This piece is marred by some unfortunate polemics, but it makes one core point very effectively:
To recap, white students in Texas perform better than white students in Wisconsin,black students in Texas perform better than black students in Wisconsin,Hispanic students in Texas perform better than Hispanic students in Wisconsin.
I can't do cut and paste on this Mac, so here is the link: http://iowahawk.typepad.com/iowahawk/2011/03/longhorns-17-badgers-1.html
I thank several MR readers for the pointer.
Next Comments →
Wow, this is taking this website to a new low, quoting a wingnut website like this. If he read Krugman's article, he was talking about spending per pupil and the state of Texas. And he was correct. Why can't these Krugman critics just make intelligent criticisms without having to make ad hominem attacks?
Tom,
"quoting a wingnut website"
"intelligent criticisms without having to make ad hominem attacks"
…
Iowahawk can be hilarious, this is a blog not an NBER paper.
that was sarcasm.
"Why can't these Krugman critics just make intelligent criticisms without having to make ad hominem attacks?"
Because Krugman is a an ad hominem attack.
I am confused, if all ethnic groups in Texas outperform their respective ethnic group in Wisconsin, how can the overall ACT scores in Texas be so much lower than in Wisconsin ?
and why do we care about the specific comparison of texas and wisconsin? why doesn't he compare all of the collective-bargaining-illegal states? oh, he's just trying to show krugman's argument is faulty? well congratulations, what an achievement.
"Only 5 states do not have collective bargaining for educators…" and Burge cherry picks one to compare to Wisconsin. I wonder why.
At first glance, it looks bad for Wisconsin in this comparison with Texas. A couple notes of caution:
1) Iowahawk says participation is pretty much universal in the tests. That may be true, but the data on which the reports Iowahawk refers to are based on a sample of about 2500 students in each state.
See: http://nces.ed.gov/nationsreportcard/about/nathow…
2) That means that any comparisons of two states are between two estimates. So any statement that a subgroup in Texas scored better than a subgroup in Wisconsin should be qualified with some measure of the Standard Error of the difference. It's quite possible that results for the entire relevant populations could be reversed in some of the subgroups.
I should note that it would seem 17-1 is a pretty bad score for Wisconsin even if some of that difference is just due to chance. Also that the polemics of the blog post by Iowahawk make me distrust him.
I grew up in Texas (Houston) and my friend grew up in Wisconsin (Wauwatosa). His school didn't have tracking and I was heavily tracked from 4th grade all the way through high school. I think there were something like 5 levels of Senior English at my school.
No idea if that's uniform throughout the state, but could explain some of the discrepancy.
NAEP is only given to 4th and 8th graders. There really isn't another test given to all students nationwide like that.(SAT/ACT aren't required).
"how can the overall ACT scores in Texas be so much lower than in Wisconsin ?"
The ACT is only popular with universities in the Mid-West. The SAT is more broadly accepted, hardly anyone from Texas would take the ACT unless they wanted to go to U of Chicago or Iowa State or something.
@ Michael Kogan:
Offtopic – your father's name wouldn't happen to have been Boris, would it?
"I am confused, if all ethnic groups in Texas outperform their respective ethnic group in Wisconsin, how can the overall ACT scores in Texas be so much lower than in Wisconsin ?"
Wisconsin: 86.2% non-hispanic white. 3.2% hispanic. 5.7% black
Texas: 32.8 non-hispanic white. 37.6% hispanic. 11.8% black
http://www.wwhf.org/documents/Demographics.pdf
http://en.wikipedia.org/wiki/Demographics_of_Texa…
That means that any comparisons of two states are between two estimates. So any statement that a subgroup in Texas scored better than a subgroup in Wisconsin should be qualified with some measure of the Standard Error of the difference. It's quite possible that results for the entire relevant populations could be reversed in some of the subgroups.
A lot of MR posts lately have made me curious about the underlying statistics. Mostly we argue about causes and effects but how reliable are all these data?
As Frank Howland points out, there is a lot of sampling and other statistical massaging that takes place behind the scenes. Are there any generic surveys about the quality and reliability of the techniques used in the social sciences datasets? Maybe we are all arguing about a non-existent effect? Lies, damned lies and statistics?
There could be a bunch of intervening variables on the Mexican statistics.
One variable affecting child performance is whether the child has school continuity–remains in the same school or goes to a different school every year or even within the year.
If Wisconsin's hispanic population is more mobile (agriculture) or even new to the community, it will have different test results than an hispanic population that is stable and in one place for a long time.
I do not see any controls in these studies…continuity, parents ability to speak English, first generation ESL speaking, immigrant v . multigeneration, rural v. urban etc.
Here is a UW report on the Wisconsin hispanic population. http://www.apl.wisc.edu/publications/HispanicChar…
Note: it is a new population, primarily composed of first generation Mexicans, half have moved to Milwaukee, 40% were born outside of the US, and they have higher poverty rates.
This is a site that supposedly is visited by economists. No one is talking about controls in comparing Wis to Tex for poverty, first generation ESL, country of origin of parent, continuity in the community, how recent the person even was in the community, language used in the school (ESL).
Bill:
This is the internet. Check the number of migrant workers in the respective states, don't speculate in a direction that supports your cause without doing the work. Are you going to seriously suggest that Texas has fewer recent Hispanic immigrants (poor, limited English, ….)?
EVEN IF Texas were not ahead, the fact that it is not behind is more than enough to make the point. If negative correlated is (again) demonstrated, well that's just gravy.
Regarding Steve Miller's point about command-C for copy and Command-V for paste, the Command button is the one that looks like a cloverleaf (the text "Command" may or may not be present on the keyboard and there may or may not be an Apple symbol next to the cloverleaf).
If you are using a Windows keyboard on a Mac, the Windows key maps to the Command key.
Hope that helps!
Bill,
Interesting point, but 40% of 3.2% is 1.28%.
By what mechanism(s) that doesn't raise costs do unions improve quality? Anyone can improve quality by raising costs, even the government…sometimes.
The NAEP data are generally reliable. One problem that we are becoming of aware of is that the sample of students who take the tests is not representative. This is because schools are permitted to exclude disabled and English language learner students. As these students are likely to perform poorly, this has the effect of inflating every single state's score. But, because states have different policies and different category identification policies and exclusion policies, the amount of the increase due to exclusions likely varies from state to state.
Looking at the exclusion numbers, it is obvious that Texas had a policy of excluding large numbers of ELL students from the 2009 tests. In the fourth grade, 21% of Texas students were identified as ELL, and 6% of students were excluded from taking the test because they were ELL. California, another state with a large number of ELL students, identified 30% of its fourth graders as ELL, but only 1% of students were excluded because they were ELL.
The effect of exclusion policies in the Texas-Wisconsin comparison is hard to gauge. Texas excluded 9% of its fourth graders in 2009 as compared to 4% for Wisconsin. In eighth grade, the total exclusion rate was the same–5%.
Based solely on a cursory look at the date, it appears that differential exclusion rates affect the fourth grade reading test to a large degree but not any of the other tests. The scores are close enough, however, that I would not comfortably draw any conclusion from the data.
There appears to be a market failure in Mac usage.
I'm so old that I remember when MS aped Apple for copy/paste, finally giving up on using the F-keys for that.
It seems pretty clear that Krugman made some assertions that Iowahawk shot down. Now Bill and others are trying to salvage Krugman's bacon.
Looking at the Pew Hispanic center site, it appears that the percentage of foreign born Hispanics is very similar in WI and TX (34% and 32%).
I love it. Krugman uses data that doesn't control for any variables to make some fatuous point about his opponents being evil. Someone controls for one variable to show that the data does not necessarily prove Krugman's point. Krugman's allies get angry that someone would dare use data without controlling for all variables.
I did not need additional proof for the uselessness of nobel prizes.
But, krugman/obama's nobels tell us far more about the prizes than about the two turduckens.
Bill,
The link you shared shows that Texas identified just over 2 times as many ELL students in the sample as Wisconsin, and excluded….just over 2 times as many ELL students. Their exclusion rates were .31 for TX and .29 for Wisconsin. The numbers for both were so small that the difference is not statistically significant. So…what are you trying to show here? That your earlier complaint about WI having more first-generation immigrants doesn't show up as an issue in the data? Or something else?
So far as I can tell, the comparison between Wisconsin and Texas (and four other southern states) that the Iowahawk blog post seeks to rebut appears to have made by the Economist's blog, not by Paul Krugman. I don't know what Krugman has to do with this, other than that conservatives don't like him.
People are always asking why am I so OBSESSED with pointing out unpleasant differences between ethnicities. Can we all just remember them in the back of our minds without mentioning them publicly. After all, everybody knows that Texas's overall performance is dragged down by its huge number of Hispanic students, right?
Except that Krugman didn't know or chose not to remember that simple fact — and the vast majority of his readers would therefore be clueless.
No, when we try to hide the truth, it gets forgotten, as in this case.
"I can't do cut and paste on this Mac"
that is a scandal
Hispanic population
Texas 6,200,000 (Thats more human beings than live in Wisconsin BTW)
Wisconisin 238,000
Assuming the same percentage are under 18 (28% in Texas) the potential hispanice student population is:
Texas 1,736,000
Wisconsin 68,000
Wisconsin better just work on this problem, really.
Krugman mentioned drop out rates of 30%
Texas hispanic drop outs 520,000
Wisconsin 19,995
Yeah, real problem in Wisconsin with 20,000 drop outs. They can always move to Texas where they can compete with the 1.7 million illegal immigrants.
Fred,
You can actually use Wisconsin statistics rather than projections. Here is a source: http://www.apl.wisc.edu/publications/HispanicChar…
Also, are we measuring Wisconsin, or are we really measuring Milwaukee county school system. Half of the Wisconsin hispanic school population is in the Milwaukee county school system. http://www.apl.wisc.edu/publications/HispanicChar…
Bill,
The 2.5 times figure you cited was from the second column. 5 excluded in TX, 2 excluded in WI. Considering the large disparity in the populations between the two, one would expect TX to have more, so that in itself is not informative. The question is, did they exclude more than would be expected, based on their respective populations? For that, one indicator is column 1, the number of ELL students identified, inclusive of both the excluded and included students. Here we see a slightly closer ratio of 2.3 times as many ELL students in TX as in WI. With only 16 in TX and 7 in WI, total, we can't really ascribe the difference between 2.3 and 2.5 to any systematic factor. It's probably just random variation.
So my point is, the number of excluded ELL students in TX is about in line with what we would expect, considering the larger proportion of ELL students in the state overall. This doesn't, by itself, show that TX is dropping the ELL students to improve their scores, unless WI is doing so, as well. If anything, this shows that one of your earlier requests, that it control for the number of ESL students, has been addressed by excluding ELL from the test populations. Thus if WI somehow has more ESL students than TX, it wouldn't affect the test scores, because those ESL students would be excluded from the tests.
One other note from the data, WI appears to be more likely to provide accommodations to the ELL students who are assessed (40%) than TX is (18%), which seems likely to result in higher scores for WI ELL students who are assessed. That still doesn't prove anything, since it could be that WI has a higher threshold for exclusion, and provided accommodation to students that TX would exclude. There doesn't seem to be much evidence from the identified/excluded numbers to suggest that the threshold is much different, though.
Boonton, If anything it appears Texas is leading in educational output relative to other states.
Podunk:
This is an Economics website, but that doesn't prevent us commentators from adding more light than heat and I think Bill has added to the discussion. I think you are right that dropout rates need to be disaggregated. However, it should be noted that dropout rates are likely overstated under current methodology. According to Heckman and LaFontaine (2007), the current approved method compares the number of graduates to the number of 9th graders. However the number of 9th graders is overstated because this is the most common grade for students to be held back, and this bias is even greater for minority students. Things may have changed, but I would still be skeptical about reported dropout rates.
Source: IZA DP No. 3216
The American High School Graduation Rate: Trends and Levels
James J. Heckman Paul A. LaFontaine December 2007
Cliff, No, that's wrong. They exclude MORE ELL on a percentage basis.
I'm so old that I remember when Apple users made fun of Windows users for having all of those "useless" keys (Ctrl, Alt, Backspace, left mouse button, etc.).
Rahul,
No. That would only be if the things you are controlling for are not relevant to the thing being measured (educational performance). Controlling for left and right-handedness would certainly make no difference, but maybe if you somehow mined the data to sufficient fineness, you could find some meaningless division that would make Texas look better even though it is not. That is clearly not the case here. When you control for something that itself is confounding the comparison, you improve the comparison.
Podunck, With this comment, aren't you in effect saying that Wisconsin scores higher in test scores than Texas for ELL students:
You said: 'One other note from the data, WI appears to be more likely to provide accommodations to the ELL students who are assessed (40%) than TX is (18%), which seems likely to result in higher scores for WI ELL students who are assessed.'
Sorry about commenting twice on the same issue! I didn't see that the first comment had actually posted.
From Bill:
Bill, using the numbers you provided:
5%/16% for Texas vs. 2%/7% for Wisconsin are very roughly equal, and, indeed, if for example, that 2% were just 2.1%, while every other number had zero after the decimal point, Wisconsin would have had a higher ratio of exclusion, so, without more precise data, you can't even call the calculated ratios statitiscally different. Sheesh!
I also want to step back and ask several logic points here:
1. Texas has had a history of dealing with hispanic populations; Wisconsin, at least from the data, has had an influx. Do you or should you handicap for change or inexperience. What if the hispanic population is spread out in a rural area school system, for example, and there is no ESL support, but there will be in the future.
2. Half of the hispanic school population from the Wisconsin resides in Milwaukee county. Are we measuring Wisconsin, or an inner city school system. How is the Texas population spread out or concentrated. Are resources concentrated as well to deal with ESL issues.
3. What is the income level of both populations (controlling for cost of living)? Income and Socio economic status effects outcomes. So does parent ability to speak english. 40% of Wisconsin's hispanic population was born outside of the US. What is Texas's?
I am not an expert in any of these statistics. Not at all and by no means. It just took one click to find a report on Wisconsin's hispanic population and a UW report to raise questions and make some obvious points. Maybe there are better statistics. Maybe the statistics we are looking at re NAEP scores should be interpreted differently. I'm no expert.
Frankly, I am an agnostic on this and want to know more, but it seems that if it gets tied into politics, or your favorite ( or disfavored) economist, or your favorite website, then logic stops and you have to wonder: did this person say that because he is defending or attacking that person, what sources should I look for because I may be mislead. Why is that? Why should we have that kind of discussion??? Is it because it is a website, talk is cheap, electrons are cheap. Yeah, sure, its not an econ journal, and none of this is peer reviewed, but there should be data for this, and its pretty clear there are a whole bunch of variables that lead to results. The world is complex, and maybe that's the problem when talk is cheap.
Cliff, I see your point. I was looking at for a population of 100, Texas excludes 5 for Wisconsin's 2. You are right though in that texas has higher ELL than Wisconsin.
Rahul,
Being a substitute for something that is causative is good enough. No way this is Simpson's rule. Race has an impact on test scores for one reason or another, the effect is too large and consistent to question.
Yancey and Cliff, Apparently you like ideology. I like facts. Yes, the world is complex.
44% of Texas k-12 is hispanic; 8% Wis k-12 is hispanic and 50% is in Milwaukee school system.
Stop hounding him into making a concrete statement! He is only following in the proud tradition of Tyler of trying to insinuate clever things. If you interpret him incorrectly, it's only your fault that you're not clever enough.
All macs have copy and paste…unless they released something brand new without that functionality.
Bill,
I am not trying to bully you, but I won't let your bullshit go uncontested. You are constantly getting caught making inconsistent statements about, or making erroneous interpretations, of data you, yourself, bring to the discussion (like above on ELLs exclusions); and when it is explicitly pointed out to you, you refuse to admit that you made a mistake, or in this case, pretend that you didn't understand the objections requiring the objectors to make further descriptions of where you went wrong-Podunk, Cliff, and myself- then you trot out the "you are only interested in ideology rather than facts" complaint to which you are constantly resorting. This is a consistent pattern with you- if the "facts" you tout don't convince people, or those "facts" are shown to no mean what you claim, then post some more "facts" that you twist, and repeat. And when all else fails, pretend not to understand that you were wrong to begin with.
On the issue at hand- sure, there are a lot of unknowns in comparing Wisconsin to Texas, but would you not agree that the blog entry of Iowahawk was at least as relevant for comparison as the data Krugman was using, and leads to a different conclusion? Yes or no?
What clown would ever think that ELL rather than all students should be the denominator for the relevant fraction of excluded students? Bill is clearly right on that.
It is funny how people are trying to turn this into a "who are you rooting for" debate instead of a discussion of what is.
Iowahawk deserves a Pulitzer.
Krugman deserves unemployment.
However my breath is not held.
…the NAEP data reflect well for Texas compared to Wisconsin.
They make all 5 of the states that restrict collective bargaining look good compared to Wisconsin:
http://www.learnedhand.com/naep2.html
Next Comments →
Comments on this entry are closed.