Income inequality and achievement: The rich test better

There’s a provocative online opinion piece, No Rich Child Left Behind, posted April 27th on the New York Times website by Stanford education professor Sean Reardon. His analysis of test-score and income data leads him to conclude that the achievement gap between the richest and the poorest has grown 40 percent worse over the past 30 years. The rich are now outpacing the middle class by as much as the middle class outpaces the poor. That’s a change from the past where the rich and the middle-class used to score similarly. Today, the achievement differences among different income groups, Reardon argues, are far more pronounced than, say, the black-white achievement gap.

He doesn’t think bad schools are to blame. It’s that the rich are different.

“High-income families are increasingly focusing their resources — their money, time and knowledge of what it takes to be successful in school — on their children’s cognitive development and educational success. They are doing this because educational success is much more important than it used to be, even for the rich,” writes Reardon.

In other words, the rich are signing their kids up for more piano lessons, reading to them more, hiring tutors and exposing them to the Louvre. No wonder their scores are soaring and these little princesses and princes are leaving everyone else in the dust.

I live in New York City amid the one percent of our nation. From my vantage point, I see more hired help , after-school activities and iPad time than parents playing with or reading to their kids. Often the nannies of the rich speak in grammatically incorrect sentences, yet the children under their care still excel on verbal tests. On one level, that gives me reason for optimism. If we could just sign the poor up for more activities and improve preschool education, maybe the poor could catch up some. But then, my pessimism returns. The kinds of enrichment and preschool programs that help the rich excel are prohibitively expensive to roll out nationally. At Chelsea Piers in Manhattan, for example, a “gymjitsu” class costs $35 a week. Plus you’d need a parent who can take off a few hours of work in the middle of the day or an $18/hour nanny to get them there and back. I personally know of several four year olds who have an activity every day.

There’s also an issue of motivation among low-income parents. Even if the government or the non-profit sector creates opportunities for their preschoolers, will they come? I often wonder why I don’t see low-income families at the amazing free weekend art programs for preschoolers at the MOMA.

Data on open enrollment school choice in New York City

Yet another study seems to indicate that white and Asian middle-class families benefit more than minority and lower-class families from open enrollment programs where students can choose to go to public schools outside of their neighborhoods.

The latest finding comes from a data analysis of New York City’s school choice program conducted by The Research Alliance for NYC Schools at NYU entitled, High School Choice in NYC: A Report on the School Choices and Placements of Low-Achieving Students. Not only do low achieving students in the bottom quintile end up at the worst schools in the city, the authors note that these bottom students are often selecting bad schools as their top choices. (Not surprisingly, most of the kids in this group are poor minorities). The most interesting chart was Table 3 on commute times. It showed that low-achieving students prioritize schools that are geographically closer to their homes than other students do. In Brooklyn, for example, eighth graders on average listed a top-choice high school that was 33.9 minutes away from their home. But low-achieving Brooklynites listed a top-choice high school that was 31.0 minutes away from home.

It could be that low achieving students don’t care enough about their education to travel an extra three minutes. It could also be that kids with low test scores are practical. After all, why bother listing Stuyvesant when you know you won’t get in? But the authors write that there are some selective high schools that will take students with low test scores. And so it may be that these students don’t have enough information to know to list these schools among their 12 choices.

Other media coverage on this study:

Schoolbook: Struggling Students Tend to Apply to Weaker High Schools

Education Week: Disparities Found in N.Y.C.’s System for Matching Students to Schools

Academic articles on school choice/open enrollment and how middle-class families take better advantage:

Educational Researcher: School Choice? Or is it Privatization?

Journal of Education Policy: Determinants of school choice: understanding how parents choose elementary schools in Alberta

Education data events

I’d like to keep an ongoing list of education data events. Please email me or post a comment if you know of others.

APRIL/MAY 2013

The American Educational Research Association’s (AERA) annual meeting is taking place Saturday, April 27 to Wednesday, May 1,2013 in San Francisco, California. This year’s theme is “education and poverty” and some of the sessions will be live streamed. Here is the full agenda. Click on “events calendar” to see day by day.

(I’ll be attending as an AERA data journalism fellow and continuing on to attend the Educational Writer’s Association’s national seminar, from May 2 through May 4, at Stanford University. I’ll be sharing what I learn there on this blog.)

JULY 2013

The National Center for Education Statistics (NCES) is holding its STATS-DC 2013 Data Conference “Discovering Through Data” July 17–19, 2013 in Washington, DC. This Friday, April 26th, is the submission deadline. The “major focus will be on Common Core of Data (CCD), data collection, data linking beyond K-12, data management, data privacy, data quality, data standards, data use (analytical), data use (instructional), fiscal data, and Statewide Longitudinal Data Systems (SLDS).”

Clean, clean, clean before you crunch those big data sets

Anyone interested in how data science might transform education should read The Dirty Little Secret of Big Data Projects.  David Dietrich, an impressive data geek consultant at EMC’s education unit who’s been involved with a big data lab at MIT, wrote that 80% of your time on a data project will be spent on the tedious, unsexy task of cleaning up the data. Often, people are so excited to start crunching their data that they end up with wrong answers because they haven’t cleaned up and prepared their data properly.

If you want to try some data cleaning at home, Dietrich suggests that unsophisticated types (such as myself) should tinker around with these two tools: Open Refine (formerly Google Refine) and Data Wrangler (from Stanford).

The politics of education data in Florida

The national debate over making student and teacher records more accessible is playing out in the state of Florida. Last week (week of April 8th, 2013) the Florida Senate voted to consolidate education records in a single, online database. It’s still far from becoming law, but the debate is quite similar to the one over the new inBloom database.

In Florida, backers of the database argue that, if they put all the data in one place, then researchers can see long-term trends and note, for example, how students fare years after a particular intervention. Parent groups fear that confidential information will be revealed or that for-profit companies will capitalize on student information.

Complicating things in Florida is a separate measure to prevent teacher evaluation records from being subject to state records laws and released to news organizations. I wonder if that will turn education data a partisan issue with pro-union Democrats opposed to databases.

Sources:

Bradenton Herald Editorial. Expand Florida education database to improve student outcomes – Thursday, April 18, 2013

Miami Herald news story. Legislators grapples with what to do with student data – Sunday, April 14, 2013

Per pupil spending by school district in the United States

At the end of March, the Hoboken school board voted to increase taxes by 4 percent to pay for the school budget, which spends $23,716 per student, the second highest in the state of New Jersey. It struck me how much school spending has changed since I went to school, when wealthier districts consistently spent more on education than poor districts. In New Jersey, for example, the state kicks in money to help raise the performance of 31 poor communities.

I got to wondering what per pupil spending is across the nation and how it varies with income and achievement. As a first step, I wanted to calculate per pupil spending in each district throughout the United States. The Census Bureau publishes a table of per pupil spending in districts over 10,000 students. (See Table 17 in Public Education Finances: 2010 — published in June 2012). But that threshold is too high to look at spending in the nation’s wealthiest districts with fewer students, such as Scarsdale, NY with 4718 students or Beverly Hills, CA with 4943 students.

A U.S. census statistician directed me to a data set from FY 2010 (http://www.census.gov/govs/school/) that lists the total current spending for elementary, middle and high school students in each school district in the United States. From this figure, I subtracted the funds that school districts pay out to private and charter schools and then divided by the number of children enrolled in each school district. (If you want to try this at home: [TCURELSC – V91 – V92]/V33.)

The data are even messier than I expected because not all school districts are the same sort of beasts that run schools and educate kids. More than a thousand of the 14,449 districts showed zero students enrolled in them. Half of these zero-pupil districts are Educational Service Agencies (ESAs), which handle administrative functions and teacher training for regions. There are also non-operating rural school districts which have suffered from depopulation and haven’t formally shut down. Or brand new school districts that haven’t enrolled any kids yet.

Once I filtered out districts without kids, the range in spending was crazy. At the top was the East Metro Integration District surrounding St. Paul, Minnesota. It apparently spent $342,500 on each of its 32 students. But I’m not sure why it has any enrolled students since the entity’s mission is to foster racial diversity in 10 different school districts that enroll thousands and thousands of children.

Supposedly, at the bottom was the Spencer Valley Elementary School District in California. The data show it spent less than $679 on each of its 2,266 students. But I don’t think the thriftiest person in the universe could educate a child that cheaply. Furthermore, the district’s website boasts a 6-to-1 student-teacher ratio, iPads for kindergarteners and Macbooks for the upper grades.

Perhaps there’s a basic data entry error with Spencer Valley. But how can I simply narrow this list of school districts down to ordinary ones that educate public school students within defined geographical boundaries? I began by filtering out school districts that enroll fewer than 200 students. Rising to the top were “special” districts that serve and provide support to children with special needs. At the top of the list was Santa Barbara County in California, where per pupil spending runs more than $84,000 per student for 701 special needs children.

Next I filtered out districts with the word “special” in them. Again, the districts rising to the top did not seem like conventional school districts. Five of the highest spending districts on my screen had the word “vocational” in them. Number 1 was the Los Angeles County Office of Education. I have no idea what that is. It has fewer than 9,000 students enrolled and I can see that it’s not the Los Angeles Unified School district, listed separately with 670,000 students.

This is my top 10…

U.S. school districts with the highest spending per student for fiscal year 2010. Excludes districts with fewer than 200 students and that exclusively support special needs students.

At the moment, I feel stuck. The school district structure has gotten so complicated that it’s really hard to compare spending per student throughout the country. Some districts include the amounts they spend on special education, some districts separate them out into special districts. There are regional and statewide districts that pool resources together to reduce overhead costs. But those teacher training and back-office expenses should still be factored in for the students they benefit.

I was also hoping to show per pupil spending throughout the nation on a map. Districts that spend the most would be in the darkest colors. Those that spend the least in the lightest colors. But I couldn’t find a publicly available school district map. I did find this terrific new school district map of the United States developed by Blue Raster and Sanametrix for the National Center for Education Statistics. But there isn’t a KML version of the map that the public can use to overlay our own data on it.

Privacy, big data and education: more about the inBloom databases

A new national database of personal student information understandably has parents and privacy advocates alarmed. As reported elsewhere, the new inBloom database houses information on millions of school children from nine states and includes names, addresses, telephone numbers, disciplinary records and learning disabilities.

One of the states is New York. Naturally, the mommy listservs in Brooklyn, where I live, are going wild with “opt-out” letters. My first reaction was surprise. Could it really be true that inBloom was going to release this private information to any ap developer who asked? (Disclosure:  inBloom, a non-profit organization, is funded by the Gates Foundation and the Carnegie Corporation of New York, which are also among the funders of The Hechinger Report).

inBloom explained to me that there are two separate data stores. One is the real data that belongs to the states and school districts. The other is a sandbox of fake data for developers.

With the real data, inBloom is functioning like an off-site storage service for school districts. inBloom says it will store any type of data that states and districts feel is relevant to their educational purposes. That’s why social security numbers could be in one of the fields. But inBloom spokesperson Barbara Roos says inBloom is not aware of any school district using social security numbers. Roos added that the regional data will not be mingled into a single national database; each state’s and district’s data will be maintained separately.

The sandbox contains only fake student data that developers can use for testing their new products. Sharren Bates, inBloom’s Chief Product Officer, emailed me that her development team generated the sample from scratch to be representative but not at all connected to any real student data set. …Some of the data was machine-generated and some was generated by hand.”

The hope here is that new applications will be able to plug into a school’s existing computer system. That would save schools hefty integration costs every time they buy a new piece of software.

But third parties can get access to the real data. That’s when a school district has directly hired a company to be an application provider of, say, a math program. The school district can authorize inBloom to release data to the vendor. inBloom is, in effect, a middle man. The school district could have just given the math company the data directly. Before inBloom’s creation, that’s exactly what happened.

Generally, there’s a contract between the school district and the vendor, defining exactly which data the vendor can have access to. The district could choose to include students’ arrest records, but that would be unlikely with a math program. inBloom says it has set up a “granular” system so that a district can specify exactly which data it wants and doesn’t want to go to a third party.

Of course, there’s plenty to be worried about. What if a school district bureaucrat makes a mistake and accidentally releases data to vendors that he shouldn’t have? What if a software company fails to protect this sensitive information?  What happens if personally identifiable information is transmitted to a vendor that goes out of business?  And how secure are these clouds, where the data is stored?

“Some people don’t want information about their children used or documented at all. Certainly they have the right to that opinion, but that’s a larger issue,” said inBloom’s Roos. “Recording information about students, storing it in the cloud and sharing it with vendors has been happening for a while.  inBloom hasn’t started that. People need to be having that conversation with their school and the government.”

Indeed, that conversation is starting. Last week the Washington Post reported that the Electronic Privacy Information Center sued the U.S. Education Department for making a change to regulations in 2011 that allowed third parties, such as software companies and foundations, to get access to private student data.

Earlier Hechinger Report coverage:

Big data and schools: Education nirvana or privacy nightmare? By Anya Kamenetz on March 6, 2013

Taking college courses in high school, new dual enrollment data

There’s been a surge in the number of high schoolers taking college classes, and it’s not the nerdy bright kids anymore. That’s the takeaway from some new data tables published by the National Center for Education Statistics (NCES) that were publicly released in March, but dated February 2013.

The new data report that about 1.3 million U.S. high school students took classes for college credit during the 2010-11 school year. That’s a 67 percent increase since the last time the NCES counted during the 2002-03 school year, when about 800,000 high schoolers took college classes.

Another way to see the rapid growth is by the number of high schools participating. The NCES data report that almost 15,000 public high schools (82 percent) enrolled students in 2 million college courses. That’s 4,000 more high schools than in 2002-03.

These are not AP courses, but supposedly actual college classes. They’re often taught by a community college, but sometimes by a four-year university. Students don’t always a taste of college life. In many cases, these “college” classes are physically held at the students’ high school and taught by high school teachers. The new data show that 64 percent of the colleges that sponsor dual-credit programs held at least some of their classes in a high school. And 79 percent of those schools use either exclusively high school instructors or a mix of college and high school instructors to teach the classes.

Of course, the rapid growth of college credits in high school and the use of high school instructors would make anyone wonder whether these are watered-down college courses. Are these students really getting college caliber instruction for their college credits? When the students get to college, do they have to take these courses again? Are they falling behind their peers? These data don’t answer that question.

I talked with Joni Swanson, who wrote her dissertation on dual enrollment programs and is actively involved with the National Alliance of Concurrent Enrollment Partnerships. She mentioned a couple earlier studies in Oregon and Florida which she says show that the students who take college courses in high school are keeping pace with their peers when they get to college. But we have nothing on a national level.

The interesting development is the transformation of these college credits, from a gifted education option to a college-readiness program. When I was in high school in the 1980s, there was one wunderkind in my high school who took an advanced calculus class at a local community college. He was two years ahead of grade level in math and completed high-school calculus his junior year. It’s a classic example of how these programs started as an option for bright high schoolers who have exhausted their high school curriculum, but don’t want to start college at age 16 and miss their senior prom.

“There are many more academically middle students. These are definitely not programs for the top five percent anymore,” Swanson told me.

Swanson says many high schools are using college courses as a way of combatting senioritis, the universal affliction of boredom that hits at the end of high school. They’re being promoted to students who are on the fence about going to college, to give them a taste of what college will be like.

And they’re increasingly being used to motivate kids who can barely pass their high school classes and are at risk for dropping out. The NCES data report that more than 22,000 high-school students enrolled in college courses during the 2010-11 academic year in programs that were specifically targeted to at-risk students. That’s more than triple the number in 2002-03 when 6400 at-risk high school students were taking college classes. We have no idea how many of these students completed the classes, received credits and eventually went onto college.

Swanson says there’s a big data challenge in figuring out if these programs are successful. That’s because students often move out of state after high school and it’s hard to track them down.

What’s clear is that these dual enrollment programs are already becoming institutionally entrenched before we know whether they work. In Iowa, for example, every community college is required to offer dual credit courses to high school students.

Data resources on dual enrollment programs:

2013 NCES data on dual enrollment, covering the 2010-11 school year.
http://nces.ed.gov/pubs2013/2013002.pdf

2005 NCES data on dual enrollment, covering the 2002-03 school year
http://nces.ed.gov/pubs2005/2005008.pdf

Oregon study showing college performance of high schoolers who took dual credit courses:
http://www.ous.edu/sites/default/files/dept/ir/reports/dualcredit/DualCredit2010FINAL.pdf

Links to the Florida and California studies:
http://www.postsecondaryresearch.org/index.html?Id=Research&Info=Dual+Enrollment

National Alliance of Concurrent Enrollment Partnerships research
http://nacep.org/research-and-policy/research-studies/

Community College Research Center at Columbia Teachers College
http://ccrc.tc.columbia.edu/Dual-Enrollment-and-College-Credit-Programs.html

Hechinger Report stories on dual enrollment expansion:

http://hechingerreport.org/content/new-dual-credit-trends-emerge-as-pioneering-post-secondary-education-options-turns-25_3238/

http://hechingerreport.org/content/low-income-and-at-risk-students-the-focus-on-dual-credit-programs_3262/

http://hechingerreport.org/content/meet-five-students-who-got-a-jump-on-college_3291/

Introducing Education by the Numbers

The fetishization of data has hit both education and journalism. And that’s why I’m starting this datablog. My aims are many. I plan to list and summarize which data sets and studies are available on certain education topics as a resource for journalists and other lay people. I’d like to write about interesting people who are crunching education data. And I will write about new data studies or stories about the use of data. At times, I will try my own hand at some data analysis and graphs.

Most of all, my aim is to become educated enough about education data to develop ideas for data projects. I’m convinced the way to become a new-fangled data journalist is to start getting messy with numbers.

I’m new at this. I’ve been an old-fashioned print and radio reporter for 20 years. I have a few graduate courses in statistics under my belt. From that experience, it seems to me that data analysis is a group exercise. At a minimum, it’s helpful to have a partner to check your work and find your mistakes. So this will be an experiment in flying solo. I’m hoping readers will comment and criticize like peer reviewers. I will keep refining and editing as we go along.

I’m fascinated with what data can tell us about which education programs work and which don’t. I’m openly curious about so-called “adaptive learning” — the idea that we can mine data to understand how each human brain learns and design personalized curriculum for each student.

For me, it all began with a data analysis course at Columbia Teachers College taught by Prof. Doug Ready. Our seminar class spent a year playing around with a large data set of kindergarten students as a vehicle for learning statistical techniques and SPSS statistical software. I loved asking the data questions. Does music instruction improve math ability? Do the children of Tiger Mothers score higher on tests? The answers often depended on socio-economic status. One answer for rich kids. The opposite answer for poor kids.

As I begin this, I’m not sure how much I trust the answers that the data spit out. I’ve learned that small changes in how you crunch the numbers can generate different answers. Sometimes the data you have isn’t really the right data to answer your question. Often the correlations you find are just coincidental and you can’t conclude that doing x will produce y result.

At its core, data analysis is unsettling. Until now, we’ve relied on the experience, wisdom and hunches of veteran educators to say how to teach our children. Many now hope the data know better and will somehow save education.

I’m curious to see what the data say and factor it in.