Thursday, February 10, 2011

Poverty, Education and Causation.

Jonathan Livengood has posted an interesting post on poverty, education and causation that I encourage everyone to read. (Speaking of causation, I remember being told in GR causation is something that is assumed in the development of the theory. Of course, I wanted to know why we should make such an assumption leaving my GR teacher speechless.  I mean, it seems fishy to me if it must be assumed because then what if our assumption is wrong? Anyways, back to the post.)

I want to just show one of his plots:
There seems to be an inverse correlation between test scores and poverty.  So the question becomes: what is causing bad education?  I think many of our policies target helping teachers do a better job.  But could the cause of bad education have less to do with the quality of teaching and more to do with general poverty?  Or in Jonathan's own words:
Moreover, if it were true that Poverty causes Achievement, the policy implications would be clear and important. In order to improve math and science achievement in the U.S., we wouldn't need to focus too much on teachers, teachers' unions, charter schools, and so on. Rather, we would only need to find a way to decrease poverty. Simple.   Okay, maybe not so simple...
Again, I encourage everyone to read his post and for me it reinforced the importance of getting causation right.


  1. Did you ever get an answer about why causation was being assumed? Or even a clear idea of what "assuming causation" meant in that context?

    I have a sort of guess, but I'd really like to hear your thoughts on it.

  2. JS,

    Only 20 points separate poor vs well to do!

    I really hate when people take two data vectors, scatter plot, draw a regression line on bunch of labels and then call it a science; it should be called BS, just like the post by NN on "liberal" "conservative" etc.

  3. A Framework for Understanding Poverty, by Ruby K. Payne, was a book that gave me a different perspective on what poverty is. She defined poverty as "the extent to which an individual does without resources." The resources she lists are not just financial; she actually focuses more on emotional, mental, spiritual, and physical resources as well as support systems, relationships/role models, and knowledge of hidden rules (meaning the unspoken cues and habits of the middle or upper classes). (I'm copying these straight from page 16). I think it makes a lot more sense to look at poverty this way because it helps us understand why people are in poverty and what they have to overcome if they want to get out of it. Even if we gave lots of money to the poor kids in schools, they would probably still be unsuccessful in school because they also need the other resources in order to succeed. I have personally long felt that if we want to improve education, we need to start with the students' families, not with the teachers. But, that's easier said than done.

  4. Ancient1,

    A 20 point difference in raw score makes somewhere between a 15 and 20 point difference in percentile: 20 points moves a test-taker from 10th to 25th percentile or from 25th to 50th or from 50th to 75th or from 75th to 90th.

    Without knowing the distribution with respect to the scale, you just don't know how interesting a 20 point difference is.

    I was about to write something conciliatory about how I never said that what I was doing was science, but you know, I don't feel like being that generous. Your description is ignorant and arrogant.

    The questions being raised by Nick, me, and many, many social scientists are interesting, they matter for setting social policy, and they are susceptible to empirical investigation. We discuss empirical evidence that bears on our questions, and we recognize the limitations of the evidence we have with respect to the questions we ask.

  5. JL,

    Let us see: From the chart at ~11% poverty rates, Hawaiians are dumber by more than 20 points than North Dakotans. Must be the air in Hawaii that must make them dumb, it is much more warmer than North Dakota most of the time, and must affect the neurons of Hawaiians, may be native Hawaiians only.

    By the way, whose responsibility is it to present data unambiguously? Let me take it one step further. Most social (pseudo) scientists draw BS charts so that politicians go on using them justifying spending tons of money, especially if it benefits them directly or indirectly. Take a look at the on-going scandals at for profit universities that have popped up in last ten years or so. Now, you can get a degree if you pay them!

    Let me ask you to re-plot your data. Instead of poverty rate, use the average winter temperature and see what happens.

  6. I have re-plotted and analyzed data that includes temperature. The results are up in a new post on my blog:

    As I say in the post, it is not especially surprising that temperature is associated with educational achievement. But in any event, temperature does not screen off poverty and achievement. So, if the causal structure is that (increasing) poverty causes (decreasing) achievement, then the policy implications are really no different than they were before, though the pay-off would be slightly less, as I point out in the new analysis.

    Anyway, temperature was an interesting suggestion for a plausible common cause of poverty and achievement. Thanks for suggesting it.

    However, I still object strongly to your antagonistic tone and your ungracious, puerile swipes at social science. Surely you can make whatever points you have to make by concretely picking on methodology and by supplying or suggesting new data, observations, or experiments, without impugning the moral character of the researchers or lumping psychologists, economists, and sociologists in with astrologers and palm readers. If you cannot have at least that much professional courtesy, I don't see any reason to go on talking to you.

  7. JL,

    After posting my comment in full, I posted the following at your blog:

    Now a few more thoughts: what would a 13 years old do in nice warm climate? Take science tests? Play outside? Hang out? Chase girls?

    Statistical analysis allow one to arm wave and blow smoke in various form, and that is what you have done.

    When I saw yout first chart, it was obvious that the colder state kids had better test results then warmer state kids. See the problem with social scientists/ They plot stuff without any rhyme or reason then arm wave.

    By the way, our govt works the same way, they do statistical analysis without any rhyme and reason and screw all of us out of our money. If you don't believe me do the analysis on state and govt employee pension benefits vs. people employed in private businesses. While you are at it, do productivity analysis as well.

    I add the following:

    Most hard sciences employ statistics to overcome inherent limitations in measurements and instrumentation, not due to lack of theoratical basis of a relationship that is usually expressed mathematically. Give a few data vectors to social scientists and they are in hog heaven!

  8. Sorry to be getting back to the discussion late.


    Thank you for taking the time to do that informative post on temperature correlation's. (Who would have known!) And what is most impressive is you were willing to do some good work digging deeper into the situation.

    As for causation.... again I am not a trained philosopher so what I say may be garbage. Also, to all readers this is just some musings. I am not putting it forward as good science. I will quote from the Wikipedia article on the issue:

    "Humans have a practical interest in their surroundings, and tend to be resistant to the idea things "just happen." If one or more sheep die, humans will attempt to discover why. Learning what has killed the sheep is an important step in protecting the herd. The question can be phrased as, "What caused the sheep to die?" The answer may be "Because the wolves broke their necks," or "Eating too much clover caused them to bloat." These explanations assume the presence of an agent of some kind. In cases where an obvious cause is not discovered, humans may attribute the events to miracles or to evil supernatural agencies. There is a learned preference for some alternative to saying that something occurred without there being a reason for it. Anything that stands as 'uncaused' may motivate us to understand the salient events in their environment."

    Anyways, in relativity you assume things don't "just happen" and you dwell on the fact that since information cannot travel faster than the speed of light then all events are caused by other events in your past light cone.

    But while the professor was talking about this I just thought: Is every event Y *really* caused by information propagating from some other event X? If so then I guess since that information cannot travel faster then the speed of light you automatically know there must be some event X in your past light cone causing Y.

    But I was just musing if it is possible for some Y to happen without there being the need for some X causing it.

    Again, I am sure this is half if not 100% of all this gibberish and is why the professor looked at me like I was crazy, but still, I just wondered. But there is this entire article that gives my hope that it wasn't a completely retarded thing to wonder.

  9. Joseph,

    Denying the so-called law of causality -- that every event has a cause -- is not crazy at all! You might enjoy this paper by John Norton, which (among other things) presents a physical example you might like.

  10. JL,

    You have misrepresented Prof. Norton. I believe Prof. Norton is addressing knowability of cause-effect; denying causality due to your lack of knowledge of cause-effect is ignorance, and worse, believing in that ignorance. Read Unreality and Time, a monogram published by SUNY.

  11. JL,

    I had a few more minutes to re-read your work on temperature, which was obvious. even though, you admit that temperature is strongly correlated then so called state of poverty, you still cling to your original notion. do I call it "denial of causality" or "ignorance"?

    You see, temperature is easy to measure, the state of poverty is not. So, I would like you to include the error bars in temperature and state of poverty and you will see yourself arm waving.

  12. Ancient1,

    This will be my last response to you. It is broken up into more than one comment for reasons of space.

    I am not "clinging" to my original notion. I am following the data. I reported some statistical relationships and then raised questions about the causal structure. Those questions are motivated by the explanatory principle that associations ought to be explained by appeals to causal structure. That explanatory principle is given mathematical precision in the causal Markov and Faithfulness conditions (for which see Spirtes et al. 2000 or Pearl 2000).

    Anyway, I didn't start with an opinion about the relationship between poverty and educational outcomes. I started with the goal of looking into a claim made by some other people and graphically representing some data relevant to that claim. If Poverty and Achievement had not been associated, I would have said so. If Temperature had screened off Poverty and Achievement, I would have said so. I am not hiding anything about my process.

    I agree that there are problems with the data, and lots of people have managed to say so without attacking my character and without calling the whole project "pseudo-science". Poverty is operationalized one way by the Census Bureau, but I agree that it does not give us a full picture. I never claimed that it does. What I do claim is that the data at hand are interesting and partially informative.

  13. You say that social scientists are happy if you just give them some data vectors, but that is just not true. Had you suggested that I look at the relationship between educational outcomes and the number of left-handed professional baseball players living within a three state radius, I would have said that even if a relationship appeared in the statistics (doubtful), we would have very little reason to think that a causal relationship obtained because there is no plausible mechanism connecting the two.

    You say that social scientists have no mathematical theory for their research. I'm not quite sure what to make of that claim. In most cases, social scientists are doing exploratory work trying to generate mathematical representations of some phenomenon. How is the curve-fitting here any different in principle from the exploratory curve-fitting done by Kepler? Was Kepler a pseudo-scientist because he had no mathematical basis for his claims about the orbit of Mars?

    I don't know why you insist on attributing morally corrupt motives to me. I don't have a dog in this fight. I do not have a research grant for studying poverty and education. I am not even a social scientist myself. I am not a policy-maker. I do not work for a PAC. I am not a K-12 school teacher. I have a question about how the world works, and I am using my limited resources to get evidence that bears on my question.

  14. Moreover, I have never asserted any unconditional causal claims. Rather, I have been cautious to couch policy concerns conditionally -- as in the passage that Joseph quotes. I also make no claim that the work I'm presenting was difficult to do. Quite the contrary. Putting up that last post took about an hour, and most of that was getting the plots to look nice. Mostly, I wanted to raise interesting questions, not give definitive answers precisely because I am not in the right evidential position to give definitive answers. If you had read carefully with even a little bit of charity, instead of pre-judging it as BS social pseudo-science, you would have seen that.

    Finally, you couldn't be more wrong about Norton. He says explicitly that he is not giving an epistemic, Humean skeptical argument. On the first page of his article, he writes, "In the negative thesis I urge that the concepts of cause and effect are not the fundamental concepts of our science and that science is not governed by a law or principle of causality." He goes on to write, "This form of causal skepticism is not the traditional Humean or positivistic variety. It is not motivated by an austere epistemology that balks at any inference to metaphysics. It is motivated by taking the content of our mature scientific theories seriously." And in Section 3, where Norton gives the physical example I alluded to, he writes, "Even quite simple Newtonian systems can harbor uncaused events and ones for which the theory cannot even supply probabilities. Because of such systems, ordinary Newtonian mechanics cannot license a principle or law of causality."

    Anyway, I have already wasted far too much time engaging your abuse. I'm done.

  15. JL,

    Here is how Prof. Norton begins:

    My purpose in this paper is to take issue with this view of causation as the underlying principle of all natural processes. I have a negative and a positive thesis.

    Regarding Kepler, it took a while to accept his theories even after he had formulated laws that were applicable to not just Mars but all other planets and beyond.

    If I can introduce temperature as more strongly correlated variable by just a quick look at your presentation, I think I have contributed to your research without charging anything.

    Instead of blowing lots of hot air toward me, why don't you figure out how would change your research approach on the grant monies you have left... Just a thought: may be sciences should be taught in winter months in warmer states...

    I will insist that after all the analysis including Kepler's curve fittings, there is a mathematical result - even a lousy look up table - that applies broadly. You have not done that.

    Finally, I suggest that you be little less emotional and stop getting triggered by little words like "arm waving".

    I hope

  16. Great post! I wish you could follow up on this topic!

    online pharmacy

  17. found your site on today and really liked it.. i bookmarked it and will be back to check it out some more later

    online pharmacy

  18. Plain and simple! I like your work!

    generic nolvadex


To add a link to text:
<a href="URL">Text</a>