Economists modify a SIR model with a spatial and also behavioral dimension

We simulate a spatial behavioral model of the diffusion of an infection to understand the role of geographical characteristics: the number and distribution of outbreaks, population size, density, and agents’ movements. We show that several invariance properties of the SIR model with respect to these variables do not hold when agents are placed in a (two dimensional) geographical space. Indeed, local herd immunity plays a fundamental role in changing the dynamics of the infection. We also show that geographical factors affect how behavioral responses affect the epidemics. We derive relevant implications for the estimation of epidemiological models with panel data from several geographical units.

That is from a new paper by Alberto Bisin and Andrea Moro.  Here is a good sentence from the accompanying and descriptive tweet storm:

In Spatial-SIR, local herd immunity slows contagion initially in the less dense city, but faster global herd immunity slows it in the denser city later

I think this means West Virginia is in for some hard times fairly soon.

Comments

Basically, the entire U.S. is in for some hard times fairly soon. Maybe the catchphrase should have been the corona pogo stick.

Respond

Add Comment

Tyler, even assuming that this new research is reliable and relevant to our understanding of the pandemic, the conclusions are tentative and none should bet on what may happen West of the NorthEast based on the paper's abstract. Anyway, it would be a big surprise if the West were to parallel what happened in NorthEast (btw, why do you ignore what happened there?).

Respond

Add Comment

Fancy, but I am skeptical about how practical it is without a ton of data.

The notion that geography can be important, including local or spatial differences in behavior and density and speed and distance of travel and contacts is indisputable. But how can we estimate all of those parameters in particular accounting for the variations of local geography (the river that runs through the middle of the city; the freeways and subways; the low-income neighborhoods near the food processing plants; the high-income highlands; etc.)?

I don't think we can. The example they give in the paper is the typical mythical ideal city, the spherical cow of urban economics. When it comes to estimating all of those local parameters in real life, they have some advice that is common sense -- and I suspect impossible to implement: "It is important to choose geographic units of analysis so that density and other geographic characteristics gi are relatively homogeneous."

In other words, neighborhoods. Even if we have decent background information on each neighborhood (the article suggests using "Big Data" sources to snoop out residents' behavior), the sample sizes are going to be small unless there happens to be a nursing home in that neighborhood, in which case the neighborhood's characteristics are largely irrelevant, it's the nursing home that matters.

Or, to get a large enough sample size, we have to aggregate data from several neighborhoods, violating the assumptions of their model of homogeneous locales.

They have additional advice about which data to gather and use:
"Proxies like the airport activity for the number of outbreaks, the distribution of socio-economic characteristics for the distribution of
outbreaks, the use of public transportation for the movement of agents, could be fruitfully used in both reduced-form and structural estimates."

Well yeah, but that's what people are doing anyway, e.g. the papers that Tyler linked to the past couple of days, that found that public transit (duh) and being Black (hmm) were risk factors.

It'd be great to account for geography and location as the paper suggests. But in real life implementation, I don't think this model gets us very far.

I think a better geographic approach is a less theoretically sophisticated, more empirical one: draw a map with cases (and their dates) plotted, and use either the human eyeball -- or whatever algorithms geographers and epidemiologists use -- to track the location and rate of spread of those cases. And sure, look at the behavior of the residents of the neighborhoods -- is this the Latino neighborhood where people are not physically distancing, or the Asian neighborhood where almost everyone is wearing a mask? But that's all common sense, I don't think this paper is giving us useful guidance.

is this the Latino neighborhood where people are not physically distancing, or the Asian neighborhood where almost everyone is wearing a mask?

Racist! How dare you say such a thing.

Whether it's true or not doesn't matter.

Respond

Add Comment

Respond

Add Comment

No, this this paper is giving us useful guidance.

But it is allowing Tyler to do at least two things at once -
1. Highlight the superiority of economists over epidemiologists in the never ending status struggle, which has completely superseded the class struggle in cosmic importance
2. Start to discuss pandemic reality, instead of studiously avoiding what has been going on since public choice considerations were allowed to run free - followed by the virus doing the same, predictably.

This this is so sloppy - and it is not giving us useful guidance.

Respond

Add Comment

Respond

Add Comment

It's why gym goers in some places in Norway don't get coronavirus: "prevalence" is the key determinant. https://www.nytimes.com/2020/06/25/health/coronavirus-gyms-fitness.html

It's a different "contagion", but it's why black children who grow up poor in a poor neighborhood grow up to be poor black adults. Why is "prevalence" mostly ignored? I suspect it's the myth of the self-made man: that everyone has an equal opportunity to be successful in life. Or to avoid, or get, coronavirus. Why do so many economic studies focus on "heterogeneity" (a word I despise because it's used to confuse) but hardly any focus on "prevalence".

Sigh...trying to educate people on this board is like playing Whack-A-Mole (TM). One reason I've cut back on posting here, it's hopeless (patents!)

Councillor, go here: https://gisanddata.maps.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6

And note Norway is only getting about 16 new cases a day. That's the real reason gym goers are not getting C-19. Nothing to do with "prevalence" whatever that is...unless you mean fewer people per capita have Covid-19.

Bonus trivia: even if you have a vaccine against a certain virus, certain viruses will actually infect you if the rest of the population doesn't have the vaccine, one reason it's important for all people to get vaccinated. The "viral load" of the rest of the population will overcome your individual vaccine. It's a 'secondary effect' (not that pronounced) and also true for parasite diseases like malaria. One reason I once read that a dozen Western researchers all took anti-malaria medications, went to a high-malaria region of Africa, and they call came down with malaria anyway. Murphy's law...

As usual, Ray, you rely on your own self-regard rather than the facts. Read the linked article: The prevalence of coronavirus varies across Norway, and so does the risk of being infected - go to a gym in a place with a low level of prevalence and there's a low level of risk of being infected. It's the same in America. Why some places have more or less prevalence than others is partly the result of "homogeneity" (I couldn't resist). In America today the South is a hotbed because the South has lots of hotheads, many having a self-regard that approaches the self-regard of our friend Ray.

My own self-regard is showing: heterogeneity not homogeneity. Opposites often attract.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Over the weeks I have enquired on this and other blogs - how do you test a test? People have tried to be helpful; I'm grateful for their efforts but none seemed to me to get to the nub of the issue. Now I have come across something that does.

'As three doctors wrote in BMJ, “No test gives a 100% accurate result; tests need to be evaluated to determine their sensitivity and specificity, ideally by comparison with a ‘gold standard.’ The lack of such a clear-cut ‘gold-standard’ for covid-19 testing makes evaluation of test accuracy challenging.” The only gold-standard for an infectious disease is purification of the pathogen from those with positive tests, and inability to purify from people with negative tests'.

My interpretation of this is that you should compare the results of measuring the relevant RNA in swabs with the 'gold standard' of extracting virions from patients and seeing whether they act as expected on monkeys or plates of monkey cells. And check against purported virions from patients who tested negative.

Have I understood the writer correctly?

All the validation work has been done and to keep raising this matter is IMO a waste of time. I continue to see lots of preprints coming out that compare various DNA and serology tests and note them in my daily newsletter: https://agoldhammer.com/covid_19/ The lamentable CDC PCR test when properly done with the correct reagents is the most specific test available and a reliable comparator. If you are trying to look for the 100% illusory best test, it's quite frankly a waste of time.

The bottom line is the virus is here and will be for a long time. We know what public health measures work to minimize the effects of the pandemic. The problem is people don't want to follow simple directions. they get sick and some die. It's all unfortunate but that is the modern reality we live in.

"All the validation work has been done": people keep making that sort of windy statement. They mostly seem remarkably reluctant to describe it in terms that a layman might follow.

Best work in an academic setting is from the Mt. Sinai group who have correlated things using viral infection in vitro studies. All 'good' diagnostic labs do their own validation studies before adopting any given manufacturer's assay system. If you want a good discussion of this, Noah Feldman on his 'Deep Backround' podcasts has a discussion with Omai Garner who is the director of clinical microbiology at the UCLA school of medicine. It is the April 22 episode and you need to access it through a podcast app. That's about the best I can do.

Thank you.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Take a look at the bottom of the Wikipedia article "Koch's postulates", especially the end "21st Century" part.

Respond

Add Comment

Respond

Add Comment

This is nothing new. Agent based modeling in networks. Look up the Santa Fe Institute and complex systems. Maybe these economists just learned something that others already knew, or are just applying models that others use. Yawn.

+1; my nextdoor neighbor's son was a post doc at the Santa Fe Institute and worked on some of this stuff. People are spending idle hours reworking the standard SIR model but to what end? So they can get a better predictability? Why not just look at total hospital admissions per unit time and measure the variance. I bet that is better than any of these predictive models.

I pretty much gave up on reading new models about six weeks ago. the only exceptions is when the paper has a catch title.

Respond

Add Comment

Can you link the similar paper from SFI I’d like to read it

Just go to the Santa Fe Institute website.

Mark Newman
Paper #: 02-04-020

The study of social networks, and in particular the spread of disease on networks, has attracted considerable recent attention in the physics community. In this paper, we show that a large class of standard epidemiological models, the so-called susceptible/infective/removed (SIR) models can be solved exactly on a wide variety of networks. In addition to the standard but unrealistic case of fixed infectiveness time and fixed and uncorrelated probability of transmission between all pairs of individuals, we solve cases in which times and probabilities are nonuniform and correlated. We also consider one simple case of an epidemic in a structured population, that of a sexually transmitted disease in a population divided into men and women. We confirm the correctness of our exact solutions with numerical simulations of SIR epidemics on networks.

https://www.santafe.edu/research/results/working-papers/the-spread-of-epidemic-disease-on-networks

Just look up SIR models on networks, and agent based models on networks and SIR. This is nothing new. If you are interested in the subject, look up Barabasi, Network Science, at Northwestern or Jackson on Social and Economic Networks.

Respond

Add Comment

Respond

Add Comment

I spent $40 on The Teaching Co.`s Complexity DVD lectures. I'm not sure how much (if any) I bought into the idea. The core properties of a "complex" system are 1. Many interacting components (i.e. separate, possibly dissimilar, "units".) 2. Feedback amongst them. and 3. Evolution of (some of) the components over time (components change their behavior based on the feedback received.)

Respond

Add Comment

Respond

Add Comment

This is the problem with model makers not actually being involved in the issue. They feel honor bound to make the data fit the model (see Global Climate Models) rather than seeing how the sensitivities from the model help guide actions. As I tell my younger engineers, information is not of any use if there is no action or insight for action arising from it.

Respond

Add Comment

Models are interesting for answering general questions about general pandemics. They are useless for a particular pandemic because when you most need the model--in the earlier stages--you things that matter you have to guess at. That include what % of the population is susceptible, estimates of Ro, longevity of virus on surfaces, virulence, etc.

Guess wrong on any of those, and your outcome is wildly different.

If January 23rd Aliens told us this was coming, but the IFR for those under 60 would be half that of flu, we wouldn't have done anything except guard the old folks.

"but the IFR for those under 60 would be half that of flu"

If the CDC is correct with the new estimate that 25 million Americans have had Covid19, then the IFR would be around 0.5%. whereas the flu is around 0.2%. So, no, it's still about 2.5x worse than the flu.

But yes, the correct response would have been to guard the old folks and beef up the hospitals.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment