSAs relative to these richer MSAs. However, I would say that's beyond the scope of our paper.

Heintjes: Understood. Maybe we can look for a sequel.

Navarrete: Yes. My co-author, Seula Kim, and I are working on a new paper-not maybe in the way that you think. We're not going beyond 2020 looking at the new inflationary periods, but going back to our earlier conversation of all this granularity, we're thinking of actually going even more disaggregated. So even though we see that these richer MSAs are experiencing lower food inflation, potentially there's a lot of heterogeneity within these MSAs-particularly focusing on the largest, richest MSAs. Potentially, these rich MSAs-let's just take New York for an example. Maybe there's a lot of heterogeneity, and actually, let's say we go to the county level. Maybe the richest county is actually driving inflation to be much lower, where you could actually have the majority of the residents in that MSA might be facing higher food inflation given that these price indices that are used to calculate inflation, they're expenditure weighted. And then these rich counties, they have a disproportionate share of expenditures.

Heintjes: Michael, I wonder if the conclusions you reached were surprising to you in any way. Did anything stand out to you with something you maybe didn't expect?

Navarrete: The thing that honestly surprised me the most was just the stylized facts, the descriptive facts that we led the conversation with: higher food inflation in the poorer MSAs relative to the richer MSAs. That really jumped out to me in the beginning, given that, to the best of my knowledge, previous work had focused on differences across households, but there hadn't been differences across regions. Seeing that stark difference and seeing it repeatedly-that was a bit surprising to me.

Heintjes: You know, the Fed is a policymaking institution, so I always like to ask: What would you like policymakers to take away from reading your research?

Navarrete: I would like them to take away two key points. One, which we just discussed: Poor MSAs are experiencing higher food inflation, at least from 2006 to 2020. And second, one of the reasons why these poor MSAs are experiencing higher food inflation, where poor MSAs have higher market concentration-so when these retailers experience a negative cost shock, they're better able to pass it through to consumers.

Heintjes: Michael, I want to thank you for spending time with us today. This has been a really fascinating conversation.

Navarrete: It was a pleasure to be here. Thank you so much for having me, Tom.

Heintjes: And I want to note again that we have a link to Michael and Seula's paper on our website-our newly designed website, I should add-at atlantafed.org. Thank you for giving us some of your time today, and let's get together again next month.