Search This Blog

Showing posts with label statistics. Show all posts
Showing posts with label statistics. Show all posts

Thursday, November 9, 2023

Conflicting Data on Crime


Richard Rosenfeld and Janet Lauritsen at WP:
On Oct. 16, the FBI released data for 2022 that showed a small drop in the nation’s violent crime rate, including homicide. That’s good news.

Unfortunately, the government’s other crime measurement system — the National Crime Victimization Survey (NCVS) — tells a disturbingly different story. Its findings, released in September, show that violent crime victimization rose — by a lot.
...

One reason might be that fewer violent crimes were reported to the police in 2022 than in 2021. We don’t know why this might have been, but increasing police response times stemming from depleted officer ranks might have made some residents less inclined to file reports. Declining trust in or increasing fear of the police might have played a role as well.


Other reasons for the discrepancy might result from the different populations covered by the two data sources. As a household-based survey, the NCVS does not include people who are homeless or those in institutions such as prisons, jails and nursing homes. It also excludes crimes of violence against those younger than 12. If people included in the survey experienced changes in violence that differ from the changes experienced by those excluded from the survey, that could help account for the different violence rates.

Wednesday, August 30, 2023

Saturday, April 29, 2023

Polling Has Changed -- Continued

 Many posts have discussed the problems of surveying public opinion in the 21st century.


Josh Zumbrun at WSJ:
At the conceptual level, there are things being done that aren’t being disclosed,” said Kathleen Hall Jamieson, the director of the Annenberg Public Policy Center at the University of Pennsylvania.

Dr. Jamieson was a coordinator of a recent effort by the National Academy of Sciences to recommend how to protect the integrity of survey research. It took particular aim at the phrase “representative sample.”

Their guidance ought to update many stylebooks: “The phrase ‘representative sample’ should not be used without explicit acknowledgment of the underlying assumptions, including weighting and modeling assumptions.” They recommend ignoring polls that don’t disclose that information.

There is a stark dividing line between the two main approaches used to build panels. Some surveys still recruit panelists randomly, such as by plucking them off the U.S. Postal Service’s master list of residential addresses or by using other offline sampling methods such as random phone dialing or text-message outreach to assemble their initial panel. This at least preserves an element of randomness. But sometimes, even the panels aren’t recruited via random outreach, but often via online ads and nonrandom methods. In this case, the surveys are even more reliant on modeling and weighting assumptions to get their respondents to mirror the U.S. population.

“Surveys can be valuable even if they’re not nationally representative, but you have to design them for that purpose and be honest about what you have,” said Arthur Lupia, a University of Michigan political scientist who worked on the recommendations.

Saturday, March 11, 2023

Government Data Visualization

 Philip Bump at WP:

The office of Los Angeles County Controller Kenneth Mejia is an exception.

Mejia, elected to the office last year, has a dedicated portal at which his team shares visualizations it has created. Below is a grim example, showing the locations in the city where unhoused residents have died in recent years.
(Los Angeles City Controller)

(Los Angeles City Controller)

Sergio Perez, chief of accountability and oversight for Mejia’s office, explained that the visualizations were part of an effort to make the large amount of information gathered by the office more accessible. There are public-facing ones, like the map above, and ones created for policymakers that remain private.
“We think that the data should be leading public policy decisions, not assumptions or untested beliefs,” Perez told me when we spoke by phone on Friday.

That means a particular focus on housing issues, obviously a key concern for Angelenos and an issue that overlaps with homelessness. The controller’s office also has a map of affordable housing agreements enacted by the city over the last decade, essentially providing an index of places that housing is not (necessarily) egregiously expensive.

(Los Angeles City Controller)

(Los Angeles City Controller)

The person behind these visualizations is the office’s director of technology and innovation, 19-year-old Kyler Chen. (“I’ve never felt older than when I’m in a meeting with him," Perez told me.) He and two interns collect and organize data before creating the presentations that are shared with the public. The intent, Perez said, was to “fill public data vacuums” that exist either because data is inaccessible or out-of-date.
A map showing areas where unhoused people are prohibited from congregating. (Los Angeles City Controller)

A map showing areas where unhoused people are prohibited from congregating. (Los Angeles City Controller)


Sunday, November 20, 2022

Lobbying, Sports Betting, and the Vitality of Mythical Numbers

 Mythical numbers and graphics distort public policy debates.  Sometimes the distortion is deliberateThey call it "juking the stats."

A 2018 SCOTUS decision struck down a federal law forbidding state authorization of sports betting,  Then lobbyists went to work.

Eric Lipton and Kenneth P. Vogel at NYT:

Gambling companies and their allies deployed a bare-knuckled lobbying campaign, showering state lawmakers with money, gifts and visits from sports luminaries and at times using deceptive arguments to extract generous tax breaks and other concessions, according to a New York Times investigation. It was based on thousands of pages of documents and communications obtained in part through open-records requests and interviews with dozens of industry and state officials.
Industry lobbyists, for example, dazzled lawmakers with projections about the billions of dollars that states could expect to collect in taxes from sports betting — projections that, at least so far, have often turned out to be wildly inflated, according to a Times analysis of state tax data.

The gambling industry managed to scare state lawmakers into keeping tax rates low, in part by trotting out data about a sprawling underworld of illegal gambling. The Times found that those figures, which suggested that Americans were placing as much as $400 billion of illicit bets each year, were unreliable.

...
Where did the eye-popping figure come from? The N.B.A. and the American Gaming Association identified the source as a 1999 report by the National Gambling Impact Study Commission, which Congress created to assess the harms of gambling.

“Estimates of the scope of illegal sports betting in the United States range anywhere from $80 billion to $380 billion annually,” the report said.

In a footnote, the report attributed the range not to an academic study or even an industry analysis, but to an Associated Press article from the month before the report was released. That article, in turn, reported that “commissioners were told” the estimate, though it did not indicate by whom.

A transcript from a commission hearing in 1998 points to the likely source. One of the panel’s commissioners, citing unidentified testimony and staff briefings, said that “there’s somewhere, depending on whose guesstimate you take, within $80 to $380 billion worth of illegal sports gambling.”

Sunday, June 5, 2022

Mass Shootings

 Many posts have discussed firearms and gun control.

Júlia Ledur and Kate Rabinowitz at WP:
Before a man killed at least four people Wednesday at a hospital in Tulsa, there had already been 232 mass shootings this year in the United States, according to the Gun Violence Archive. It is the twentieth since last week’s shooting at an elementary school in Uvalde, Tex., left 19 children and two teachers dead. 

Quoctrung Bui, Alicia Parlapiano and Margot Sanger-Katz at NYT:

If the key gun control proposals now being considered in Congress had been law since 1999, four gunmen younger than 21 would have been blocked from legally buying the rifles they used in mass shootings.

At least four other assailants would have been subject to a required background check, instead of slipping through a loophole. Ten might have been unable to steal their weapons because of efforts to require or encourage safer gun storage. And 20 might not have been allowed to legally purchase the large-capacity magazines that they used to upgrade their guns, helping them kill, on average, 16 people each.

Taken together, those four measures might have changed the course of at least 35 mass shootings — a third of such episodes in the United States since the massacre at Columbine High School in Colorado, a New York Times analysis has found. Those 35 shootings killed a combined 446 people.
Morning update:

Wednesday, January 19, 2022

Congress and Polarization


Judge Glock, "The Mismeasurement of Polarization," National Affairs, Fall 2021:
It has become a commonplace of political discourse among academics, pundits, and politicians to claim that one or the other party has become more extremist, or that it has taken a radical position on some particular issue. Most often, those making the charge fail to distinguish the baseline against which the other side has supposedly radicalized.

When those on the left say the right has radicalized, they tend to measure it against the perceived will of the people at the current time, or against some new status quo — such as the Affordable Care Act — that was itself considered leftward radicalism just a few years earlier. When those on the right say the left has radicalized, they tend to measure it against some point in the distant past, or against the long-standing status quo.

What both measures of party polarization fail to take into account are long-term shifts in policy and public opinion that have occurred over America's history. When we look to these changes, it becomes clear that the baseline against which partisanship is measured has not remained static; instead, it has shifted to the left. Both parties have followed this shift over time, with Democrats largely leading the way and Republicans following from the rear.

This changing baseline helps explain why both parties can plausibly claim that the other side has radicalized. For the left — whose adherents measure polarization from the baseline of current policy and opinion, which has itself moved left — Republicans may appear extremist, even if their positions have remained fixed over time. For the right — whose adherents measure polarization against policies and opinions of the distant past, which fell more to the right than they do today — Democrats have radicalized, in the sense that they have followed the shift of policy and opinion to the left of where it once was.

On the whole, American politics has moved, and continues to move, along with policies and the preferences of the American electorate. This may be cold comfort for conservatives, who have witnessed both drift decidedly to the left over the past century and a half. But it can help assuage the alarmism over one-sided polarization that has long plagued our political discourse.

The Poole-Rosenthal scores provide a distorted view of ideological change over time.  

Frances E. Lee, " Patronage, Logrolls, and “Polarization”: Congressional Parties of the Gilded Age, 1876–1896," Studies in American Political Development 30, Issue 2 (2016).  The abstract:

According to the quantitative indicators scholars use to measure political polarization, the Gilded Age stands out for some of the most party-polarized Congresses of all time. By contrast, historians of the era depict the two major parties as presenting few programmatic alternatives to one another. I argue that a large share of the party-line votes in the Congress of this period are poorly suited to the standard conceptualization as “polarization,” meaning wide divergence on an ideological continuum structuring alternative views on national policy. Specifically, the era's continuous battles over the distribution of particularized benefits, patronage, and control of political office make little sense conceived as stemming from individual members' preferences on an underlying ideological dimension. They are better understood as fights between two long coalitions competing for power and distributive gains. In short, the Gilded Age illustrates that political parties are fully capable of waging ferocious warfare over spoils and office, even despite a relative lack of sharp party differences over national policy.

Sunday, December 26, 2021

Misleading Graphs on Labor Force Participation

Philip Bump at WP:
Earlier this month, [Tucker] Carlson ran a segment focused on the decline of men in the workforce. To bolster his point, part of his broader effort to cast American men as endlessly embattled — “The thing about men is they kind of need to work,” he said in the segment — he showed graphs of the labor force participation rate by gender. (That’s the percentage of working-age Americans who are working or looking for work.)

Here, as reported by the Daily Beast, is what Carlson showed.
Man line go down; woman line go up. Done and done. Cable news success.

Yet there are two big problems here. The more immediate is that the vertical axis on the female participation rate graph is mislabeled. It doesn’t range from 65 to 90 since the labor force participation rate for women has never been higher than 60.3 percent. It’s not really clear what happened here. Notice that the grid lines don’t actually line up with the axis labels anyway. It’s just a mess.

But that leads to the bigger problem. The labor force participation rate for men has always been higher than that of women. Here are the same data, with two changes: a vertical axis that runs from 0 to 100 and both measures shown at once. There is still a long-term decline in the participation rate for men — but also for women over the past decade or so.


Wednesday, December 8, 2021

Survey Samples: Bigger Is Not Automatically Better

Surveys are a crucial tool for understanding public opinion and behaviour, and their accuracy depends on maintaining statistical representativeness of their target populations by minimizing biases from all sources. Increasing data size shrinks confidence intervals but magnifies the effect of survey bias: an instance of the Big Data Paradox1. Here we demonstrate this paradox in estimates of first-dose COVID-19 vaccine uptake in US adults from 9 January to 19 May 2021 from two large surveys: Delphi–Facebook2,3 (about 250,000 responses per week) and Census Household Pulse4 (about 75,000 every two weeks). In May 2021, Delphi–Facebook overestimated uptake by 17 percentage points (14–20 percentage points with 5% benchmark imprecision) and Census Household Pulse by 14 (11–17 percentage points with 5% benchmark imprecision), compared to a retroactively updated benchmark the Centers for Disease Control and Prevention published on 26 May 2021. Moreover, their large sample sizes led to miniscule margins of error on the incorrect estimates. By contrast, an Axios–Ipsos online panel5 with about 1,000 responses per week following survey research best practices6 provided reliable estimates and uncertainty quantification. We decompose observed error using a recent analytic framework1 to explain the inaccuracy in the three surveys. We then analyse the implications for vaccine hesitancy and willingness. We show how a survey of 250,000 respondents can produce an estimate of the population mean that is no more accurate than an estimate from a simple random sample of size 10. Our central message is that data quality matters more than data quantity, and that compensating the former with the latter is a mathematically provable losing proposition.

Sunday, October 10, 2021

Research 2021

 Watch this clip from The Wire.  It is the best description of research, ever.

The Internet Archive -- if there is a broken link to what you need, this site might help you find it.  (A newsworthy example, the International Chiropractors Association scrubbed this anti-vaccine diatribe from its website: http://www.chiropractic.org/?smd_process_download=1&download_id=3683

Great stuff at Honnold Library -- which students usually overlook! (password required)
  • Nexis Uni:  news sources and law journals
  • Political science journals
  • Dissertation abstracts (search for "California" and "redistricting" in abstracts, and you will see a couple of Rose Institute names)
General Statistics 

Public Policy and Finance
California and General State Politics

Elections, Parties, Campaign Finance

Public Opinion

Thursday, September 30, 2021

Faulty Data About COVID


Data is key to an effective pandemic response — and the lack of it has hobbled the U.S. response again and again. The lack of testing and then, of standardized reporting of cases and deaths left U.S. officials slow to grasp the scale of the crisis when the virus first began to spread. Insufficient data also meant supplies arrived too late in hard-hit cities. State and federal officials made decisions about travel restrictions and reopening policies with an incomplete picture of what was happening.

Many places were forced to shut down before they had substantial outbreaks, former FDA Commissioner Scott Gottlieb told The Washington Post, and when the virus finally arrived, some resisted a return to restrictions.

“Early on, CDC couldn’t even tell us how many people were being hospitalized for covid,” Gottlieb said.

Multiple factors underlie this data deficit. First and foremost: The U.S. does not have a national health system like Israel or the U.K., and in a pandemic, must rely on a vast and decentralized public health infrastructure that is notoriously underfunded and full of holes. As a result, there is no simple way to track infections or outcomes across a wide swath of the population.

Another obstacle to data aggregation may be the siloed computer systems and the self-interest of medical institutions. Some hospital systems want to hang onto their data, said Michael Kurilla, director of the division of clinical innovation at the National Institutes of Health’s National Center for Advancing Translational Sciences.

“They don’t necessarily want to give up all that data because they see that as a potential future revenue stream,” Kurilla said.

The CDC compiles national statistics by collecting data from every state and locality, but these jurisdictions often have different ways of counting tests, infections and even deaths. The data may not be submitted to the CDC for days or weeks. Many smaller jurisdictions still share that data via outdated fax machines.

Friday, May 28, 2021

Lockdowns Worked


There is copious evidence that lockdowns reduced transmission of the coronavirus. Some types of social distancing restrictions are more effective than others, and some sub-populations benefit more than others, but overall, lockdowns did limit the spread and saved lives....And lo and behold, when we look at evidence, we find that lockdowns accounted for only a small percent of the economic slowdown. For example, economists Austan Goolsbee and Chad Syverson looked at the state border between Illinois and Iowa. On the Illinois side, the towns issued stay-at-home orders, whereas on the Iowa side they did not. And guess what — economic activity fell almost as much on the Iowa side as on the Illinois side!
This is very similar to the results of a comparison of Sweden and Denmark. Denmark locked down and saw its economic activity decline by 29%; Sweden chose not to lock down, and saw its economic activity decline by 25%. The biggest economic destroyer by far was not government policy; it was fear of COVID.

Saturday, September 12, 2020

California Crime

From the Center on Juvenile and Criminal Justice:
A fact sheet released today by the Center on Juvenile and Criminal Justice finds that California’s crime rate fell by 3 percent in 2019 from the year prior, reaching its lowest level since the state began compiling these statistics 50 years ago. Crime rates have continued to fall steadily over the past decade amid major criminal justice reforms, including some that reduced prison and jail populations and lessened penalties for low-level offenses.

As California and the nation reckon with ongoing and historic harms caused by the justice system, declining crime trends offer an opportunity to consider steps toward inclusive community safety. These data offer an important reminder that crime is falling, not rising, which has persisted through an era of progressive policies and substantial reinvestment in communities.

California crime rates*, 2010-2019

California crime rates drop by 12 percent from 2010 to 2019.

*Total and violent offense rates exclude rape because the definition was broadened in 2014, hindering comparisons
across this period.

The fact sheet finds:
  • Statewide crime rates fell by 12 percent from 2010 through 2019, including a one-year decline of 3 percent from 2018 to 2019. Over the past decade, California jurisdictions reported reductions in most major crime types, including burglary (-38%), larceny/theft (-3%), and robbery (-16%).
  • Crime declines accompany a period of transformational criminal justice reform, including the passage of Public Safety Realignment, Prop 47, Prop 57, and Prop 64. Despite initial concerns that these law changes would boost crime, most communities were safer in 2019 than at the start of the decade.
  • Approximately 70 percent of California counties experienced crime declines from 2010 through 2019. Variation across counties indicates recent crime trends likely reflect local practices and conditions far more than state policies.

Read the full fact sheet >>

Saturday, September 5, 2020

Wednesday, August 5, 2020

Crime and COVID

The Dispatch:
A new report from the Council on Criminal Justice—which tracked changes in 11 different criminal offenses across 27 American cities—found the homicide rate increased 37 percent from the end of May through June and the aggravated assault rate rose 35 percent over the same period. Chicago, Philadelphia, Milwaukee, Louisville, Nashville, and Detroit accounted for some of the biggest spikes. “In general,” the report concluded, “property and drug crime rates decreased, while violent crime rates increased during [the spring and early summer].”

A Wall Street Journal analysis of crime statistics casts a broader net, gathering data from the nation’s 50 largest cities. Its findings? Homicides have risen by 24 percent so far this year. The murder rate has increased by double digits in 36 of the 50 cities included in the study. Robberies, conversely, have decreased by around 11 percent year-over-year.
Despite the surge in recent months, the homicide rate remains far closer to its historic low-point than its peak. According to data from the FBI’s Uniform Crime Reporting Program compiled by the Disaster Center, the homicide rate over the past 60 years topped out at 10.2 murders per 100,000 people in 1980. It was less than half that—5.0 per 100,000—in 2018. The statistic reached its nadir in 2014, when there were just 4.4 murders per 100,000 people.

Wednesday, June 17, 2020

How to Look Things Up and Find Things Out

Watch this clip from The Wire.  It is the best description of research, ever.

Lots of links on my homepage

The Internet Archive -- if there is a broken link to what you need, this site might help you find it.   For instance, a campaign website from the past: http://www.johnkerry.com/

Great stuff at Honnold Library -- which students usually overlook! (password required)
  • Databases
  • Nexis Uni: news sources and law journals
  • Dissertation abstracts
General Statistics and The Census

US Budget

Writing, Literature, and Film
Science, Health, and the Environment

California and General State Politics
Elections, Parties, Campaign Finance

Sunday, May 24, 2020

How to Lie with Statistics -- COVID Edition

Neil Irwin at NYT:
Did you hear about the booming air travel industry? It’s up 123 percent in just the last month!
Technically, that’s an accurate number. Over the seven days ended Sunday, an average of 212,580 people went through U.S. airport security checkpoints, up from 95,161 in the week ended April 17.

But of course, that is all wrong if you know anything about the underlying reality of the air travel industry. This time a year ago, 2.4 million people a day went through those same checkpoints. By any reasonable measure, these remain disastrous times for air traffic. It’s just that the shutdown in March and early April made even the slight recovery that has taken place seem like an enormous surge in percentage terms.

Get ready for the same effect to apply to all sorts of numbers — most notably with economic data. These swings are artifacts of the arithmetic of percentage change. But if you aren’t attuned to the yo-yo effect that we are likely to see in crucial data in the coming months, you could get a misleading impression of where the United States stands.
...
When something falls by 10 percent and then rises by 10 percent, it might seem as if it ends up back where it started. But that’s not how the math works.

A 10 percent drop from 100 to 90, followed by a 10 percent gain, would return it only to 99. With bigger swings, those effects become more striking. A 40 percent drop followed by a 40 percent gain would result in a quantity 16 percent below the starting point.

At even greater extremes, you end up with bonkers numbers like those in the air traffic example, in which a 96 percent drop followed by a 123 percent gain leaves you with a number that is still 91 percent below normal.

Monday, April 27, 2020

Watch Out for Graphic Deception

Andrew McGill at Politico Nightly:
If you’ve spent time on Twitter or in front of cable television in the last month, you’ve probably seen a lot of charts.
While they’re vital in visualizing the pandemic — how else can we grasp the spread of a virus approximately 100 nanometers wide? — charts aren’t impartial. They have assumptions; they leave things out. And when they’re deployed to make rhetorical points, it can be devilishly difficult to find the lie.
Here are three things I look for when a new chart flashes across my Twitter feed: 
1. Does the unit of measurement make sense? Bad news can easily be made to look good, if you don’t actually show the bad news.



A few weeks back, the administration deployed this positive-looking chart to show how well testing was going in the U.S. Generally, charts that show an up-and-to-the-left trend mean good stuff is happening. But sharp-eyed chart-readers noted the graphic showed the growth in cumulative tests conducted — not an increase in the number of daily tests, the more important metric. That trend was much more grim, as redrawn charts showed. 
2. Has anyone monkeyed around with the scale? A common trick used to exaggerate — or minimize — a trend is to play around with y-axis. Take a look at this polling chart below:

Wow, looks like Candidate B is way ahead! Not really — the chart starts a number other than zero, exaggerating the gap. If the numbers on the left side of the chart look funky, beware.
3. Does the chart acknowledge uncertainty?







No one truly knows what’s going on. This is as true for coronavirus projections as it is for political polling, or even the fluctuations of the oil market. Good charts acknowledge this by showing uncertainty zones, or prominently displaying margins of error. Bad charts hide this stuff, and pretend to be all-knowing.
Charts help us understand this crisis. We just need to make sure we’re understanding it correctly.

Thursday, March 26, 2020

Tracking Political Money


From the National Institute on Money in Politics:
It's March, and Congress and many state legislatures are busy! If you are a reporter covering proposed bills, be sure to remember this arrow in your quiver: the National Institute on Money in Politics identifies the members of legislative committees in the states and in Congress, compiles data about who contributed to those committee members' election campaigns, and makes tools available to inform your work as you decipher it all. Institute researchers have nearly completed the 2020 legislative committee membership lists; you can access this essential information in three ways, all for free at: www.FollowTheMoney.org.
My Legislature allows you to identify who has contributed to the committee members considering legislation you care about, in addition to seeing who has given to sponsors of any piece of legislation. This enables you to analyze how political contributions correlate with actions by bill sponsors, legislators, and committees. My Legislature also contains information about what actions have been taken on each bill. Simply select for congressional delegates or select your state from the drop-down menu at My Legislature and use the tabs in the upper left to navigate your own very specific search.
Power Mapping for states or for Congress empowers analysis of how members of a legislative chamber or a committee may be inclined to vote on specific legislation or general issues based on their campaign donor pools. If you are an advocate, this tool can be extraordinarily useful for identifying legislators who may be open to persuasion either for or against your issue based on patterns in their fundraising. Journalists may find interesting stories here, as well, particularly in cases of legislators who have donor pools that are dominated by interests pushing a specific agenda. The Power Mapping page includes links in the lower right that describe the tool functionality and provide a video tutorial.
Finally, the Institute's powerful Ask Anything search function serves fully tailored searches. For example, with only a few clicks it is possible to see the largest donors to members of the New York State Senate Housing, Construction, and Community Development Committee or the largest donors to members of the U.S. House Energy & Commerce Committee.
With the Institute's freshly updated legislative committee membership lists and comprehensive campaign finance data, the possibilities are truly endless.
As always, do call us with any questions. We actually answer the phone: 406-449-2480.
###
The nonprofit, nonpartisan National Institute on Money in Politics collects and analyzes campaign contribution information on state and federal candidates, political party committees, and ballot committees. Its free, searchable database of contributions is online at FollowTheMoney.org

Thursday, January 30, 2020

School Demographics

A neat tool from the Urban Institute:
US schools have become more racially and ethnically diverse over the past few decades, but these changes have played out differently across the country. Understanding a school’s demographics can inform conversations around school segregation, redistricting, funding, and equity.
Search for your PK–12 school (prekindergarten through 12th grade) to see how student enrollment by race or ethnicity has changed.

 See, for instance, Mountain Avenue Elementary in La Crescenta, CA: