Question about the Census and Census data

From an anonymous reader:

You consume more economic research and data than anyone I can think of and you always ask great questions, so I wanted to ask: Do you have any experience working with census data and any thoughts about the kind of data, timeliness of data, etc, that you think is lacking from government?

…Crises like the pandemic can force governments to make some changes…I think Census has a “no turning back” sense about knowing they need to try some innovative things (like the pulse survey), so I’d be happy to hear about any thoughts you have on economic data and surveys. Is there any low hanging fruit? Is there something that frustrates the hell out of many researchers? Are there moonshots in data or data-linkage that the census could attempt that you think could be valuable?

Please do leave your suggestions in the comments…

Comments

I was a census pre-list enumerator in 1979. Clearly the technology has changed. But an issue that surely is still true today is the magnitude of the task, and how relatively infrequent (say 1%) of peculiar circumstances suddenly become very big deals across 100s of millions of people.

So data on how many people, undocumented or not, generally live in a particular dwelling, how many homeless people are there, how many people living in "things" or "places" most of us wouldn't call a dwelling, but which they call home, and so forth, loom very large.

Respond

Add Comment

Apologies, don't want to wreck any serious conversation, but Hong Kong doesn't collect much of any economic data, and has been extraordinarily successful.

There's this https://fred.stlouisfed.org/tags/series?t=hong+kong, of course, but much is repetitive and/or irrelevant [such as Balance of Payments statistics :-)].

Anyway, Hong Kong seems to have succeeded without much data. Maybe so-called advanced countries can do the same.

Respond

Add Comment

You may want to see Andrew Whitby's recent book on the census and its history, The Sum of the People

Does he discuss Numbers? A Jewish guy I knew refused to take the census for religious reasons. As for the OP, it's kind of naive, in that the questioner does not seem to know that Census data routinely lags market data, meaning the market data will tell you 100 people live on a certain street while the official census only lists 80. If anything, but for the Constitution, they should do away with the official census. Kind of like 2700chess.com vs the official FIDE rating list for chess (the former is a private initiative and more accurate for real-time ratings).

Respond

Add Comment

Respond

Add Comment

Census should be:

0. Focusing more on availability and linkability than "timeliness", which they have historically excelled at [at the expense of other, in some ways more important objectives].
1. Trying to finalize its data sharing agreement with IRS/BLS with respect to the American Community Survey, where income imputation has been *bad*. In general, it is totally unclear why the Census and BLS are separate operations altogether.
2. Trying make sure all of its historical data is preserved and making old historical datasets more widely accessible. The 1948 Current Population Survey (or its predecessor) should be entering the public domain, but I think that National Archives basically screwed up and haven't retained any of the microdata or any of the tools needed to actually recover this stuff - so it won't be, despite the 72 year rule. There are 40 years of American data (basically 1950 - 1990) and firm-level surveys which have been "lost" [despite the records being available in some microfilm format] because the computer technology to read these surveys were not retained. If Trump wants to do some kind of stimulus, you could spend 2 billion+ Census data recovery and not waste it because Census archival systems heretofore were so broken. This likely originates in the fact that the Census promotes workers based on their inspiring work on whatever the latest survey is. Anyway, based on this data lying around, you could more than double the statistical power of a lot of the standard time series standard analyses by doing this.
3. Setting up an API. Current data access is barbaric.
4. Taking inspiration from IPUMS in its transparency and data availability.
5. Putting Census proposed differential privacy protections - which threaten the entire microdata enterprise with dubious marginal gain - on hold.
6. The BLS has done a poor job tracking the gig economy and I doubt that the CPS is an appropriate instrument for doing so. Census should think about putting gig economy and "new economy" questions on the SIPP, which is better designed to solve the questions Krueger and Katz originally wanted asked. Furthermore, the SIPP is linkable to tax records, which would enable the reconciliation of several puzzles in the data.
7. Chetty's work on neighborhoods is starting to attract controversy. One way to put that controversy to rest would be to digitize and link the 1980 Decennial to Social Security Records (again, a project that could be done for tens of millions of dollars, and which, if Chetty is right - would yield trillions in returns) and simply track outcomes to 2020.
8. Look into the "Data ACT" and Data Coalition's work. Often the best data is sitting around in government silos - the Census has a role in making that accessible.

+=42

Respond

Add Comment

Respond

Add Comment

Someone asks Tyler for his opinion, and Tyler punts the question to a comment section he considers a sewer. Maybe the anonymous reader should ask the excellent Kevin Lewis instead.

and the idea that Tyler is some paragon of data consumption is comical. tyler consumes more research and data than anyone this commenter knows. classic! high-level trolling there.

Respond

Add Comment

Respond

Add Comment

Does asking people their race in a census cause (correlate with?) greater segregation? Less intermarriage0

Respond

Add Comment

The issue I have with the Census and its impact on representation at the Federal level is that there is no tie into public economic efficiencies. MN is on the cusp of losing representation at the Federal level due to a census count. Yet MN is a net outflow state, meaning we send more money in taxes than we receive back in benefits. How does it make sense that the states with higher levels of public economic performance have a lesser voice in the negotiations of the public purse?

Moreover, it is becoming abundantly clear that our efficiencies may have worked for the greater group, but is sorely hurting the minority population which has grown from around 5% of the population in 1990 to pushing 21% today. While the greater group has established patterns that maintain safe streets, good schools and adequate transit; the inner group is at conflict with established police forces, struggling with public schools and catching the virus on public transit. The ‘defund’ movement is an economic one. They want resources reallocated to better address their suite of public goods, rather than funding another cop who comes into their neighborhood as a show of force.

Somehow it feels like MN’s success is ripping them off of their funding.

If you want representation tied to tax levels, here's your solution:

Representatives and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective Numbers,

But is any state at this point really "net outflow". We're running trillion dollar deficits as far as the eye can see. I think it quite probably every state gets more back than it sends in taxes.

Also what about the methodology? When Target pays corporate income tax, does it all get credited to MN, even though the profits were earned in all 50 states. What about Target's employer's share of payroll taxes. Does that also get credited to MN, or does it get credited to the state where the employee actually lives or works?

These are excellent points. I think what the 'defund' situation is telling us is that there maybe a more accurate way to match funding with social objective; whether that be for a subway system in NY or for a higher quality of life in a neighborhood in North Minneapolis.

Respond

Add Comment

Respond

Add Comment

when the fed. government runs > trillion dollar deficits for 10 years
not sure any state can be accurately defined as a new outflow state

Theoretically MN will pay a disproportionate share of the interest on the debt as well. Though now that we are all MMT that shouldnt matter soon.

Disproportionate on what basis? Population i.e., that everyone should pay the same amount, regardless of wealth/income??

It's weird to see extreme flat-tax positions showing up from (the U.S. version of) the Political Left

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Minnesota's representation in the Senate will be the same after the census. Losing a House seat because state population has not grown fast enough relative to other states makes complete sense. Members of the House represent people and Senators represent states. You should know this. Your Senators could put their feet down and say no more to net money flowing out, as you say. But like other Democratic Senators, they quite like this system of growing the federal government's power to redistribute money. Why don't Democratic Senators propose lowering federal taxes so that the Democratic states can raise their own taxes and then spend the money themselves, rather than sending tax money to DC and then bitching when they don't get that same amount back?

'You should know this.' Hmmmmmm

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Define the problem ... before seekin solution.

The "Census" itself is merely an arbitrary 18th Century procedure to answer a simple population question.

What's the specific Question -- and how else could it be answeed in this 21st Century ?

The Federal government now operates a vast domestic surveillance system 24/7... monitoring 97% of the population, right down to their phone calls, financial transactions, and health records.

It's a cynical charade for the government to pretend that national Census question sheets and door-knocking are necessary to find out how many people are where.

For one thing, the Constitution requires an enumeration.

Just because the government has access to information doesn't mean it is cleaning up and linking all of this information reliably to individuals.

Two examples: when the Nevada Parole Board considered the case of O.J. Simpson, it missed the fact that Simpson had a prior criminal conviction in California for domestic violence. This "vast surveillance system" did not pick up on public records linked to one of the most famous men in the country. Then, you can google the phrase "U.S. citizen detained by ICE" to see several examples of naturalized U.S. citizens who are wrongly accused of being undocumented immigrants and are detained -- sometimes for weeks -- until a relative can locate paper evidence of their citizenship. Again, the vast surveillance system of the federal government apparently does not capture the basic details of naturalization proceedings that may have been held in the past.

Respond

Add Comment

Respond

Add Comment

Just the other day it occurred to me the increased availability of quality DNA lineages could support some interesting longitudinal surveys. Have fun.

Respond

Add Comment

Raj Chetty's new project is ostensibly about tracking the economic impact of COVID-19, but he's made no bones about wanting it to serve as a prototype for real-time national accounts.
http://tracktherecovery.org/

Respond

Add Comment

For my purposes Census data is now 5x or 10x harder to use than it used to be. The main tool I used to create grant proposal needs assessments used to be available at http://factfinder.census.gov. Go there now and you’ll see a re-direct to data.census.gov.

The new system looks superficially better but it’s a nightmare for usability. Each query can take as long as 10 seconds to complete. The old American Factfinder system happened almost instantaneously, and it was easy to quickly generate main tables for most topics (housing, poverty, income, employment, etc.).

I have been in touch with a guy named Tyson Weister, who is the main feedback version for the new way of accessing the Census. His email sig block says,

Dissemination Outreach Branch
Census Enterprise Dissemination Services and
Consumer Innovation (CEDSCI)
U.S. Census Bureau

He has been polite and interested in the topic, and he seems keen on improving the system, but I’ve not seen big changes in the last few months. Simply re-enabling American FactFinder as an alternative would be a tremendous usability improvement.

I have a huge amount of additional, very granular detail on what might optimistically be called the challenges of using the new Census interface for data work. It’s so bad that we’re telling clients to choose only one or two zip codes, and only one municipality, whenever possible, because the data work is now so hard and laborious. I made some screencasts demonstrating why the new data.census.gov interface is so dang hard to use, and I need to get those up as well.

I'd be happy to see pulse surveys and the like, but the fact that the UI/UX of the main Census interface has taken such a gigantic step backwards mitigates against my optimism towards that.

I’ve been meaning to write a Grant Writing Confidential (http://www.seliger.com/blog) post on this subject.

This is useful; I get statistics from the census.gov website only rarely. So rarely that each time I go there I have to re-learn the query tools, which often have changed in the interim. Thus I hadn't been aware of the new UI, sounds dreadful.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment