 Law and Society, he gave a much briefer presentation, based on this paper at the PAA meetings, and it was so interesting that I invited him to come and give the full-bore version today. With that, I will hand it over to John. Thank you. Thanks, Susan. Thank you. That was very nice. But full-bore, I guess, didn't want to talk like a New Yorker. I took two breaths, I think, during the presentation at PAA. Segregation through the lens of housing transition is the talk that I'm giving here. Most people know that I study neighborhoods, how neighborhoods change over time. A lot of my work focuses on crime as one outcome. That's just one way we can measure the quality of the neighborhood, what it's like, is it good or bad? If I live in a neighborhood with little crime, that's better than having a lot of crime. But my research, because I'm interested in how neighborhoods change, a fundamental process of how neighborhoods change, of course, is people living in them, and people who move in and move out of neighborhoods. So unnaturally, part of my research gets pulled in this direction of residential mobility decisions. Some of my work has looked at segregation because that ends up playing into, that's kind of an outcome, one more outcome of the process. Some of my work has looked at how mobility decisions are affected by crime as well. This one is, this particular paper working on here is focused on residential mobility decisions, mostly looking at moving into the neighborhood. So again, if you think of segregation, segregation can only happen basically when at the level of the housing unit, when we see a transition, it's the only way you're gonna get to a segregated outcome. It takes over time, it takes a process, but that's fundamentally what it takes to get there. So for my interest here, I'm interested in focusing on housing units. I'm gonna be following the data set. I'm gonna be using American Housing Survey, follows housing units over time, so it's a little different than other studies using following longitudinal samples of people who they're following where the people move, which is interesting, but it's a different question. So here it's a focus on people moving into the unit. So the question, how do people choose what unit to move into? What is the process? Do people care about the racial ethnic composition? I'm not gonna be focused here on trying to get at the perceptual sorts of things as a whole literature on that. I'm more just looking at some of the patterns in what we get from it. There's some interesting implications I think from some of those patterns that I'll talk about in a bit that some of the perceptual information out there focused on a very small unit of analysis. I'm actually gonna be using that small unit of analysis and showing how it captures part of the process, but not perhaps all of it. So when people make a decision where to move, what unit to move into, what are they looking at? If racial ethnic composition matters, that's a question we could even back up and ask, but if they're looking at that, what area are they looking at? Are they looking at some broader neighborhood? Of course that's a mushy concept, that's something I work with a lot and some of my work trying to get at what is a neighborhood? How do we define what it is? But if you set that aside for a moment, say well there's some sense of a broader area which is a neighborhood, is that what people look at? They look at the racial ethnic composition of that area and let that guide their decision. Or is it something much smaller? I refer to the micro neighborhood composition, this term is mushy itself as well, but nevermind. More and more this idea is kind of popping up. Sometimes you see it referred to as the street block where Granis has used that term a lot. The idea basically you can think of a street segment both sides of the street, where bounded by intersections at either end, kind of a small area in the suburbs, this is often what we think of as a block party, you see this real cohesiveness. Granis has done some work recently looking at the formation of what are referred to as neighborhood ties, neighborly ties, which is a vague concept in and of itself, but nevermind. But there you see this real strong geographic boundary where people are much more likely to form ties with people next door to them across the street and then it decays really quickly. So again that suggests that this local street really matters. And so maybe we would think when people make a decision on where to move to, that they're looking at the racial ethnic composition of the street. Or is it, do they look at the current residents, the racial ethnic composition of the household that they're moving into? This is something we're looking at here. I've not seen much in the literature looking at this, but it's something we're gonna consider as well in this paper here. How does segregation come about? This group here, I don't think I spent a lot of time I would imagine on this, but there's different theories out in the literature that have suggested how we get segregation coming about. Here's a few that I'm gonna talk, I'll just kind of go through briefly. And in particular I wanna say, well what implications would they have for these different units of analysis? Whether we're talking about whether it's the racial ethnic composition of the household or the micro neighborhood or the broader neighborhood that matter. There's the place stratification theory, the household preferences theory, which is this idea of people having a preference what the racial ethnic composition of the people that live near them are. Place stratification is the notion more that race matters even above and beyond the economic resources we have. Differential economic resources is more the classic assimilation model idea that as people gain economic resources you'd see less segregation and whatever you do see is just driven by this economic process. Discrimination and steering are the notion that you'd see access to certain units being denied or steered away from, if you will. Social networks, the idea that there's this networking process that goes on that somehow contacts are important for bringing about information about units. And then the last one that I'm gonna talk about a bit here that I'm gonna suggest at least throw out there as a possible part of the thing to think about is signaling, an idea from economics, I'll get to that in a bit. So again, talking about each of these, what do they mean for these processes? Place stratification theory, if it is indeed the case that this process works, what we really should see arguably is that the racial ethnic composition of the broader neighborhood is what you would see mattering. This is much more of a larger structural theory. It doesn't focus so much on micro areas, micro neighborhoods and that kind of thing. As much. Household preferences on the other hand is a theory that gets a little more micro as you might well imagine. Again, it's this idea of preferences. So preferences that are driving what the area looks like. So you might well think that the micro composition of both the local area and the broader neighborhood would matter for someone's preferences. They wouldn't just look at their local block and then ignore the broader area. According to this theory, it'd be less reason perhaps to think that the household itself composition or the characteristics of the previous household would matter to them. But that's what we'd expect from that. This notion of differential economic resources to classic assimilation model concepts. Well one, if arguably if we account for economic resources we shouldn't really see much other effect for race ethnicity. So it's of less interest to us here. Discrimination and steering a bit trickier to think about. What are the consequences? Discrimination, it depends how you model this. What do you think how the process works? If it works at the level of the housing unit then that's where we would see it. We would see this process where there'd be reduced likelihood that a housing unit would transition from a white household to a non-white household. So you wouldn't necessarily see it at the broader neighborhood or even the micro neighborhood. But it would happen right at the level of the housing unit. Steering, which is this notion of real estate agents playing this important role of guiding prospective residents to certain housing units, certain neighborhoods. What might matter there? Well you might think that the composition of the micro neighborhood or the broader neighborhood would matter. Although at the, originally I wrote this as the more I've thought about it, could you make an argument that steering might happen based on the race ethnicity of the previous, or the current occupants of the unit? Something that might occur, I don't know. It would require the real estate agent knowing the race ethnicity of the previous people would they use that as a cue of who to show the unit to. If that were the case, you could see that process occur as well. Social networks, social networking, the idea of who people talk to, how they get information about different units, different neighborhoods. How would we see this process play out? You could think about it in a couple different ways. One, we might see that the composition of the micro neighborhood might matter, or we might see that the housing unit matters. The housing unit, you can imagine this process if people find out the availability of a unit. Someone's moving out of a unit, they send information out through their networks, we see this differential likelihood of networks being members of the same race, but then you could see that information going to people of the same race ethnicity, and it could play out that way. John, are you talking about both turnover and rental units and owner occupied, or? Yes, I'm mushing them together, that's a great point. But we can talk about that. So it seems to be about the different processes. Exactly, yeah, that people raised that, posed that, yeah. What are you thinking of, in any particular? Well, the flow of information about finding out about units, and both the process by which the next occupant is determined, is it a real estate agent who's selling it, or a rental agent who's trying to fill it? So I think that there might be differences here. Yeah. For the analysis here, this initial analysis was mushing them together, if you will. I was pushing that direction of trying them. Some of the models I tried differently didn't show too many difference, not as many as some people expected to see, if you're not seeing, but it, there's definitely a different process that goes on there, and we'll get to that in just a moment when I talk about this idea of signaling. When we start thinking about that, how that plays out in a rental, or an owner situation might be different. So it's definitely worth, yeah, to the extent that that plays out differently, that's an important empirical distinction to make. And so the idea of the micro neighborhood is this idea of information you find out about others nearby, that could play out perhaps in a rental building. You know, you see a unit become available, you send out information to your network ties, and again, if they're more differentially, you're the same race ethnicity, you can see that process play out that way. And likewise, on the micro neighborhood, if you know a unit is available. The final idea that I'm kind of going through that I'm gonna kind of put out there is an idea to think about is this notion of signaling. It comes from economics, and at root, it's this idea of information asymmetry. It's been used, initially it was proposed for labor markets, those who are familiar with the idea, this, the classic case, the principal agent problem, where this person coming in wanting to get employment, knows how, I know how good a worker I am, but the person hiring me doesn't know that about me. They're trying to interview me, they're trying to figure out something about me. So they're looking for signals from me. What kind of a worker am I? And the early literature, the one idea was this idea of getting a degree, right? And a degree of some sort, economic, oh, I can't, I can't, I can't, I can't, I can't. Brody, you sit there. Can't I forbid? Can't I forbid? I have one, I'm happy. So you get this degree and it's a sign of certain quality that you have. And that sends a signal to the potential employer. So again, it's this information, a symmetry's popped up in other literature as well, having to do, basically any time you can see this process come up, they've done it in the investment literature, it's popped up a bit there. One implication of it that then was raised early on by economists, this notion that you can get this sort of statistical discrimination, they pointed out, if there's a tendency for one group to be poorer workers, well then if I see someone as a poorer worker, I can statistically infer that they're more likely to be a poorer worker given their characteristics. So that is a potential problem that can pop up with that. I'm suggesting the same situation in a sense can occur here, that the person living in the housing unit knows more about it than I do as a potential person moving in. I'm trying to figure that out, right? It's like buying a car, right? You're trying to figure out what kind of shape is new. And as well, they know more about the neighborhood than me as the perspective person moving in. And again, I'm kind of a detective when trying to figure that out. A lot of my work focuses on crime and there you can see the same potential thing becoming an issue. Someone who lives in the neighborhood knows, they may have seen things happening, it seems like the neighborhood's going downhill. Me as someone coming to consider the unit has less information about that. I'm trying to pick that up. I'm looking for signals, right, if you will. Maybe I'm looking for disorder or that sort of thing if I'm interested in being concerned about crime. If for some reason I have a particular preferences about race, ethnicity, would I use the race, ethnicity of the current household as a potential signal for something like that? What it would suggest to me, might suggest that, A, this is an indication of the race, ethnicity of the area, perhaps. It's a crude estimate, but other than me going door to door walking around and trying to see what the race, ethnicity is, that people might be one simple clue. Or another idea is that even if it's a racially, ethnically mixed neighborhood, this might be a sign to me that this is a welcome area for someone of my race, ethnicity, if that was something. For whatever reason I was concerned about. Because you just mentioned a minute that I was interested in some of the earlier versions where if I have a preference for living in a neighborhood that has people of this race and not people of that race, how do those theoretical approaches deal with that the information problem? How do I find out what the races of the people in the neighborhood are? I can't go to a door, they're not necessarily standing out on their porch steps. How are you supposed to, clearly if I lived in the neighborhood long enough I would eventually see people leave their houses unless they live in a suburb and they'll leave or see their cars. But how am I supposed to find out about this as a potential home buyer? I can't ask. I mean that's an excellent question. I don't know, are there people in here who know that have a sense of that? It's great I can. Years ago they did the National Survey of Black Americans and they were especially interested in trying to find Black families in largely Black neighborhoods. And it was an incredibly reliable process to ask any White resident. There were any Black residents in the neighborhood. Yeah. So that's the kind of search you're interested in. It's not that difficult once you start to just ask a few questions. Okay I guess that's the question is like as a potential home buyer is that something that people feel is politically okay to ask somebody or does that sound like a racist thing to ask? So are there any Black people in this neighborhood? I don't know. Anything you can imagine asking somebody that? Right. Yes. I'm sure. That's why you get a real estate agent, right? And then a real estate agent finds those things out. Well the real estate agent knows the neighborhoods, right? If they do, yeah. Well that's their business to assess. But it's illegal for them to tell you. Right, but they can show you certain houses and they can say, okay go. That it violates the Fair Housing Act because it's steering if you start talking about the racial composition or the crime level. But they can talk about the school district and the other amenities in the neighborhood. Okay. That's what I can only touch on. I don't know if you're a sociologist or you're not. I mean you touch, I mean that instrument, the crime is interesting because there's a little side note. I said crime in the neighborhood that I'm trying to collect data. And a lot of cities, some cities will provide it on the web, others won't and I think the concern is just that, that people will use that as a cue for inferring something else from it. But it's a tricky question at how you would determine that. Yeah, a real estate agent presumably would know that but their information would be imperfect I suppose. So then the question is, there might even be a question. Some have posed, I've said that, given this talk before and some of them immediately shot back, no way do people know the race ethnicity of previous residents, which I don't, I don't, I've heard it was such a sturdiness that it kind of, oh God, maybe I don't know something here. How could people know it or how could they infer it? One, you might meet them, it's very possible that you meet the previous residents. Two, they might be described by the real estate agent in some way. Three, I pose this sort of joking, new residents may be cultural anthropologists. I say half joking, but maybe not, right? We don't, we may be better than we realize and there's interesting evidence about this, but if you go walking through a house, can you tell the race ethnicity of the people living there? Susan mentioned it, what's this little museum up in San Francisco? Oh, it's called the Exploratorium. It's a kid's science museum essentially, but one of the exhibits there gives you pictures of a house and it says, can you guess who lives here? And they'll give you some of the Latina woman, they'll give you say an Asian couple or pictures of people who are very different and then they'll show you the house and you have to guess by the clues and one will have a picture of Milagros on it so let's see who might live there. Or another will have pictures on the refrigerator that are of an Asian family. Well, let's see who might live there. And you could very easily guess who lived in one kind of house. And so it's an exhibit in the museum. How common is staging, I mean for home sales? That's a key question, that's a key question. And on the Bay Area, you don't sell a house without having it staged. Yeah, and describe it, what is staging? So basically you have a professional come in, take all your stuff out and bring in stuff to make your house look appealing. Better for you. And it's appealing to sort of a certain kind of, I mean, so there again, it's a certain kind of buyer. It's still a cultural, yeah, yeah, yeah. But it's sending a certain kind of signal, right? And they say, you know, yeah, my understanding is they take all the pictures, all the family pictures out and yeah. I mean, they might, you know, if they're really good, they might bring in pictures of the prospective buyers' families and put them up. Right, right, right. How common is that up there? Do you know? It's pretty much universal. Yeah, I'm trying to get, this data set I'm using in his back would be the 80s into the very early 90s, do you suppose that was going on back then, you know? That's something that kind of need to get a sense of to what extent, if that is indeed prominent now, that would pose a bit of a problem for this, but. I know the rental market was tighter in the early 90s, but I don't know about the buying market. Yeah, okay. So again, how might we do this? Different ways, photos in the wall, unless it's been in a professional stage. Magazines line about, right? Magazines are an interesting cultural reference. And the type of magazines that people read can give you a clue. The art, music that's sitting around, decorations that are up and about. Empirical evidence, so this is a book back in the early 80s. So again, using late 1970s data found that 96% of respondents knew the race, ethnicity of the previous residents. So we're pretty strong empirical evidence. Now, does that still hold now? That's something, you know, now we're. That's after the fact, right? That's after the fact. Yeah, after the fact. Yeah, whether it holds now is another question as well. I'll just briefly talk about previous work that's focused on the same idea of housing unit transition. Marulo did a study back, again, using the same data set as what I'm gonna be using here, showing transitions in housing units, but didn't take into account the context at all. So he was just focusing this narrow question, do white people tend to replace other white people, Asians replace Asians, et cetera, and found really strong homophilia. But again, not taking into account the context. Rosenbaum and colleagues using New York City data done a number of studies, but there they've done the housing unit, but then they've also taken into account a context. But for them, there are neighborhoods, if you will, as they term them, are 100,000 people or more, which you can argue is a pretty big neighborhood. Regardless of how dense New York City is, it's still pretty darn big. Ellen used census tracts as the context and found this effect. Again, the racial ethnic composition of the tract, effect of who moved in, and at the same time, she still found the same household effect, even controlling for the composition of the tract, there was still this effect where people of the same race tended to replace one another. In there, she speculated that she wasn't sure if this was entirely a household thing, it could be the micro neighborhood as well as driving that. And with my data, that's when I'm gonna be able to tease out, we're gonna be able to take into account characteristics in the micro neighborhood. Sam Friedman has a forthcoming paper looking at micro neighborhoods and tracts, and then studies of racial ethnic preferences, it's interesting to point out that they've shown these cards to people and the cards are basically a micro neighborhood. And that's sort of, and it's interesting and then juxtaposed out with the segregation literature that uses tracts. So certainly the preferences have focused on a real small unit, so I'm gonna try to measure that here. The data, again, American Housing Survey, the neighborhood sub-sample, this was drawn in three years, 1985, 89, 93. They did this extra little sample, I didn't even put the numbers on there, did I? I have to do this, I've done this so many times, there's many others, over 670 micro neighborhoods, I think in 85, and then they supplemented that in 89 and 93 with new micro neighborhoods that came along. So I've got two move periods, I can study people moving from 85 to 89, do they move, 89 to 93. And movement being just does the head of the household change over that four year time period, so I don't know, I mean just at the beginning of the end, I don't know how many moved or anything like that. And so I'm gonna have three outcome measures of people that I'm looking at, basically looking at the race ethnicity, the new residents in the unit, are they white, are they Latino, are they African-American? Very clear for you, you said that the head of household changed, could the head of household only have moved out and the rest of the household stays the same? I don't know. I mean, they get divorced in a few weeks and the rest of the family's still there. That's not really a household unit change, but it would be a change in the head of household. Well that's a tricky one, yeah. If they were to a family person to point that out, you're right, that's an excellent point, but yeah, I don't know how that happened, I agree. Let's check if that gets caught or not in that case. I'm banking usage, I'm all moving. I got a check on that, that could be, not quite right. Most of the time. So again, focusing only on households who move in these over the two years, 5,700 moves we have. I don't know if I, I think I do mention, here we do, they do some models accounting for the non-movers and they don't change their results at all. Our independent variables, we're most interested at the household level, the race ethnicity of persons, white, black, Latino, and other. The other category is pretty small. Again, this is the mid-80s. This year there's just not enough Asians or other groups to look at them at all meaningfully. Micro-neighborhood and tract in each of those, we include the percent white, black, Latino, and other. In tracts we can split out Asians because there's enough variability to do they're not so in the micro-neighborhoods. We also include a measure of the racial ethnic mixing, the heterogeneity, based on herping dog index. How do you operationalize micro-neighborhood? Micro-neighborhood, yeah, good point. I ran through that. This, this American Housing Survey follows housing units over time and what they did in 1985, they came along, they randomly pulled about selected 670 of these units and then what they did is for each of those units they went and they interviewed the 10 closest units to them. That's what they have. So you have this center point and then the 10 closest. I've actually read through this documentation exactly how they do it. It's based, they don't go over a railroad track but it's whoever's closest, they go around the block sort of thing. If it's in an apartment structure or come back to there, they'll, there's a real complicated thing but they'll do all the people in the same floor. There's 20 on the same floor then you'll have 20 people in the micro-neighborhood. If it's less, if there's five, then they'll go down. I think they go downstairs and upstairs. I've been asked about your upstairs and downstairs neighbors because those are the ones you meet first, right? If you're here first. Or here first. Yeah, if you're there by the stairs then I'm gonna test it here. So that's pretty much exactly what you want. In a sense, yeah. It's pretty darn close, yeah. Amazing. And it's, the data's not been used very much. It's something, I stumbled along, I grabbed it for my dissertations when I started working on using. I did the analyses at, the US Census has these Census data centers. And I did the work there. Little bunker or no sunlight or anything like that. Other controls taking account income stability, crowded and perceived crime. Other neighborhood characteristics, school quality, bars and liquor stores. That kind of thing just to account for, maybe these are sort of spurious things that are driving it. Not many of them have much of it. Interesting fact that we did that. The models are just logit models with corrected standard errors for this clustering. I did do multi-level logit models but the results were similar. There's some estimation issues with the logit with SAS. It's a whole, it's ugly mess where you have to use starting values. Results can be different and the whole thing. And we're limited in the census data center. All you have is SAS there. So the results were the same. So I'm just gonna present those results. Multiple imputation for missing data. And as I kind of alluded, we did do a HECN selection model including the inverse mills ratio to taking into account nonmovers. The results were very similar, extremely, extremely similar. So it didn't have much of an effect. Let me show you a few of the results. What we found, first question again, asking is the new household black? It's moving into the unit. Talking about predicted probabilities. So here, again, the question if the black household replaces the white household, this previous unit had a white household. Now what is the probability that the new household will be black is what we're asking here. And for this, to look at the effects in these different micro-neighborhoods, we start picking the average micro-neighborhood racial composition for black residents in the sample entirely. So that's the, we start with that as the baseline, if you will. And there the predicted probability is 0.48 that a black household will replace a white household that was living in that micro-neighborhood. Now if you increase the standard, the percent white in that local micro-neighborhood, one standard deviation, that probability drops to 0.23. If you increase the percent Latino, a similar amount, it drops to 0.32 so it's not dropping quite as much. And conversely, yeah, go ahead, Greg. So this is, this can't be of all the white, original white households, what fraction switch? No. That's gonna be an incredibly small number, right? So. Yeah, of all the white, right. That would be a different thing. Yeah, do it like that, yeah. What is, what is this? Like, I don't understand that. What's the 0.48, for example? That would be, that's, that is gonna be the predicted probability. If we had a white household living in an average, a micro-neighborhood that has the racial composition of the average black, there's not, for one thing you imagine, there aren't a lot of white houses there, just that, yeah, off the top of your head, you say that's kind of surprising the big, somewhat of a big number. Why would it be almost a 50% chance that it would switch to black? That would be in a sense be a bit of an indicator of this sort of process. We see that white households living around more blacks are more likely to leave, and then more likely to be replaced by blacks. So that's what is, it's that white household living in that context. Then what is the probability of the new residents who are black? The average is about what, 12%? 13, 15%? Say that what? The average is 15%. What is that value? The average micro-neighborhood. Oh, what is the racial composition of that? Right. What's the, so what you're saying is like, if this is like 15% black, this neighborhood, and then they have one. No, it's much more than that because it's the average micro-neighborhood of a black resident in the sample. So that's, it's not the average micro-neighborhood, yeah. It's the average context of a black household. Oh, okay. So put that in context. Yeah. So it's a bit, yeah, you could write how to present the results. You could, it's. And then the point A that would look really high, right? Yeah. Right. And that. So you can, as you can well imagine, the numbers are gonna vary. It's particularly gonna vary when I get to tracks because we've got this, it turns out there is this heterogeneity effect, particularly for blacks. We don't see it as much for the other groups, but we do see it for them. And then that, so then it's gonna matter whether you're at a low percentage or a higher percentage is gonna make a difference. So it's not a straight linear effect, basically is what we get. And at the micro neighborhood level, again, if it's more a higher percentage black than the idea, then the probability of that transitioning from a white to a black household is even that much greater. And what would that, those numbers be if, or maybe I'm not just saying where you're going, but if that was the black household that was being replaced. Oh, a black replaced in another black household? Yeah, in the average black. I should just show you these. Don't show that or not. I don't think you can see this, maybe. Maybe so. Here's, yeah, here, the number is 0.87. Okay. Yeah. If it's, so it's 0.48 chance if it's white, 0.37 if it was Latino, 0.87 if it was black for that. And so I don't know if it's easy to follow this or the other one which you can see better, but so that's this micro neighborhood context we see this kind of straightforward effect. For tracks, we actually see a little effect for from in movement of a black household. And again, it's because of this hydrogenated effect that where you are, here we're starting with a track that's relatively high percentage black. So when you push it up higher, it doesn't have much of an effect at all. Probability is still 0.48 of course, but when you increase percent white, there's hardly a change, Latino hardly a change and percent black going up. It's even less of a likelihood. So blacks tend to be being pushed more into a heterogeneous. Are standard deviation changes comparable at those two levels? No, they're a little different. They're a little different. I went with, I chose to go with standard deviation change because the argument being that this represents really the amount of change because as you're intuiting, there's always gonna be more, I think I can say always, nearly always gonna be more variability at the micro neighborhood than at the track. This track's a larger agitation so it's not gonna move as much. So that makes a little bit hard to think about 0.46, right? Either way, yeah, I don't, it's a tough question to back up and say, well what number do we wanna get it? Do we want it? Because a 10% change in each are different things. So I don't know if that's comparable. So I went with this. Either way, you're gonna, if you push it more, it's still, when it's not gonna make a big change, it's still gonna be. If we double that, it's gonna push it down even a little bit more. So it won't make that much of a change. Again, this is the only one where we found this effect. Now, controlling for that neighborhood, micro neighborhood, do we see the same race effect? Yes, and it's actually extremely strong, even controlling for these. Here, just presenting these as odds ratios. Compared to previous house will be in black, the odds are reduced. Of this new house will be in black, 86% of the previous were white and 91% of the previous household was Latino. So it's still a gigantic effect controlling for the track, controlling for the micro neighborhood. Still something going on about the race, ethnicity of the housing unit. And the pattern of results, very similar in general, for the other two racial ethnic compositions people will be in. Here, asking this new household white and here, micro neighborhood, a white household replacing the Latino. I just chose this one rather than going to that whole table I had before, although I've got that for people who want to come through that. And the average micro neighborhood of 0.31, chance that a white household replaced a Latino. If you increase the percent black or Latino, that probability goes down quite a bit. If you increase the percent white, it goes up. So again, a fairly substantial micro neighborhood effect. But here we also see a stronger effect for tracks, for white. So at the same time, controlling for them at the same time going on, the average is 0.31. Again, increasing the percent black or Latino goes down even more. And again, this is a little less of a percentage change. So it would be even more if we did an equal percent change. And then whites going up even more. So that's enormously more important, right? The track level than micro neighborhoods or whites and blacks? Yeah, or whites and blacks, absolutely. Blacks that had almost no effect, depends, yeah, where you're at. You could see a bit of an effect if it was a low percentage black. But for whites, yeah, it's a huge effect. So here would be interesting to look at home ownership, right, versus renting? Because if you're interested in house value, you might be interested in a larger neighborhood. Larger area. Yeah, it's important if you see blocks changing. That's an excellent point. That's an excellent point. Out to it. Looking at that one. Nonetheless, even controlling for it, there's huge effects for tracking micro neighborhoods. Still a huge effect for the housing unit, raceiveness. Again, it's 80%, odds are reduced 80% if the prior household was black or Latino. So finally, looking at it's new household Latino, similar pattern for blacks. Same story that it goes down if you increase the percent white or black and the micro neighborhood goes up, increase the percent Latino. So again, same race effect. Tracks were similar, similar kind of pattern going on there. So you can plot out this reinforcing effect to be increased to percent, one standard deviation increase in percent white in the micro neighborhood and the tract. This probability of 0.25 goes down to 0.06. So it's a big drop. Same way with percent black, drops quite a bit. So reinforcing, even though it's just a linear or an additive effect, but it's still strong when you combine the two. But again, this strong household level effect. Do you think a fraction of people moving in know the people they're replacing? Has anyone asked about that? I don't know. That's when people oppose that. If people know of literature, I've been scouring the literature trying to find this sort of a vast people about this. I've not found much. I've found that, I found a couple sites that talked about out to the fact that people know the race ethnicity of people, but whether they've met them or not, I don't know. Is that my account for some of the huge effects that are showing up at the individual unit? What do you mean? If you're recruiting the replacement or play a role and your own network is very singly, singly ethnicity. Right. Right. So as that was my disclaimer coming into this, I'm looking at patterns here. And since it leaves off in some ways a bigger question, why is this? And I've not seen literature looking in so to me it's a natural direction to go to try to figure out some of these things to what extent do people know who was in the unit, et cetera. Just a couple comments, quick comments on the control bureau's income. Higher income in the micro neighborhood affected, increased the probability that a white household movement moved in. Less likely to have a Latino household move in. The income in the tract had no effect. Residential stability in the micro neighborhood also increased the probability that a white household will move in. And lower the probability that a black household will move in. So these more stable, local areas of the tract had no effect. One final thing I tried doing a little bit is following up on this signaling idea. Like someone said, oh, people don't really know the race, ethnicity of people, how would they know? And then it becomes a question, well maybe there's certain contexts where it's more important to me to know. I mean you can imagine certain people care more than others. Some don't care at all. Some might be somewhat concerned about it. The people who are concerned might be particularly concerned in certain contexts is where they really, it becomes important for them to figure out this, use this signaling information. So I tried a couple things, interactions, kind of contextual things. Residential stability, thinking that there might be less signaling information in a case like this. Because there's a large amount of change going on anyway so you'd figure out all this change going on. The present household is less than an indicator of the household. They just haven't had a chance to move out. They might be getting ready to leave. So it's not as strong a signal. Racial ethnic composition, this may, again leaving aside George's question of how do you know what the race ethnic composition is. You trust your real estate agent. That might really increase for whatever reason the salient said this to you and you might use that signaling more. Or the economic environment might be important. There's less uncertainty perhaps. If I'm going into a high income neighborhood, I'm probably safe to assume it's a pretty quality neighborhood. A lower income one, it might be more dicey. Maybe it's a nice cohesive neighborhood that's just a little lower income or maybe it's a really problematic one. So there I might want to use the signaling much more. Test these out. For stability we do see somewhat of an effect that looking again at that micro neighborhood, there's no effect for the track but looking at these interactions, what effect do they have on this same race effect for the household. Neighborhoods with more stability, I'm sorry, micro neighborhoods with more stability, the white to black transition another 34% less likely. So even less likely to see that transition occur when you see a lot of stability in the micro neighborhood. And likewise Latino to white transition 20% less likely again in a very stable micro neighborhood. So again, somewhat consistent with that idea. Asking about the racial ethnic composition, if you increase the percent black in the micro neighborhood over the track, either one, black to white transition another 35% less likely. More Latinos in the micro neighborhood attract again, Latino to white another 34% less likely. So again, before that, whites seem to be more affected by the broader area. Here we're seeing it, it seems to affect this same race effect even more. Income had no effect, thought it would, but it didn't. One last thing we tried was looking at the broader context. Our viewers said, oh, look at central city areas. That turned out not to have an effect, but we thought, well, central city, really what that is a proxy for when people talk about what they mean, low income, high crime. We tried race as well, doesn't have an effect, but it's such a big context. But we do see a strong effect for violent crime and income. And it turns out, if you look at the county violent crime rate, that does in fact affect this same race transition. The transition from white to black decreases, it's a weird freezing, like black transition. The odds decrease 40% of seeing this transition or black to Latino. Again, that decreases quite a bit for a high violent crime county, so in that larger context. And, flip, you see the same thing if you look at the lower income counties. Same thing, white to black decreased, black to Latino decreased. So again, in these broader contexts, either high crime or lower income, that same race effect is the same. Almost out of time. So let me quickly summarize what I've been showing here. What does this mean? So we see this strong household effect that persists even controlling it from micro neighborhood and the broader track. That's kind of inconsistent with some of these models. Place stratification, the preferences, classic assimilation or not. There's something else going on here. Of course, the strong micro neighborhood effects also imply that stratification, assimilation, signaling. Signaling model doesn't account for that, but we still see this micro neighborhood effect as well. Of course, it's possible that there's more than one of these going on, quite likely. The micro neighborhood effects are consistent with some of these theoretical models, preferences, steering, networking. It could be that the case, but which one is at work? I can't say here is more. Again, I was looking at patterns. Again, it's possible that more than one is at work. So again, that might be something for future research to kind of look at. And again, this question of what is the important context which I started with, this idea trying to pick out what seems to be an empirical pattern, what reveals itself based on behavior. There's still this question of exactly how does it work? What do people look at? Both the micro neighborhood and the tracks appear important, especially for whites. But again, this issue of the race, ethnicity of the previous residence seems important, but I've seen there's not much literature out there talking about that. Or even these questions of how this might take place, how they know race, ethnicity, race, ethnicity of people. How do they know that? Even there, there's not a lot of literature on that. And different theories that might explain it. We touched on this notion of social networks might be in effect. Whether that differs between homeowners and renters, I think is a useful direction to kind of look at and say that if that is indeed how it plays out, the steering effect could be the case, but it's hard to say, do real estate agents. But it's tricky to find out because no one's going to admit or acknowledge that. So it's a real difficult thing to tease out. So just to conclude, the ratio of the composition of the track matters. So even though these little cards look at the local micro neighborhood, which does have an effect, and in fact, we found that for most groups, it was almost stronger, but the track did matter, especially for whites. The race, ethnicity of the previous household still has a strong effect. And why? It doesn't seem like it's discrimination because we kind of sat across the board. It's not just whites not transitioning to other groups. So it doesn't seem like it's that, but as it networks, it's the signaling. It's not clear, but either way, it's something is going on, something worth studying more. So I'll conclude there. Thanks. Did you just sample one micro neighborhood per track? Essentially. So you never get enough multiple micro neighborhood? At all. There's like about four or five doubles. There's just no variance on that. Yeah. Have you saw urban areas or I missed that? I didn't say it. I should have far-branded whistleback in Carolina would be jumping all over me, but it's urban and suburban basis. So it's non-rural is what they did. So they slopped off them and then they did a random sample from the urban and suburban persons. Again, they were sampling those housing units and then building the micro neighborhood around them. And that was the sampling device. So then that's why it ends up that there's almost none that are two or more in a track, which is nice for standard areas, but it's not great if you wanted to try it. Get this additional variability going. So you know whether it's on or, whether it's around the unit or on the unit, right? I do know that, yeah. And it seems like there are a lot of reasons to want to do this separately. Yeah. Because it might help you to distinguish between some of your theories. I agree. I'm kicking myself. The reason I'm kicking myself now, well, for many reasons, but one of which is now, this work I did back at Duke in their Census data center and now to do the analyses now, I go up to UCLA and I first have to beg and plead for the census and that to show me I have an R and R and I need you. So I need you basically. I'll send the paper to you, right? Critique it, give me an R and R and say I need to go back and with that, then I can go back. That's what I need is I need that and I can go back and say I need to redo these analyses. But it's something that's become startling clear to me that at the time, I thought, oh, it won't make that much of a difference, but that was a pretty hindsight, not a great decision. Because it makes sense that these processes would play out, at least potentially differently. I mean, I did some, I've got the public use data where I can't do the contextual effect, but I can look at same-raised transitions by renters versus owners there and there were some differences there, not huge. But again, that's not really getting that, what you really wanna be telling, you know, testing. Because it's more this idea, some of these contextual effects we might well think would play out differently for those two groups. I don't know, I'm trying to think if that causes problems for estimation, but I'll set this. Do you know how far people have moved? Do you know anything about whether it's a local, relatively local group or a distance? Well, I'm trying to think of the information that people would have available that most moves are relatively short distance. Do we know how they found the unit? Yeah. That's a great question. Almost certain, I don't think they ask where they're coming. Cause all I know is I don't know where the other people laugh, but I know the new people coming in, I mean the people coming in, that's the question. I don't know how far they've come. I don't think I have that. I'm not 100% sure I need to go look on that. They do ask questions about why you chose this unit. You know, what characteristics of the neighborhood, what characteristics of the unit. And I'm trying to remember now if they ask, was this... Like prior residents? Yeah. I guess the main thing we'd probably want to know, I think, where you're going with is, are they, were they in the same metropolitan area or not? I think you could probably split it based on that. If you're in the same metropolitan area, you have relatively decent information. If you're coming from outside. You might have a lot of things. You might have information, access to particular realtors. You might know the neighborhoods. You might also have the network connections. Yeah, right. Whereas if you're coming from outside the metropolitan area, those things... Right, and it might be a much grosser picture of what's going on. You might have heard about, oh, this area, but not that one, but not down to... A real flying grain? You know, Camille Charles has done some work on this, and what she discovered was that whites had more information than blacks. Interesting. Just in general, you mean... About the potential neighborhoods. Either, I mean for people living in the same metro area or even... You're living in the same metro area. Okay. So people coming from another metro area are kind of equally uncertain. They're something, yeah. Yeah, that's worth thinking about, too. I guess I'm biased coming from the Detroit area, but I'm trying to think about what would be a lot of the concerns about these white movers who are thinking about moving into this neighborhood. Why are they concerned about whether their minorities live in there or not? Yeah. And one of the issues is this neighborhood is churning. Yeah. It's going down. The property value's gonna go down. Right. And so that's one of the reasons you'd want to look at it. I would think that you'd see a really strong difference there in terms of like if you have a high SES neighborhood that's clearly being kept up and clearly very wealthy, that in a sense it might matter less what the races of the people are than if it's a neighborhood that's more urban or more sort of questionable in a sense. So you're not really sure where it's going up or down. Right. But you said that you didn't see an income difference. Yeah. That income in terms of the individual person's income or the SES of the neighborhood or... The neighborhood and micro neighborhood. Are you thinking? Yeah. So if this doesn't matter at all? Not for those two. The micro neighborhood, I mean measuring the average income in the micro neighborhood or the tract did not affect that. I thought the same thing. It seems pretty non-plausible to me, so I buy it. And that's controlling for race, ethnicity, and crime and other things? Yeah. Yeah. Right? Yeah. That's what we're testing on the system. Yeah. But I'm not clinging to it for the next 30 years and I give it up. So I just, Grace, I mean so one of the ideas is like, people, race is sort of socially salient, right? Sort of this idea like race is a master's status or something. Yeah, right, yeah. Do you have other thoughts about like, I mean so like the one that sort of jumps to mine is like language, right? Would be potentially really interesting to look at. I mean I was thinking of like other things like sexual orientation potentially, some neighborhood sometimes or class. Yeah. But I just wondering what other kinds of information you might have that you could look at to potentially get information about. I mean, and I don't know if like the different sort of group identities would give you different ideas in terms of like networks, discrimination, signaling. Yeah. But it seemed like there might be some leverage there. So are you going, I suppose, are you not thinking for this particular study, is that what you're referring to? No, no, no. It's like the next group. Exactly, right, yeah, that's, no, right, thanks. I don't have enough to do. Yeah, I mean, absolutely. I mean, race to me in some ways is horrifically boring is what we, you know, it's like, oh my God, it's easy to measure, we kind of put it in. And it used to be tremendously important and it's still somewhat important no matter what we might want to think. But I think, yeah, there's a lot of reason to think that some of these other groups would be important. And that signaling is getting fundamentally at that, right? I mean, that is, well, I mean, people in general, you move to a neighborhood, you want to get a sense of what's going on there. Just talking to someone there, saying there's someone who's, we're trying to do a paper looking at gentrification. And how do you measure gentrification? Someone came up with a proxy, the number of Starbucks. Yes, which I thought was interesting. That might be an easy way, you just drive on, oh, lots of Starbucks, I'll move here. It's tricky though, but yeah, for other groups, yeah, how do you, you know, for gays, who, I mean, that would be a fascinating group to look at because they have, I don't know what the word is, but they've moved into all sorts of neighborhoods in all sorts of different cities, a real process, like how does that go on? I think some of that is, I'm speaking really off the top of my head, but I think there's been some pretty clear signals that are given at points there. But that, yeah, you can imagine how that plays out. And in some of those instances there, you know, it's, there's reasons like that would be pretty important to know that, perhaps, right? Some neighborhoods might be downright inhospitable. That, you know, that's, that would be a concern as well here. And so in that case, again, that signaling is even more important than how it plays out. So then it raises all these, yeah, it raises these questions of how, you know, are you just, in that case, are you just looking at the, the characteristics of the household? Probably not, in that case, if, if we're talking about a case of transition, right, because presumably the people moving out are going to be different than those moving in. So then it doesn't hold up as well. Well, and you're looking at margin-free measures, right? So like, because I think like one of the things, like you have like more immigrants coming in or something like that, like that's essentially that's going to be taken into account. Cause I mean, the likelihood of a white to Latino transition is going to be increasing, but that's not going to show up in there, if I understood your models correctly. Yeah, I have to think about it. Yeah, I know, I know, and I need to think about that cause I don't want to, cause I know, yeah, that's a huge issue for like a lot of the segregation stuff, like, because yeah, certain movement in, you're going to see certain patterns are just a mathematical identity. I don't think that's a problem here, but I'm not going to commit to that, especially since it's on tape. I want to hear a look that other household members like children, I would imagine the people that are small, small-school-age children that are different than in certain ways, they would consider different options on moving in. Right. Like slaves? Yeah, that's a great point. I mean, other stuff I've done, I've looked at no mobility out in looking at crime matter, you know, some of my work has found that crime does matter, but yeah, you do see this strong pattern in here that those with children less than five years are more likely to move. Those with children six to 18 years are less likely to move. And you argue that that is the school factor, that's one simple story, is the school story going on. Yeah, but also when they move, they would sit in that place. Yeah, you're going, yeah, I know it. And it's funny, because I rattled that off and then, no, I have not thought about that for this process, but why would it not work? Of course it would, but certainly. And that's an excellent point. And I've not looked at that, but yeah. It just makes all the pictures, they're really different from people who don't have kids, I mean, they haven't. Yeah, those people with kids just mess everything up, I mean, the restaurants, it's noisy, I don't know how it's going to start. John, our time is up, it's been wonderful. Thank you very much. Thank you. That's all. Thank you.