« Ahmadinejad's rough week | Main | 300 »

March 11, 2007

Comments

Jesurgislac: I've pointed out several other criticisms of the study; I haven't limited myself to the criticism that the study unjustifiably claims that Bush is the "first" (although this should be a good clue to the discerning reader that mischief is afoot, as it proves that the study's authors are either too stupid to know what the word "first" means or too dishonest to care).

In any event, other correct criticisms of the study are: Doesn't specify its methodology for ascertaining how many investigations are in existence; counts investigations of Democrats that actually began under Clinton; apparently counted news stories that mention who was subpoenaed, but doesn't even consider the fact that witnesses are often subpoenaed (as opposed to only targets); omits many investigations of Republican officials.

On top of that, the study makes absolutely no attempt to discover whether 1) the investigations of Democrats were justified, or 2) any Republican instances of corruption were left uninvestigated. After all, if US Attorneys investigate all significant local corruption cases that involve violations of federal law, and it just happens that most local cases involve Democratic officials, then a disproportionate number of Democratic investigations is precisely what you'd expect. Indeed, if local Democrats do take bribes more often, etc., the real bias would be if US Attorneys tried to rig the numbers so that they investigated an equal number of Republicans and Democrats.

So I took a few names at random from the list of local officials and tried to Google them.

First was Abe Beltran of Colton, California. Oops, turns out that he pleaded guilty to accepting bribes. And it turns out that that link shows that two more people on the list (Karl Gaytan and Jim Grimsby of Coltron, California) either pleaded guilty or were found guilty of accepting bribes. Maybe that wasn't a witchhunt after all.

Next one I googled: Xochilt Ruvalcaba of Southgate, California. One document shows that the investigation here began with the LA district attorney (and moved to the FBI only later). An LA Times story describes some of what Ruvalcaba and the other people from Southgate, California (also on the list) did:

South Gate political leaders thrown out of office in a recall election last month drained the city's reserve in a weeklong, $2.1-million spending spree that didn't stop until hours before the new council was sworn in. tarting on the day they were overwhelmingly rejected by voters in the Jan. 28 recall election, former Mayor Xochilt Ruvalcaba and Treasurer Albert Robles signed the checks, with most of the money earmarked for attorneys the ousted officials had hired. . . . The revelations form a closing chapter to the former leaders' turbulent two-year reign. During that time, the southeast Los Angeles County city endured corruption investigations, civic unrest and unruly council meetings that were topped off this month when Ruvalcaba slugged a council member in front of hundreds of residents. . . . The next day, incoming Vice Mayor Henry Gonzalez ordered stop payments on $500,000, but he said the payment of millions of dollars to attorneys has left the city on the verge of bankruptcy. This week, city officials, citing the budget crunch, announced the layoffs of 60 employees and said there would be more to come. . . . "In some ways we still don't know the myriad of ways that they really defrauded the public," said Jaime Regalado, executive director of the Edmund G. "Pat" Brown Institute of Public Affairs at Cal State Los Angeles. "It's hard to believe anybody could be so brazen."
Hmmm, maybe that wasn't a witchhunt either.

Anyway, that's all I have time for. Feel free to use Google yourself (www.google.com) to try to figure out what went on in the rest of the cases. As for the flip side of the equation, you might also try to dig up cases of local officials who were Republican, who broke federal laws in a similar manner, but who weren't investigated federally.

Without that kind of factual investigation, neither you nor the study's authors have any idea whether the Bush administration has been unfairly targeting local Democratic officials. Do you admit that much?

Hilzoy, in my post at Stubborn Facts, I deliberately left the criticism of the "study" cited by Krugman about the "first administration" claim to last. It's the most minor point. As I've already pointed out, the study is deeply flawed for the purpose for which Krugman cited it. It simply is not reliable about its fundamental point. That's the big deal. An hour of fact-checking found SIX data points that the authors omitted, a 10% error. More time would almost certainly have found more omissions, and the other biases Tully and I have noted provide still more reasons to disregard the study as worthless.

PatHMV -- it's darned odd that even after those flaws have been repeatedly pointed out (both here and through links to more comprehensive essays elsewhere), Hilzoy's 12:40 post 1) takes it for granted that there is a problem with the overall "numbers," and 2) assumes that the only flaw alleged with the study is whether it claimed that the Bush administration was "first" to engage in political witchhunts. Darned odd, I say.

Pat,

First Tully unexpectedly turns up, and now you. Things should be a lot more fun around here. I know you both from elsewhere, pre-Stubborn Facts. My id is secret for now for the reason noted upthread.

PS: Bernard - so you were up in arms when Clinton et al fired all those US attorneys en masse?

livermoron,

No. I was not. I quote nell:

How many times does someone in this thread have to explain that U.S. Attorneys are political appointees, that it is absolutely routine for an incoming administration to replace them, whether en masse or in a leisurely fashion?

Still, I'm sure you'll get credit for memorizing and repeating the talking point correctly.

I did not take it for granted that there was a problem with the numbers. I did take it for granted that it had not established that Bush was the first person to do this. I then noted that that was not its point.

I will also add: it would take a lot of errors to get around a 7:1 disparity.

Clarification: You seem to take it for granted, and indeed still do, that the numbers are wrong, i.e., 1) that this study really does show a 7:1 disparity (when it does no such thing), and 2) that any disparity that does exist is problematic (which may or may not be true, depending on facts of which you seem to be entirely ignorant).

The (reported) 7:1 number applies only to local officials. The overall ratio is closer to 4:1. I make no comment on the accuracy of those numbers, only that those are the numbers put forth by the study.

I'm not sure how useful the study is. It seems that the facts of the individual cases, of which I have very little knowledge, are what matter.

"I will also add: it would take a lot of errors to get around a 7:1 disparity."

Or a lot of crooked Democrats. Too bad they didn't bother to find out which it was.

Hilzoy, you're really being careless with that 7:1 figure. As hairshirthedonist points out, that's only the local officials, when state officials are omitted. It would be an odd sort of bias to show up against the lower-level politicos but not the higher, state-level ones. When you look at ALL elected officials and candidates, or at least all the ones found by the study, and the ratio is 4.45 to 1. Add in the 6 Republicans and 1 Democrat I found omitted by the study (with just one hour of work), and the ratio falls to 4.09 to 1.

You continue to assume that the numbers of the study have any validity at all, despite the fact that I have demonstrated both specific omissions and the potential for systemic bias. When the basis for the raw data is destroyed, the numbers become meaningless. There's nothing to "get around."

As hairshirthedonist points out, that's only the local officials, when state officials are omitted.

oh fer chrissakes, Krugman himself makes that point. read the last sentence of Hilzoy what quoted.

Hilzoy: I fail to see why this is a big deal.

it's all they've got.

My suggestion would be to forget about the study. If the study were never done, would there be nothing to talk about? I don't think it's this study that has made the firings an issue.

I also have a question. What does this all mean in the end? If there's nothing illegal about the allegations being made about the admins conduct, and assuming the allegations are all true, what then? Is it simply another reason to dislike the Bush admin, perhaps to the point to think that it may be one of the worst ever, or is there something more? I haven't read a clear statement on the implications of all of this.

I agree with all the skeptics of the study - there cannot possibly that few corrupt Republican officials in this country.

jesurgislac, Krugman's become a hack. I won't waste the off-topic electrons here showing it, but he is. He demonstrates hackdom just by citing a study he has to know is bogus. If he's not aware the study's junk, then I take back anything nice I said about his intellect--he's either stupid, or a hack, or a stupid hack. Pick 'em. You wanna worship him as a fellow traveler, go ahead. I don't kneel in that church, not for either party.

Tully, who are you to talk about junk social science?

Someone with extensive experience in designing and analyzing valid studies who knows blatant junk when he sees it. Feel free to attempt to defend the study on its academic merits--it has none. QED.

Krugman cites evidence that USAttorneys investigate seven times as many Democrats as Republicans as evidence that something might be going on. But the study is not his main point, nor is it mine in citing him.

Hilzoy, if his "evidence" is so blatantly bogus, how does it support his point? It's not evidence of anything but the study author's [a] incompetence, or [b] intent to decieve, or [c] both. Without his "evidence" what is his point, anyway? What else does he offer that we know for fact? Stripped of semantics and reduced to objective statements?

As I pointed out upthread, the numbers in the study for federal & statewide officials show NO disparity from random selection, even though they were apparently hand-picked. The local numbers show 7 to 1, but we already know the "research" itself is horribly flawed at best, in data acquisition, methodology, selection bias, etc. At all levels--I picked out bad data by eye without ever hitting Google at all, just glancing through the data tables. Before I even got to the (complete lack of) required methodology descriptions and such. Were I a peer-review reader and that crossed my slushpile, it would've been immediately roundfiled as utter junk.

As I noted upthread, even starting with the assumption that Democrats and Republicans are equally inclined to be corrupt, you could still easily end up federally investigating and prosecuting more Dems than Reps without any partisan selection bias by USA's at all, simply based on the demographic distribution of locally elected Dems and Reps in relation to metro size and proximity to USA's offices.

Meaning we know absolutely nothing from the "study" of any relevance at all to the subject of the firings, as the study is completely worthless from any empirical standpoint. The firings themselves are certainly worthy of discussion, but the study is junk and Krugman puts himself on record with it for some guilt by association. Far as I'm concerned that makes his opinion worth the same as any partisan wing-ranter's, left or right, namely zilch. Nothin' but emo-angst dump. If he had any relevant and verifiable facts in his pay-per-view column, I'm open to hearing them.

Tully,

"As I noted upthread, even starting with the assumption that Democrats and Republicans are equally inclined to be corrupt, you could still easily end up federally investigating and prosecuting more Dems than Reps without any partisan selection bias by USA's at all, simply based on the demographic distribution of locally elected Dems and Reps in relation to metro size and proximity to USA's offices."

Where to begin?

a. US attorneys as a group cover the entirety of every state -- there is no place which is not subject to the jurisdiction of a US Attorney. Your objection based upon the "metro size and proximity to the US Attorney's office" is just silly. Are you seriously suggesting that if there were criminal events going on in Las Cruses, NM and Inglesias decided it was too far from Albequerque to investigate, that would be normal and expected conduct for a US Attorney?

b. The fact that Democrats tend to cluster in large metro areas means there are far more, not less, Republican local officials than Democrats, if each municipality has roughly the same number of elected officials. Certainly a city of 100,000 would not have 100 times the number of elected officials as a town of 1,000.

c. Ultimately, any discussion of this study is a distraction from the real issue here -- this Administration wanted the US Attorneys to ignore their proper role, and instead prosecute innocent Democrats and protect guilty Republicans. To their credit, at least some US Attorneys did not, and were fired for it.

Dantheman, Tully and I are both well aware of that no part of the U.S. is outside the jurisdiction of some U.S. Attorney. That's not the point. Where the prosecutor's office is physically located does indeed impact what they investigate. Even the best investigations are sometimes only sparked by gossip. There's not as much personal interaction with folks in the hinterland, so the FBI and the prosecutors are less likely to have sources and snitches tipping them off to improper behavior.

More noticeably, small town embezzlement and corruption tends to involve dollar amounts measured in the thousands, not millions. A U.S. Attorney is unlikely to bring the full resources of the FBI to bear on a crime involving $10,000, while a crime involving $1 million will certainly get his attention. It's primarily the large urban (mostly Democratic) areas which have that kind of money at stake.

The critique of the "study" is highly relevant, because Krugman charged that it was a "bigger scandal" than the 8 fired prosecutors. Krugman claimed that the STUDY was the "real issue" here, not the 8 individual prosecutors. So if you think the study is a distraction, blame Krugman and Hilzoy for bringing it up.

Also, note that even according to the "study," Democratic local officials outnumber Republican ones by a noticeable amount.

First of all, I have yet to see argument against Krugman's cited study that isn't bull****. I'm waiting for someone to correct me.

We have Tully pointing out that a 7:1 ratio of investigated/indicted Democrats to indicted Republicans doesn't prove that the Bush administration was witch hunting. Fine. But that doesn't make the study inaccurate. The purpose of the study was to determine the ratio of investigated Republicans to investigated Democrats. Period. It did so. Period. It doesn't prove anything about anyone's motivation or related unethical/illegal behavior. Not the purpose.

Then we have Tully not liking that there was no published metholodogy. So? So you think you have objectively disproved the conclusions, because you don't know how they did it? Further more, I see the Chi-square calculations right there in Appendix D. What more methodology, exactly, do you want?

Lastly, we have unfounded allegations as follows: but we already know the "research" itself is horribly flawed at best, in data acquisition, methodology, selection bias, etc.. Excuse me? You know this how? Would you like to come up with some examples, please? Or should I just assume you're making it up out of literally thin air?

Or are all three of those allegations related to PatHMV's unintentionally hilarious declaration that she's already found six Republican officials indicted who weren't mentioned in the survey? PatHMV, are you bleeding kidding me? Do you know what a sample is? Do you know what samples do - survey a random grouping of individuals in the population, rather than surveying literally every person in it? You could find 1000 officials that weren't in the sample. That's why it's a sample. It's not a systematic compilation of every indicted official in the country. For Pete's sake.

So, if anyone comes up with a legitimate, evidenced reason why the sample demonstrating - period - a 7:1 ratio of investigated/indicted local Democrats to Republican officials is factually incorrect - we're still waiting. In the meantime, it seems to be rightwing pushback (i.e., serial dishonesty) period. Although, to give Tully the benefit of the doubt, he might just be a psychological contrarian with a serious problem sticking to the subject - which is not whether or not you like the survey, but whether you have any specific evidence that its conclusions are factually incorrect.


Second, this survey is a bogus red herring. Subtract it out, and it makes not an iota of difference to the meat already on the table - Gonzales firing eight US attorneys who refused to indict - frame - innocent Democratic politicans solely because indictments make Democrats look bad.

The survey stuff is just icing. What we have on the table is, at minimum, a huge scandal, and depending on what we find about Menendez's bogus indictment last year, potentially Watergate-level infamy.

"Gonzales firing eight US attorneys who refused to indict - frame - innocent Democratic politicans solely because indictments make Democrats look bad."

Um, you know that they're innocent? Really? How do you know this?

Glasnost's comment is particularly stupid, as is shown by the contrast between these two statements:

Do you know what a sample is? Do you know what samples do - survey a random grouping of individuals in the population, rather than surveying literally every person in it? You could find 1000 officials that weren't in the sample. That's why it's a sample. It's not a systematic compilation of every indicted official in the country.

And:

Then we have Tully not liking that there was no published metholodogy. So? So you think you have objectively disproved the conclusions, because you don't know how they did it?

Anyone who has even the slightest familiarity with social science will perceive the stupidity here: If the study is supposed to be based on a random sample, then it's pretty dang important what the methodology was. Specifically, how do you get a random sample of "local-officials-indicted-or-possibly-subpoenaed-as-a-witness-in-corruption-investigations"? The authors of the study are clearly dishonest -- as shown by their claim to have proven that the Bush administration is the "first" to do [something] even without looking at any pre-Bush evidence. Why the heck should anyone trust that they picked a random sample (even if such a thing is possible here)? If they weren't trying to create a list of all corruption investigations, then who says they weren't just cherry-picking the sample?

Brett: Um, you know that they're innocent? Really? How do you know this?

Well, all of them are legitimately entitled to be presumed innocent until proven guilty, yes?

And, it does seem likely that when a Republican-appointee (a Bush-appointee) prosecutor declines to indict a Democratic politician, it means the case against the politician is shaky to non-existent. Especially, when the attorney's job is on the line and it has been made clear to them: indict or get fired. If there was any kind of case, surely these Bush appointees would have indicted.

Brett, you have come across as reasonably fair-minded, so far. This comment rather spoils that effect.

Um, Glasnot, did you bother to read my post at Stubborn Facts? Or for that matter the study itself? Where do you find a claim by the authors that they were studying a sample, rather than the entire population? The heading on their data source table reads: 375 Candidates & Elected Public Officials (All Offices) Investigated by the Bush Justice Department Identified by Political Party Affiliation 2001 thru 2006.

No reference to "Sample" there. And any randomized study ABSOLUTELY requires a disclosure of the methodology used to select the sample. Did they pick at random the names of 375 elected officials, and then check to see whether they were investigated or not? Did they compile a list of all "investigated" officials and then random select from that population? Did they pick a random assortment of 10,000 officials (and candidates!) and then discard them until they had 375 who had been "investigated"? No, it's clear from their study that their figures claim to be ALL the elected officials investigated by the Bush Administration.

As for sticking to the subject, the subject of this post by Hilzoy was this study dug up and touted by Paul Krugman as the "bigger" problem. And the study is bogus. If you don't want to see that, fine. But first back up your assertions that the authors were making a "random sample" and tell me how and from where they drew that random sample.

As for sticking to the subject, the subject of this post by Hilzoy was this study dug up and touted by Paul Krugman as the "bigger" problem.

Well, actually, no...but carry on.

As for sticking to the subject, the subject of this post by Hilzoy was this study dug up and touted by Paul Krugman as the "bigger" problem.

err... no. that's what y'all keep saying it should be. but, it's not.

though... maybe if you can prove Jamil Hussein wrote the study you're so concerned about, you could prove that not only is every A.O.K. with the D.O.J. but with Iraq, too.

Sorry, I must have gotten confused when Hilzoy wrote: "Paul Krugman raises another crucial point:" Pardon me for focusing on just one of the "crucial points" made in the post.

Sorry, I must have gotten confused when Hilzoy wrote: "Paul Krugman raises another crucial point:"

Yes, you did get confused. Or, at least, your writing is confused.

Usually, when someone uses the term "another", it's used as a butressing point for a larger, overall point. In no way can that be confused as the subject of the post, particularly since the "point" is introduced in the last third of the post.

Such sloppy construction sorta implies that the underlying rhetoric structure is equally sloppy, and I don't think you want us to think that.

Well, I tend to give relatively equal weight to "crucial" points. If they were of lesser importance, not really needed, than they wouldn't be "crucial."

This is a problem with motive speculation when there isn't strong evidence--if you speculate on a motive that turns out to not be very strong, it distracts from an attempt to point out the initial problem.

It is odd that Bush chose to fire these people in mid-term.

That is worth investigation.

After reading through the thread, I'm not at all convinced that the reason these people were fired was because they weren't sufficiently investigating Democrats, nor am I at all convinced that Democrats are being particularly unfairly investigated (the idea that Democrats tend to have power in big cities and thus when corrupt [and whatever the 'natural' rate of corruption is] have the ability to make make a bigger splash on the local corruption level than corrupt Republicans who tend to be important in smaller towns amkes a good deal of sense).

That shouldn't detract attention from the fact that the firings are odd and wanting of an explanation.

Just because this motive isn't as explanatory as it seemed on the first pass doesn't mean that the action of firing these people is from a pure motive.

Sorry, I must have gotten confused when Hilzoy wrote: "Paul Krugman raises another crucial point:"

Uh, the crucial point wasn't the study, but, if these 8 were pressured to put the heat on democrats, refused and were fired for it, isn't it possible that other U.S. Attys were pressured and didn't refuse to put the heat on democrats? He then points to the study as "evidence" (which, obviously doesn't prove anything) that such a hypothesis is true.

SH, sure looks bad to me.

I have to agree with Sebastian, for the most part; I don't understand this focus on almost insisting that the cause of the firings must have been, or likely was, in each case, insufficient willingness to prosecute Democrats.

There's only specific evidence indicating that in one case, that of McKay of Seattle; in the other cases, it so far seems likely that there were other (political) motivations for pushing out the Attorneys.

This fact in no way indicates any lack of malfeasance on the part of the DOJ leadership, or the White House, so why we're winding up with people focusing on "if these 8 were pressured to put the heat on democrats" when there's no evidence at all to support that point (it could be true, but for now, 1 =/ 8), I don't know.

Besides, anyone who thinks that Gonzalez and his minions would limit themselves to only one political reason for pushing out U.S. Attorneys they'd like to replace to be able to give the slot to another up-and-comer, is being deeply unimaginative, IMO. :-)

Or better, JMM's central post.

I'm not at all convinced that the reason these people were fired was because they weren't sufficiently investigating Democrats

Sebastian,

At Volokh, anderson provides this link to a story in the Seattle Times about one of the fired USA's. It sounds convincing to me, but maybe my biases interfere. What do you think?

Gary- I think Iglesias in New Mexico fits in that list as well.

oops. rilkefan beat me to the punch.

PatHMV:

Where do you find a claim by the authors that they were studying a sample, rather than the entire population?

Right here

We compare political profiling to racial profiling by presenting the results (January 2001 through December 2006) of the U.S. Attorneys' federal investigation and/or indictment of 375 elected officials. The distribution of party affiliation of the sample is compared to the available normative data.

Fish in a flipping barrel.

I await your apology and retraction, Pat.

As for the methodology of selecting the random sample, I tell you what, I wouldn't mind finding that out myself. Perhaps someone should try emailing the authors and politely asking them for it, eh?

I read your website, Pat. It's a collection of speculation and gripes. None of the others, to your credit, are as wildly off the mark as this one, but to talk around here as if you've somehow "debunked" the study is ludicrous. What you've done is listed a bunch of complaints about what you think the study should have done that in no way disprove the conclusions of the data it presents.

I don't know if the random sampling they took might have been in some way biased. You don't know either. So, because I can't prove that the sample was bias free, you think you can declare it 'junk science?'. Give me a bleeding break. You don't know the methodology of newspaper presidential polls. And even if you knew the announced metholodogy, you don't know if they're lyyyyying. Unless you can come up with some evidence that the sample was biased, you've.... got..... nothing.

And Ugh beats me to Iglesias.

Gary, you're (a) being too literal and (b) missing D.N.M. WRT (a), there's no functional difference between pressure to indict Dems and pressure not to indict Reps, and so that a firing falls into the latter category doesn't really say anything positive for the Admin., nor does it really refute the 'one true motive' thesis.

That said, I think you'd have to call the story developing. People who think that the particular motive of one-sided corruption lies behind these firings will likely be proven right, at the end of the day: this is why the government keeps changing stories. More simply: they're covering up because they did something wrong.

In point of fact, as has been much reported, Harriet Miers initially inquired of Gonzalez as to whether all 90+ U. S. Attorneys could be fired and replaced at the beginning of Bush's second term; Rove said it wasn't practical, and suggested a smaller list.

This, along with the other reasons the fired Attorneys were apparently fired for, would seem to clearly demonstrate that there were a number of reasons these firings were engaged in; not just the single reason of not prosecuting Democrats.

(I should have mentioned Iglesias, as well as McKay, though.)

Carol Lam, for instance, prosecuted Duke Cunningham. Paul K. Charlton investigated Republicans Jim Kolbe and Rick Renzi. Bud Cummins investigated Governor Roy Blunt. Daniel C. Bogden investigated Governor Jim Gibbons.

It seems that investigating Republicans was as bad for your job health as not investigating Democrats with no grounds.

Margaret Chiara, on the other hand, didn't ask for the death penalty, and Charlton also had a fight with Washington about it. Coincidence?

"missing D.N.M."

?

"Gary- I think Iglesias in New Mexico fits in that list as well."

Yeah, I already posted about that, before your comment crossed mine.

Sorry, and multiply pwned at that.

"You don't know the methodology of newspaper presidential polls."

Well, so long as one avoids reading about them; the methods are publically available on the web, as a rule, and even (usually) linked to from the newspaper's site.

signed,
guy who has done a lot of polling

"People who think that the particular motive of one-sided corruption lies behind these firings will likely be proven right, at the end of the day"

Well, of course; I'm not sure why this is addressed to me.

Pointing out that the corruption is almost certainly -- all the indications so far point to it -- for far more widespread reasons than the single one of not prosecuting enough Democrats, is the opposite of saying it isn't corruption!

Gary, I don't think anyone at all is saying that 'failure to indict Dems' is the one true motive. Refutation of this is beating a straw man. One sided corruption is the proffered motive, and 'failure to indict Dems' is sometimes used as a shorthand.

"Gary, I don't think anyone at all is saying that 'failure to indict Dems' is the one true motive."

I'm perfectly willing to believe that that's not what people are thinking, but it's what some are saying.

Example: Dantheman, 3:05 p.m.: "Ultimately, any discussion of this study is a distraction from the real issue here -- this Administration wanted the US Attorneys to ignore their proper role, and instead prosecute innocent Democrats and protect guilty Republicans. To their credit, at least some US Attorneys did not, and were fired for it."

I agree with the first phrase, but the rest clearly states only that sole reason for the firings; I think that's myopic and almost certainly wrong.

If it's just "shorthand," I don't think it's helpful for people to be arguing in a way that leaves out all the other motivations and reasons for the purge, when that spread of corrupt reasons is clearly relevant. I think that "shorthand" is misleading, and will lead to problems when someone sucessfully points out that it fails as an explanation in some cases. That's an argument no one should be bothering with, let alone setting themselves up for. IMO.

That's all.

Yes, Glasnost, I saw that word the first time I read the study. It does not say "random sample." In context, it can easily be read to be referring simply to the data they studied. Had they been referring to a statistical sample there, they would have said, "we sampled X number of elected officials, whose overall partisan make-up reflected the nationally known partisan distribution of ___%.

As for e-mailing the authors, you are welcome to. It's obvious to me from the language they use and the manner in which they have "released" the study that they aim at furthering a partisan agenda, not exploring real facts. In political science and sociological research, indeed in any rigorous academic research, it is the responsibility of the researcher to discuss these issues upfront, not the responsibility of the reader to assume things or inquire further as to their fundamental methodology.

will lead to problems

Only if people are excessively literal. A decided danger given all our new friends . . .

Well, so long as one avoids reading about them; the methods are publically available on the web, as a rule, and even (usually) linked to from the newspaper's site.

Okay. Perhaps a bridge too far. But you basically take the newspaper's word on faith that the methodology they actually used is the methodology that they say they used. You don't ask to see the source code for the software that was used to randomly dial.

The metaphor here is that not knowing the sample selection methodology when looking at a study done by university professors with no obvious reason to be dishonest, and who have obvious competent backgrounds in their field, may be something you'd like to know, but not knowing it doesn't make the study "junk science". It's ridiculous, spurious, and politically motivated to say so. You don't even know that that methodology hasn't been made available to others. It simply isn't on the link for epluribusmedia's web presentation.

Just coming up with reasons off the top of your head why some study somewhere might have a flawed sample and then saying "so this study is liiiiies" is a joke. You haven't proven a thing. You've made accusations that have neither been refuted nor vindicated. Making accusations is really easy.

Of course, someone could argue in the reverse vein, that just because the Bush DoJ has investigated 7 times as many local Democrats as Republicans, isn't proof that they did it for partisan reasons. By itself.

No, for proof of that we need to turn to the eight DoJ firings that just happened...

Yes, Glasnost, I saw that word the first time I read the study. It does not say "random sample." In context, it can easily be read to be referring simply to the data they studied. Had they been referring to a statistical sample there, they would have said, "we sampled X number of elected officials, whose overall partisan make-up reflected the nationally known partisan distribution of ___%.

Pat...
My anger is almost spent. I'm trying not to be more nasty than neccesary here. But you really need to slow down before you say foolish things.

We compare political profiling to racial profiling by presenting the results (January 2001 through December 2006) of the U.S. Attorneys' federal investigation and/or indictment of 375 elected officials. The distribution of party affiliation of the sample is compared to the available normative data (50% Dem, 41% GOP, and 9% Ind.).

You found "missing" investigations and claimed that it somehow debunked the study.

I pointed out the obvious fact that it was a sample, not a population.

You asked me where I got the idea that it was a sample. I quoted the weblink saying just that.

Now you say, if it was really a sample, they would have given the total N (#) of elected officials, and stated that it matched the national distribu... oh wait, they did that too.

It makes me wonder - is this really Pat? Is this someone spoofing him? Making his arguments look even more ridiculous than they really are?

It's obvious to me from the language they use and the manner in which they have "released" the study that they aim at furthering a partisan agenda, not exploring real facts. In political science and sociological research, indeed in any rigorous academic research, it is the responsibility of the researcher to discuss these issues upfront, not the responsibility of the reader to assume things or inquire further as to their fundamental methodology.

Oh, I see. Because they released their findings in a manner that suggests they don't like President Bush, they must be dishonest liars. Thanks for clearing that right up. Additional snark here self-edited.

Well, I give up. You've got your mind made up, glasnost, and the facts matter not. They did not "do that too." Their statement does not say that the sample is "comparable" to the normative data. They compared the investigations they identified to the national sample. The sentence about the 50% to 41% normative make-up is designed to establish that there is not a 7 to 1 disparity between Democratic and Republican officials. You are reading the words but not understanding what they are saying.

If you're so convinced they did a random sample, why not e-mail them yourself and ask them? Or just point me to a source telling me what the TOTAL number of all political officials and candidates under investigation in this country. The number 375 strikes me as in the right order of magnitude.

The metaphor here is that not knowing the sample selection methodology when looking at a study done by university professors with no obvious reason to be dishonest, and who have obvious competent backgrounds in their field,

1. They are professors of communications. This makes them competent to study federal prosecutions . . . how?

2. No obvious reason to be dishonest: Well, we've already established beyond a shadow of a doubt that these particular professors are willing to lie about their results (i.e., by claiming that they've proved that Bush is the "first" to politicize prosecutions, when they didn't even study anything pre-Bush). If they lie through their teeth about one thing, I chalk them up as untrustworthy. Why be so gullible?

As for the methodology of selecting the random sample, I tell you what, I wouldn't mind finding that out myself.

Correction: The authors never say "random" sample. Why are you defending the authors by making a claim that even THEY are not dishonest enough to make?

We compare political profiling to racial profiling by presenting the results (January 2001 through December 2006) of the U.S. Attorneys' federal investigation and/or indictment of 375 elected officials. The distribution of party affiliation of the sample is compared to the available normative data (50% Dem, 41% GOP, and 9% Ind.).

You seem to be reading this to mean: "We took some sort of sample that matched the political demographics of the overall population." It obviously means nothing of the sort. The authors say that they're comparing the 375 people under investigation to the 50/41 split. It doesn't even make sense to claim, as you do, that the authors collected the sample of 375 by matching the 50-Dem/41-Repub split -- the whole freaking point of this bogus "study" is that the 375 don't match the 50-Dem/41-Repub split.

Instead, the comparison to which the authors are referring occurs in (for example) Table 2 here, where they compare the number of Democrats and Republicans under investigation to the 50/41 split in the population of local officials.

Obviously, this comparison of the final results does absolutely nothing to make the original sample random.

In short, you have absolutely zero support for claiming that the sample was randomly selected. Given that the authors themselves, dishonest as they can be, don't even pretend that the sample was random, why should anyone believe your desperate and post hoc scramble?

The metaphor here is that not knowing the sample selection methodology when looking at a study done by university professors with no obvious reason to be dishonest, and who have obvious competent backgrounds in their field,

1. They are professors of communications. This makes them competent to study federal prosecutions . . . how?

2. No obvious reason to be dishonest: Well, we've already established beyond a shadow of a doubt that these particular professors are willing to lie about their results (i.e., by claiming that they've proved that Bush is the "first" to politicize prosecutions, when they didn't even study anything pre-Bush). If they lie through their teeth about one thing, I chalk them up as untrustworthy. Why be so gullible?

As for the methodology of selecting the random sample, I tell you what, I wouldn't mind finding that out myself.

Correction: The authors never say "random" sample. Why are you defending the authors by making a claim that even THEY are not dishonest enough to make?

We compare political profiling to racial profiling by presenting the results (January 2001 through December 2006) of the U.S. Attorneys' federal investigation and/or indictment of 375 elected officials. The distribution of party affiliation of the sample is compared to the available normative data (50% Dem, 41% GOP, and 9% Ind.).

You seem to be reading this to mean: "We took some sort of sample that matched the political demographics of the overall population." It obviously means nothing of the sort. The authors say that they're comparing the 375 people under investigation to the 50/41 split. It doesn't even make sense to claim, as you do, that the authors collected the sample of 375 by matching the 50-Dem/41-Repub split -- the whole freaking point of this bogus "study" is that the 375 don't match the 50-Dem/41-Repub split.

Instead, the comparison to which the authors are referring occurs in (for example) Table 2 here, where they compare the number of Democrats and Republicans under investigation to the 50/41 split in the population of local officials.

Obviously, this comparison of the final results does absolutely nothing to make the original sample random.

In short, you have absolutely zero support for claiming that the sample was randomly selected. Given that the authors themselves, dishonest as they can be, don't even pretend that the sample was random, why should anyone believe your desperate and post hoc scramble?

"the subject of this post by Hilzoy was this study dug up and touted by Paul Krugman as the "bigger" problem."

No. This was not the subject of this post. This was the subject of Tom M's Instapost linking here, but it was not the subject of my post.

The subject of my post was the attorney firing scandal itself. The 'crucial issue' Krugman brought up was not the study; it was the question: since we know that there were US Attorneys who were pressured and then fired for not yielding, it stands to reason that there were US Attorneys who were pressured and yielded and were not fired. That is: that some US Attorneys carried out politically motivated investigations and/or prosecutions under pressure.

Please do not attribute to me points I did not make.

1. They are professors of communications. This makes them competent to study federal prosecutions . . . how?

Quite a bit, if you're familiar with academic studies. That is, if you're honest about it.

I tell you what, Pat and John. You're right: I don't know that the sample was a random sample. Yeah. I'm assuming that the authors, whatever method they used, didn't deliberately skip Republicans when they went looking. On the other hand, it's definitely represented up front as a sample, and not a comprehensive list, so, as said from the beginning, finding 8 officials not listed doesn't have a whole lot of relevance, if you're looking at a sample. All that matters is that, whatever method they used, Republicans weren't skipped.

The local official subtotal of investigated/indicted officials is 262 to 037. The p value for this is less than .0001. It only has to be .05 to be statistically significant. In other words, that ratio would have to be unrepresentative of the population by *massive* amounts for the study to be substantively wrong.

So, you've found 7 Republicans and 1 Democrat that weren't listed? Do you think that there are, oh, 200 more investigated Republicans, waiting to be discovered, that the study's authors somehow skipped? If you found 200, and there were not another 200 Democratic investigations also then the study's conclusions would be substantively wrong. And then, and only then, do you have a case for arguing that the study's authors had any reason at all to act dishonestly. And you sure don't have, yet again, any evidence that their sampling methods were improper.

1. They are professors of communications. This makes them competent to study federal prosecutions . . . how?

Um, professors of communications - statistical surveys are a pretty large part of communications - their chi-square calculations are posted, and how many in 100 random Americans know how to do those? One? Two?

If you're so convinced they did a random sample, why not e-mail them yourself and ask them? Or just point me to a source telling me what the TOTAL number of all political officials and candidates under investigation in this country. The number 375 strikes me as in the right order of magnitude.

It seems like the right order of magnitude to me, too. So there probably aren't another 100 investigated Republicans, waiting to be found, you say? So, the study's conclusions were substantively correct?
So, in conclusion, like I said, a bunch of bomb-throwing and smoke-blowing to detract from the very probable truth
that the Bush DoJ has investigated.. about 7 times as many local Democrats and Republicans?

Glasnost, for the last time, in the context in which the word "sample" is used, it is not at all clear that it means what you claim it means. It reads to me as "the sample of investigations we were able to find," rather than an intentionally created subset of the total. The titles on the data source tables support this conclusion, as they do not use the word "sample" at all.

Today alone I've found 3 or 4 more Republicans investigated but not named in the study. Are there 200 more? I have no idea. What is the bar you'd like me to cross before you accept that the study is worthless? Do you admit that the 7-to-1 claim is false, and are you now resorting to a claim that there must be some bias?

Not being privy to the professors' methodology (which is normally and routinely disclosed as part of actual academic investigations), I have no way of knowing whether the other biases I described actually impacted their results. Failing to define "investigation" in any way beyond "some reporter said it" by itself renders the study meaningless. There is no legal category of "under investigation."

A more rigorous study might focus on indictments, which are public knowledge and are very clearly defined. Counting "investigations" is mostly counting characterizations made by reporters and pundits, not actions by U.S. Attorneys.

But hey, it backs up your biases, so you're convinced. Revel in it.

Hilzoy, fair enough, but in the comments you've defended, incorrectly, the study (denying that it claimed that the Bush administration was the "first" to do this), and you did cite it to bolster your argument. It would be incumbent on you, then, to either defend the study or admit that it's flawed and unreliable to support your primary point.

More cases to keep in mind for the next fundraiser for the victims of BVA (Bush-induced Voluntary Autism).

Um, professors of communications - statistical surveys are a pretty large part of communications - their chi-square calculations are posted, and how many in 100 random Americans know how to do those?

Um, I'm not questioning their calculations. I'm questioning their competence at collecting data on federal prosecutions. Did these "scholars" winnow out a sample from TRAC's database of federal prosecutions? Or would they even be aware of such a database? Their reference to "local beat reporters" makes one think that they just searched news stories. Which might seem fitting to simple-minded communications professors, but is not a good way to conduct a study like this, and it's certainly not a way to get a random sampling of federal investigations. Garbage in, garbage out.

Speaking of databases, let's not guess about the 375 number being a reasonable estimate of the total number of prosecutions, OK? DOJ's Public Integrity Section files a report to Congress every year on the public corruption prosecutions undertaken that year (and in previous years). Here's the report for 2005. If you scroll down to page 53, you'll see that between 2001 and 2005, there were between 1087 and 1212 prosecutions each year (that's the sum of federal, state, and local). It's actually unclear whether this represents new prosecutions initiated in each year or the total number of open prosecutions (some of which obviously last for more than one year).

In any event, there have been many more prosecutions under Bush than 375 -- up to 6,000 or so. In which case it's pretty doggone important how these so-called scholars came up with the sample.

I'm questioning their competence at collecting data on federal prosecutions.

and your job title is ?

I'm questioning their competence at collecting data on federal prosecutions.

and your job title is ?

As a gentle cough all around, the argument from authority has both uses and limits.

Amen. If I can't criticize the scholars' lack of knowledge of legal databases and their crappy data collection because I'm not a professor of communications, then who is "cleek" to question my arguments?

If you don't want an infinite regress of ad hominems, cleek, try addressing matters of substance. If you can.

who is "cleek" to question my arguments?

Someone who comments as "John Doe" really has no right to put the assumed name of another commenter in quotes.

He/she is cleek. You are John Doe. No need to make an issue of it.

"...then who is 'cleek' to question my arguments?"

Someone who has been posting here for something on the order of two years, and thus providing a long record we can all decide how much or little credibility to give to, since you ask.

You? Under a week.

Hey, you asked.

Power of argument, particularly as noted via logic and citations of verifiable facts, usually triumphs around here, or any other fair site, FWIW.

Um, I'm not questioning their calculations. I'm questioning their competence at collecting data on federal prosecutions. Did these "scholars" winnow out a sample from TRAC's database of federal prosecutions? Or would they even be aware of such a database? Their reference to "local beat reporters" makes one think that they just searched news stories. Which might seem fitting to simple-minded communications professors, but is not a good way to conduct a study like this, and it's certainly not a way to get a random sampling of federal investigations. Garbage in, garbage out.

Actually, I think this might be a better description of everyone HERE than the professors (me included).

This looks more and more like a study of how the MEDIA is reporting on these issues. Which makes the professors neither simple minded nor incompetent--they're studying exactly what their competence lies in...no more, no less. That's of interest certainly (I'd say of academic interest, but....).

I think everyone is reading a bit more of this than is warranted.

OK, Mr. Farber, but that's really a pretty meaningless criterion. Just in this one thread, I've offered evidence about the existence of legal databases; the DOJ's annual reports on public corruption prosecutions; and the exact nature of a few of the prosecutions included in the study. I'm one of the few people who seems to have actually looked at the study (which is why I don't make embarrassing errors like claiming that the study had a "random" sample, or that it did not claim anything about Bush having been the "first" to engage in political prosecutions).

Cleek, on the other hand, has offered nothing but snarky questions and remarks. So if he/she has a two year history of equally non-substantive posts, that doesn't really say very much, now does it?

"So if he/she has a two year history of equally non-substantive posts, that doesn't really say very much, now does it?"

Perhaps not. Probably you win the argument. Yay, you.

Gee, no one got my Walter Brennan joke, back when we were on helots.

John Doe, thanks for the link to the Public Integrity Section report. I should have gone there myself. Those stats on page 50 and following are of people who have been formally charged, by either indictment or bill of information, and of people who have been convicted. There is nobody "under investigation" included in those figures.

Note that the corruption figures in the DOJ report include non-elected officials as well as elected. In 2005, there were 96 state officials and 309 local officials charged. Almost all the federal cases involved non-elected officials. These figures include both local assemblymen and mayors as well as the non-elected officials who work for them, cabinet secretaries, chiefs of staff, accountants, bookkeepers, treasurers, contract review officials, etc. That's a much larger pool of folks than the elected officials, so the total of elected officials and candidates is going to be lower than the totals for "local corruption" given in the report.

John Doe, please stop by Stubborn Facts and comment over there if you have a moment. We're always looking for commenters with sharp analytical minds and some common sense.

We're always looking for commenters with sharp analytical minds and some common sense.

Also known as: people who agree with me.

John, that's an interesting report and all. Hoorah. No, really, nice job. However, as Pat helpfully points out, it's not really relevant to what we're discussing - investigations, and indictments. First of all, they don't have party affiliation in the stats in the report, secondly Pat's point about both elected and unelected officials, thirdly, only indictments and convictions are listed here. So, no, actually, the professors couldn't have used this for "investigations". And the fact that the DoJ's public stats don't mention "investigations" is sort of convenient, isn't it? Most importantly, it's not helpful or credible as a piece of evidence to suggest that the pool of investigated elected officials is much larger than the sample. So there's your embarrassing mistake for the day. You can lose the smirk now.

Now, rather self-evidently, just because, as Pat says, There is no legal category of "under investigation.", that doesn't mean that there isn't an administrative category of exactly that type. We can all agree that Congress should subpoena those administrative records pretty darn quickly, shouldn't they? How many local elected officials does DoJ have a file on, from which parties? That would clear up whether the good profs are right very quickly.

Pat also says: it is not at all clear that it means what you claim it means. It reads to me as "the sample of investigations we were able to find," rather than an intentionally created subset of the total.

At this point, I don't know that anymore, either. John is right that I thought that first and don't see a strong suggestion of it now. However, this is also very unlikely to be important, because:

Pat:
Today alone I've found 3 or 4 more Republicans investigated but not named in the study. Are there 200 more? I have no idea.

Come on. If you had to bet a loved one's life on the answer, where would you come down? Because another 200 (150-200) is what it would take for the study's conclusions to be substantively wrong. The 3 or 4 you found do.. not... matter... for a survey which never claimed to have found them all to begin with. Whether or not it was "random" as a randomized subset of a larger controlled population of data, it's claimed to be a "sample". And I, heart on my sleeve here, believe in the absence of any evidence to the contrary, that it wasn't cherry-picked. That would be very obviously stupid and self-defeating.

What is the bar you'd like me to cross before you accept that the study is worthless?

Reasonable evidence, or compelling and suggestive logic that the conclusions - that a statistically significantly larger number of Democratic local officials than Republican ones have been investigated by the Bush DoJ - are substantively wrong.

Not griping about how much work they showed. Not hating on their choice of websites to put up their findings, or their choice of extraneous words. Not that you can think of ways in which some person doing some study on this would have some (unknown) level of bias, considering that any accidential bias would have to be mind-blowingly enormous to wash away a disparity of 7 to 1.

After all this twisting in the wind, someone come up and state clearly: "I think the Bush Admin DoJ really investigated as many, or about as many (let's say, within 50%) Republican local elected officials as Democratic ones".

I'm standing up here and saying I believe the study's conclusions. Even if genuinely unplanned, systemic bias factors are eventually discovered, the ratio might fluctuate, but not very much. You can't even bring yourself to directly say, "I think the underlying conclusions are wrong". By all means, go for it.

This looks more and more like a study of how the MEDIA is reporting on these issues.

Gwang, this sounds a lot like Pat's:

Counting "investigations" is mostly counting characterizations made by reporters and pundits, not actions by U.S. Attorneys.

This is, no offense, pretty unlikely. How many stories are run that the DoJ is investigating someone, without hearing from the DoJ, or checking with the DoJ, to see if the DoJ is actually investigating them. If there were press stories about DoJ investigating 290 Local dems and 35 local Republicans, to claim the DoJ isn't involved in that very strange discrepancy is ... very charitable. Very, very, very charitable.

I'm sick and tired of seeing recent statistical evidence - usually the best or most complete information on record - dismissed out of hand, and its creators villified, assumed away as liars and quacks, because they can't jump through an infinite number of post-hoc hoops put forth by Republican spear-carriers. So, there's another way of getting me to change my mind: quote me a better study. Maybe one with a larger total data set. I'm all ears.

The same thing went down with the Lancet.

The authors of the study are clearly dishonest -- as shown by their claim to have proven that the Bush administration is the "first" to do [something] even without looking at any pre-Bush evidence.

I would not like to say that the author of the phrase above ("John Doe") or the others who made a similar argument were "clearly dishonest," so I won't. But they are clearly wrong.

The "smoking gun" in the original report, as quoted numerous times above, is that, as JD himself quotes: "The current Bush Republican Administration appears to be the first to have engaged in political profiling."

"appears to be . . ."

Anyone with any familiarity with social science research - or the English language, I am tempted to say - should recognize this phrasing as indicating, "We think this is true but haven't bothered to research it, because it's not important."

Condemning such a statement on the grounds that the authors haven't bothered to research it is indicative of, in the kindest possible light, an inability to comprehend what one is reading.

In the worst possible light? . . . well, I said I wouldn't call anyone "clearly dishonest."

Dr Ngo, this thread's been running for two days. During that time, a small number of people who either don't often comment here or who have never commented here before, have been suggesting that it matters whether or not Bush/Cheney is the first administration with this kind of bias in US Attorney's investigating politicians: matters to an extent that if it is not true that Bush is the first one, this somehow casts doubt on proven facts (firing of Bush's own appointees mid-term) and likely allegations (that this was being done because they were not prepared to investigate Democratic politicians against whom there was no evidence of wrong-doing).

Each time this suggestion (that Bush is not the first, and that this matters) has been brought up, it has been refuted quite thoroughly, by several different people, from Hilzoy to myself.

None of these new commenters have paid attention to the refutation, but continue to buzz about Bush possibly not being the first, as if that mattered.

At this point, I'd say that bad faith is proved.

Since no-one has mentioned this, I guess it's incumbent upon me to note that the first clause of the op-ed -- which is not the actual article, since this also appears to have escaped everyone's notice -- is the following:

Our ongoing study of the Bush Justice Department (to be published in 2008)...

A closer read shows that: a) it's an ongoing, longitudinal study from which the data has yet to be fully collected; that b) this material was...

...presented the preliminary data through August 2004 at the Southern Speech Communication Annual meeting in April 2005 in Baton Rouge and as a refereed panel paper with data through December 2004 at the November 2005 annual meeting of the National Communication Association.

so presumably the full data set including methodology is available somewhere, having been refereed -- although I'll grant that I have no idea how rigorous the refereeing of the National Communication Association is, and neither do you -- and that c) the authors' claims are implicitly to be verified in their actual article rather than in the précis they provided in the op-ed.

None of this is to say that the op-ed, as it stands, is dispositive; it clearly isn't. And hell, maybe they really did just pull the first 375 cases out of a Lexis-Nexis search, in which case this study really would be "junk science". Those claiming to have "read the study" simply haven't, however, and those claiming "dishonesty" should reduce their wild-eyed claims to something more tenable like "as-yet unjustified". Oh, and they should also stop impugning the literacy and integrity of others, but that should really go without saying.

[I should also note that I tried emailing the authors of the article but something screwed up in my registration and I couldn't get through. Should anyone else succeed, I'd appreciate hearing the response.]

Addendum:

Oh, and they should also stop impugning the literacy and integrity of others, but that should really go without saying.

This should really go for everyone on this thread, myself included.

If you don't want an infinite regress of ad hominems, cleek, try addressing matters of substance. If you can.

see, i'm not the one whinging:

They are professors of communications. This makes them competent to study federal prosecutions . . . how

so, please, tell us your job title, so that we can all decide if you're qualified to criticize them, or anybody, on anything, ever. or, you could admit that the issue of job description in this matter is yet another red herring, and has no bearing at all here as a matter of substance.

then who is "cleek" to question my arguments?

cleek is a programmer.

but it was you, "John Doe", who brought up the issue of job title vs. competence, and seems determined to use it as a way to discredit the study you apparently seem to think is the key to this whole issue.

that's about as far away from a "matter of substance" as you can get.

and, i apologize for two similar posts in a row.

cleek also can rip a mean guitar riff, once his cat got everything worked out

my cat really is the key to it all.

Glasnost:

Most importantly, it's not helpful or credible as a piece of evidence to suggest that the pool of investigated elected officials is much larger than the sample. So there's your embarrassing mistake for the day. You can lose the smirk now.

The DOJ report that I linked to showed around 1,100 to 1,200 prosecutions every year. I'd assume that the pool of investigations would be even larger than the number of prosecutions. Yet you interpret this to mean that the "pool of investigated elected officials" could be close to 375? Wow. Do the math there: 1,100 per year is more than 375 per a 5-year-period.

"The current Bush Republican Administration appears to be the first to have engaged in political profiling."

OK, they use the phrase "appears to be." So what? Given that nothing pre-Bush was even purportedly studied, there's still no basis whatsoever for claiming that Bush "appears to be" anything.

Jesurgislac: There's been a long discussion of points other than the one you seem hung up on, i.e., whether Bush was "first." Of all the criticisms that can be made of the so-called "study," that doesn't even make the top 10. Why are you still pretending that it's the only criticism that's been made?

Cleek: it was you, "John Doe", who brought up the issue of job title vs. competence, and seems determined to use it as a way to discredit the study you apparently seem to think is the key to this whole issue.

Not true. Please read before posting. I do not use their job titles to infer their incompetence. I point to their obvious incompetence (i.e., searching news stories by some completely unspecified method and claiming to have found a "sample" of federal prosecutions). From their incompetent choice of methodology, I conclude that these professors of communications weren't even aware of the existence of legal databases.

From that, I also conclude that it's silly (as glasnost suggested) to assume that the study is just fine because the professors have "obvious competent backgrounds." No, they don't, and their "study" shows it.

But all this is a red herring -- I'd never have even mentioned their qualifications had glasnost not tried to play the qualifications trump card in a completely ridiculous and unjustified manner, as if anyone who is a professor of anything is therefore "obvious[ly] competent" to investigate something entirely outside their field.

Glasnost, I agree it's unlikely that the authors intentionally cherry-picked their stories, which is why I went to such lengths to explain why a variety of forms of selection bias could have a significant influence on their results.

Go read a few of the stories cited by the authors as their sources, and see for yourself what the DOJ had to say. Take the Baltimore example I gave in my post. The DOJ's response was "no comment." Why? Because the prosecution should never comment about on-going investigations, and US DOJ policy is not to do so.

You seem prepared to admit that the extent of the study may well be flawed, that the number is not 7 to 1 as claimed, but instead wish to reclassify its basic point as being that there is some bias. Well, that's not the claim the authors make. They claim it's 7 to 1, not "some." Their conclusion is defeated if the ratio is only 6 to 1, or 5 to 1. Your conclusion of bias may or may not be correct, but it's not made by the study. You've asked me to disprove not the study but your own conclusion of some bias.

Still, let's narrow it down some to see whether it's possible to do that. How do you define "under investigation"? Are each one of the Baltimore city council members "under investigation" because they each got a subpoena for some records? Can I count investigations begun during the Clinton Administration but continued during the Bush Administration? If they subpoena the records of a bunch of local contractors and the press assumes that means they are investigating the local city council, do I count those, and whom do I count, the mayor, the entire city council, who? Do I count only investigations confirmed by the U.S. Attorney? Give me the criteria which will convince you, laid out ahead of time, and I just may gather the data to do it.

That's why laying out the methodology is so important. In science, you lay out your experimental criteria, methodology, and definition of success before you do the experiment. By analyzing those factors, we can begin to determine whether the experiment produced valuable results or merely confirmed what the experimenters wanted to hear.

Anarch, as for the "preliminary" nature of this study, all we have to go on is what the authors put on their website. They chose to release it, unfinished, and it is now being used to support political charges. I didn't go digging through unreleased studies and bring them to light, their study was cited by a prominent, partisan columnist. If it's going to be used now, then it must be supported now.

If nothing else I have to award points for stamina all around.

I'm still wondering why Republicans aren't running with this study to show how much more corrupt Demorcrats are, weird.

Not true.

i'm willing to accept that you didn't actually mean to imply that, but your words certainly suggest that's what you meant.

Please read before posting.

please make sure your writing says what you intend.

And a pretty good read to boot! Although I'd issue demerits for excessive sarcasm and personal attacks. If everyone ulitimately kisses and makes, those can be remediated.

I always find it interesting that these sort of protracted discussions evolve (or devolve) into finer and finer hair-splitting, and in parallel, become more and more heated. Someone should study that.

Pretend my last comment immediately followed OCSteve's. I'm slow.

I always find it interesting that these sort of protracted discussions evolve (or devolve) into finer and finer hair-splitting, and in parallel, become more and more heated. Someone should study that.

I definitely notice the hair splitting on this one. When it gets down to parsing verbiage/methodology that is at least somewhat ambiguous there is no way for one or the other to really ‘win’ (whatever that means in the context of blog comments).

People dig in and set their feet but it’s not really going anywhere. I usually give up at that point (thus the comment about points for stamina).

Anyway – Hilzoy has a fresh post up on the topic without Krugman or this study involved. Suggestion: Retreat to your corners then come out swinging there :)

Anarch: presumably the full data set including methodology is available somewhere, having been refereed -- although I'll grant that I have no idea how rigorous the refereeing of the National Communication Association is, and neither do you -- and that c) the authors' claims are implicitly to be verified in their actual article rather than in the précis they provided in the op-ed.

OK, maybe it's the case that in a year or two, we'll all find out that the authors really did look at a database of all 6,000 or 7,000 prosecutions, that they really did select a sample of 375 randomly, and that the 7:1 ratio in local prosecutions really does exist. To be sure, that would still leave open the question whether there was any bias in the process at all. Perhaps those large urban cities controlled by Democrats really are more likely to experience corruption of the level that deserves federal investigation. [Note: the authors don't even seem to be aware of this possibility.]

But we're nowhere near that point yet. All that the "study" consists of now is: An op-ed, and a few PDF files that merely list 375 "investigations" with zero explanation of how they were chosen.

It's embarrassing that Krugman would cite and rely on this as "statistical evidence" of anything.

Which leads to a broader point: Academics probably shouldn't write regular op-ed columns (or blog, for that matter). They all too often end up sacrificing intellectual standards for the sake of making a polemical point. There's absolutely no way that Krugman, in one of his scholarly works, would claim that "statistical evidence" proved something, when the evidence consisted merely of an op-ed backed up by a few charts with zero explanation of methodology. Krugman the academic knows better than that. But Krugman the columnist will cite anything and everything that supports a polemical point, no matter how preliminary or unsubstantiated.

[Unrelated hint for Cleek: Someone else brought up the issue of "competence" first. I was merely responding to it.]

I hope this thread somehow magically disappears soon, but I will add my two cents worth anyway.

It seems this thread has devolved from the subject of the post to a critique of a study, as if showing the study is flawed means there really is no problem.

Somehow this reminds me of the whole uproar over the allegedly forged memos aired by Rather. They became the focus, rather than the substance of his total report.

Of course that is an easy way to avoid looking at the reality. Find a flaw, no matter how trivial and focus on that. Magically, everything else becomes suspect.

I, for one, am glad hilzoy posted a new post on this whole affair.

It seems this thread has devolved from the subject of the post to a critique of a study, as if showing the study is flawed means there really is no problem.

You're reading something that's never been said or implied. I'll lay it out for you:

1. The attorney firings look like a problem, at least in a few cases, albeit one of very limited scope.

2. If this study is right, however, there's a massive and nationwide problem of politically-partisan investigations by federal prosecutors.

Do you see why it might be important to know whether 2 is true, even if that doesn't affect 1?

JM, sometimes misdirection is all you've got.

Here's the only thing I could find on the web that discussed the Cragan/Shields methodology. It's by someone who actually called up Shields:

I asked Shields about his methodology. How did he arrive at the conclusion that the Bush administration targets Democrats?

It seems the research done by Shields, or Cragan, or both consisted of simply hitting the Search button on Google.

Specifically, their work consisted of a tabulation of the number of times the words or phrases federal grand jury or public corruption AND elected produced a result, along with "a census search of extant press releases available at each U.S. Attorney's Office official Web site." The names of Democrats come up far more often than those of Republicans.

From that, he asks us to believe that Democrats are investigated disproportionately and presumably unfairly.

I question the conclusion because of what is not addressed, namely:

What about investigations that never get media coverage (and thus don't get on Google)? Grand jury proceedings are by definition secret, so we can never know the true extent of what's really going on.

What about the disparity of media coverage between large and small towns? The Inquirer covers Philadelphia like a glove - but local outlets may be less well-equipped to look into secret or putative investigations of their local leaders. This is an important point because investigations often are sparked off by media reports in the first place.

What about the disparity in Internet participation among newspapers? Same here. Many good investigative stories in local papers don't make it to the Web and therefore don't make it to Google.

Where is the confirmation of actual investigations? Shields and Cragan got some of their data from U.S. Attorneys' office press releases, but evidently never called those offices back to check their figures.

Why not address actual prosecutions, not just simplistic media coverage of investigations? Granted, Sprague's claim, and the study it's based on, concern investigations, but it's the cases that make it to the prosecution stage where the rubber meets the road.

If this account is right, the study is garbage. Again, both Krugman and Hilzoy should be embarrassed for blithely citing it as evidence without even attempting to figure out what methodology was used.

This looks more and more like a study of how the MEDIA is reporting on these issues.

Gwang, this sounds a lot like Pat's:

Counting "investigations" is mostly counting characterizations made by reporters and pundits, not actions by U.S. Attorneys.

This is, no offense, pretty unlikely.

For communications professors??? Hardly. That's bread and butter, meat and potatoes, tenure-granting research there.

This is not to say that these resarchers AREN'T stepping beyond their competence here. But as a former communications academic, calling this study garbage is unwarranted--it's the generalization beyond the scope of the media that seems to be garbage.

People dig in and set their feet but it’s not really going anywhere.

Oh, I don't know. Pat agreeing that the data set is unlikely to have been intentionally cherry-picked (or at least, shouldn't be assumed so without specific evidence) gives me room to consider this a civil debate.


Anyway – Hilzoy has a fresh post up on the topic without Krugman or this study involved.

Yeah, but it looks like we mostly agree about that one. Thus, not so much fun. I like debate. Heated debate, even.

The DOJ report that I linked to showed around 1,100 to 1,200 prosecutions every year. I'd assume that the pool of investigations would be even larger than the number of prosecutions.

DOJ report = elected & unelected officials

Data set = elected officials

1100 = tells us little about subset, "elected officials"

Thanks = anyway

The local official subtotal of investigated/indicted officials is 262 to 037. The p value for this is less than .0001. It only has to be .05 to be statistically significant. In other words, that ratio would have to be unrepresentative of the population by *massive* amounts for the study to be substantively wrong.

So, you've found 7 Republicans and 1 Democrat that weren't listed? Do you think that there are, oh, 200 more investigated Republicans, waiting to be discovered, that the study's authors somehow skipped? If you found 200, and there were not another 200 Democratic investigations also then the study's conclusions would be substantively wrong. And then, and only then, do you have a case for arguing that the study's authors had any reason at all to act dishonestly. And you sure don't have, yet again, any evidence that their sampling methods were improper.

This is horrific statistical analysis. The p value is .0001 only if Democrat/Republican is the only possible independent variable. Which it clearly is not. Analyzing it like that is the kind of statistical nonsense that gets people to believe that Saggitarians are 38% more likely to have broken arms than people from other astrological signs.

This is pretty far afield, glasnost, but what's your reason for thinking that in a 5-year period where around 6,000 officials were actually prosecuted by the federal government, the number of investigated officials who were elected (rather than appointed) is anywhere near 375? The pool of investigated people has an absolute minimum at the number of prosecutions, and is probably much larger. What makes you think that only around 6% of officials -- who are in a high enough position to be worth bribing -- are "elected"?

Jesurgislac: At this point, I'd say that bad faith is proved.

Oh, I know, Jes, but sometimes when a particular point (in this case a linguistic one) has not yet been made, I feel obliged to make it anyway. In the faint hope it might make a difference.

Although I'm an ex-Christian, I suppose I still tend to believe in the possibility of redemption, even of those who are clearly wrong wrong wrong.

Up to a point. Beyond which, I begin to suspect the Sin Against The Holy Spirit. No cookies for them. Forever.

So I shall probably not be responding to John Doe, johnt, Brett and the rest of that mob hereafter. Pity, that.

John Doe... On that last point, I would suggest that it's not always the elected official who is in a high enough position to be worth bribing. Very often it is the staff or the contract officer who have the biggest potential to influence contract awards and so forth. Is it cheaper and safer to bribe the one guy writing the bid specs (so they will be written in a way which makes your company the only real choice) or a majority of the aldermen who will be voting to award the contract?

Glasnost, I too enjoy a good vigorous debate. I think you're as wrong and biased as I'm sure you think I am, but you debate civilly and stick with it, and I appreciate that. Wrong as you are, you actually engage in argument rather than contradiction.

Wrong as you are, you actually engage in argument rather than contradiction.

No he doesn't.

300

The comments to this entry are closed.

Blog powered by Typepad