See update, below.
Because Matt W. asked, and we don’t diss Matt W. here, if we can help it.
A quick effort at determining the incidence of “God” in the New York Times from 1/1/1901-12/31/2005. (Couldn’t think of a way to limit it just to the editorial page.) I searched ProQuest’s NYT historical database for the word “God” and recorded the number of articles in which it occurred for each calendar year. Then to control for maybe different sizes of NYT over time, I did the same for the word “January”, and created a ratio of number of articles containing “God” to number of articles containing “January”. (I got this method from this paper.)
Here you go; click on it to see it bigger.
Update:
Well, Vance made me doubt the suitability of “January” as a denominator, so I went back and tried two others: no search term, which should capture all articles in the database, and “the”. Results below.
It looks to me like there’s no huge difference made by different denominators.
50 comments
December 19, 2008 at 12:12 pm
Vance
Proof of the advances made by the War on January.
December 19, 2008 at 12:16 pm
Levi Stahl
Hmm. Somehow we got through World War II without talking about God all that much. Maybe he asked us to be all sneaky-like about him being on our side? And starting in the 1950s he hired a PR firm?
{Speaking of which: I used to play on a softball team whose second baseman was a rabbi with a full beard. We kicked around the idea of naming the team God on Our Side, but ultimately didn’t have the guts to deal with the offense that would surely have been taken.}
December 19, 2008 at 12:22 pm
kid bitzer
but i hope at the very least you called your softball gloves god mit-uns.
December 19, 2008 at 12:26 pm
eric
Actually, Vance may be right, dammit. Which is to say the absolute incidence of the word “God” does increase over time, but the pattern in the use of the word “January” is making the increase look bigger. Maybe I’ll do it over with “the”. No I won’t.
December 19, 2008 at 12:29 pm
Vance
What else holds steady? “Corruption”?
December 19, 2008 at 12:30 pm
eric
Well, I mean, I should stick to my methodological guns, right? “January” should proxy for the number of words in the newspaper?
December 19, 2008 at 12:38 pm
Vance
Seriously, I’m surprised there’s any kind of secular pattern to the incidence of “January” — I was just being silly. Your suggestion of “the” seems sound.
December 19, 2008 at 12:39 pm
Levi Stahl
Maybe “Times” would work? Or “United States”?
December 19, 2008 at 12:40 pm
TF Smith
Along with WW II, also seem to have gotten through WW I, the Depression, and Korea…jumps up during SEA, however.
LBJ?
December 19, 2008 at 12:43 pm
Anonymous
Fascinating, but what about accounting for “Lord” and “Almighty”? They were more colloquial earlier in the twentieth century.
December 19, 2008 at 12:46 pm
transportinburma
May I make a gentle recommendation? When making a figure, always label your axes, even if you think they represent is obvious. I was chastised for presenting some unlabeled axes at my first thesis committee meeting, and in retrospect, I think justifiably so.
(I only say this because I noticed unlabeled axes in your “Make-work” as well.)
December 19, 2008 at 1:28 pm
eric
Fascinating, but what about accounting for “Lord” and “Almighty”? They were more colloquial earlier in the twentieth century.
I think the problem would be, they get used for other purposes than referring to God in the NYT.
December 19, 2008 at 1:28 pm
eric
When making a figure, always label your axes
Duly noted, thanks.
December 19, 2008 at 1:36 pm
eric
Seriously, I’m surprised there’s any kind of secular pattern to the incidence of “January” —
Unless perhaps the New York Times got bigger 1901-1950 and smaller 1950-2005. Which is plausible. The whole point of using it as a denominator is to control for change in size of the pool you’re drawing from.
December 19, 2008 at 1:49 pm
andrew
Is there also a metadata issue? January might not be indexed unless it’s mentioned in an article rather than at the top of a page. I don’t know how ProQuest does it.
I’m not convinced that corruption did decline substantially 1870-1920, by the way, but that might just be because I think it continued to decline into the 1950s (which of course doesn’t mean it didn’t already decline a lot by 1920). But corruption is really hard to count, and I’m just guessing, so what do I know? I suppose I should finally read that paper you linked to.
(This looks like a free link to that paper. From this book.)
December 19, 2008 at 1:53 pm
andrew
And this paper specifically indexes corruption based on newspaper reports of corruption, if anyone else is actually interested.
December 19, 2008 at 2:05 pm
eric
Right, I think that’s the same study as the one I cited.
December 19, 2008 at 2:05 pm
Sam-I-am
Totally not relevant, but I heard Eric on NPR morning edition today. Something about FDR and the Great Depression, I think. ;)
December 19, 2008 at 2:11 pm
andrew
The second paper I linked is different than the one cited in the post (they’re in the same collection, though).
December 19, 2008 at 2:34 pm
eric
I heard Eric on NPR morning edition today
Really? I didn’t even know I was on. I’ll have to hunt it up, otherwise Dave will be sad I didn’t post it.
December 19, 2008 at 2:59 pm
dana
I’m confused. Is “size of the paper” referring to circulation or words deemed fit to print?
December 19, 2008 at 3:04 pm
eric
I think the point of counting “January” and dividing by it is that we think it’s a value-neutral word that occurs, more or less, in the same proportion to the total words in the paper every year. So the value of “January” should go up and down as the number of articles in the paper goes up and down. So the NYT might increase or decrease in size both in the short term—maybe wartime paper shortages?—and in the long term, as the market for news changes. So we want to account for that, otherwise the number of articles using the word “God” might increase simply because the paper has more articles.
Right?
December 19, 2008 at 3:05 pm
eric
Although, it appears the ProQuest search engine will allow one simply to count number of articles in a given year by not entering a search term. Hmmm.
December 19, 2008 at 3:10 pm
Cosma
Is it really so hard to just get a total count of the number of words per year? That seems odd but I could believe it.
I can imagine stylistic variations which would change how often “January” shows up (e.g., “the first month”, or using “this month” in January, “last month” in February, etc.), and that sort of style might change over time. It’s a lot harder for me to dream up ways of changing the frequency of “the”. (Hello, Prof. Zipf!)
December 19, 2008 at 3:10 pm
Cosma
My post crossed with Eric’s @ 3:05.
December 19, 2008 at 3:20 pm
PorJ
1. ProQuest on the daily newspapers is not comprehensive (even if its the best we’ve got). Much better on academic stuff.
2. Counting specific words is not content analysis, by a long shot – and there’s some conflation going on here, in terms of conclusions drawn.
3. So the value of “January” should go up and down as the number of articles in the paper goes up and down. This is a strange assumption, from a methodological perspective. There are holes in it – for instance: Do we know whether guidelines for copy style were changed at any point – and whether that could have had an impact (on the use of either “God” or “January”)? Could the paper (as somebody mentioned above) have synthesized “Lord” or other words into God in the last few decades? Perhaps it once used “Last month” more commonly in February than “January” – too many assumptions.
But you did a good job making the data come out the way you wanted it to.
December 19, 2008 at 3:25 pm
PorJ
Also, if you want to limit the search to editorials (or any type of article) go to “More Options” on the search page, and grab the dropdown menu of “Article type.”
December 19, 2008 at 3:26 pm
PorJ
Oops, I mean “More Search Options” and “Document Type”
December 19, 2008 at 3:42 pm
Matt W
w00t! Thanks Eric!
Fascinating, but what about accounting for “Lord” and “Almighty”?
Well, actually, what I was thinking was that it would be interesting to see whether the incidence of “God” in presidential speech, as plotted by Lim, was any different from the incidence of “God” in some other speech. Do presidents start talking about “God” more because it’s something we expect of the president, or because everyone is? Which means coming up with a graph that’s comparable to Lim’s graph — which means counting “God” only, because that’s what he did. If the use of “God” shot up in presidential addresses because people started using it instead of “Lord” or “Almighty,” you’d expect to see the same effect on the NYT editorial page.
December 19, 2008 at 3:47 pm
jazzbumpa
Though, try as I might, I cannot apply Elliott Wave labeling to this data set (too much overlapping,) it’s clear there are 2 domains: the first where G/J is decreasing secularly, and then the second, where G/J is increasing secularly. (The P/E of the DJI exhibits a similar phenomenon.) I take the occurrence of the inflection point (min, not Max) at the time of my birth (+/- a year – it’s a little hard to pin down precisely) to be a strange and improbable coincidence.
I have no fundamental problem with a war on January, but I think a war on February would be a wiser allocation of resources.
December 19, 2008 at 4:34 pm
eric
But you did a good job making the data come out the way you wanted it to.
Wow, that’s exceptionally obnoxious.
December 19, 2008 at 4:41 pm
eric
Just to come back on this, PorJ: wow, that comment was really, actually, gobsmackingly obnoxious, and wholly unwarranted. I didn’t want the data to come out any particular way, I wanted to test what Matt W. wanted to test—hence the use of “God”, just as in the previous post.
And the methodology of “January” I borrowed from an article looking at similar incidences of words in computer databases. Which seemed sound to me.
But then when I wondered if it actually was sound, I went back and picked a couple other denominators—and guess what, no real change!
You’ve really no evidence at all of bad faith on my part.
December 19, 2008 at 4:47 pm
andrew
Incidentally, I wonder what the God stats in the Congressional Record/Globe would show.
December 19, 2008 at 4:49 pm
eric
I wonder what the God stats in the Congressional Record/Globe would show.
The Globe/Record isn’t online after Reconstruction, is it? Also, I’m done making graphs of God for now, thanks.
December 19, 2008 at 4:53 pm
andrew
Still not? Oh well. The Globe (pre-1873) is online, but without the later numbers there’s no point in trying it.
done making graphs of God
Next challenge: graphing the mind of God.
December 19, 2008 at 5:25 pm
PorJ
O.K., I’m sorry for the snark. It was shorthand for a much longer comment I wanted to make but I’m feeling really sick today. The longer point is: So What? I.e.: what’s the larger purpose of this exercise? To prove that the use of “God” in American culture and society is more widespread today (or trending upwards) than it once was? To argue that this data is indicative of religiousity? What, exactly is the bigger picture being drawn, the larger inferences?
If you are honestly just looking at the (relative) use of the word “God” in the New York Times over time, and are not asking us to draw conclusions, then I double-apologize (although I note that if you don’t segregate the results, New York Times historical returns advertisements as well as articles. Just for kicks, I got 339,109 hits for God (not in quotes) between 01/01/1901 and 12/31/2005. Does this match your total?)
But my point is that any larger conclusion except for the use of that single word is useless. Management of theNYTimes has changed over time; how do we know that changes in the use of “God” don’t have more to do with the upper management (or average editors’ religiousity) than larger societal forces? How do we know that the NYTimes is the same barometer of society and politics today as it was in 1962 when the NY Herald Tribune and the World-Journal-Telegram were still around? The paper is significantly different in terms of production, stylistic conventions, political purpose, etc. I just don’t buy that this exercise teaches us anything significant about the historical times we’re living in (or reconfirms our suspicions).
I think historians use journalism to prove larger points when it suits them but aren’t reflective enough about the problem of assuming journalism is an inherently accurate mirror or reflection of society. I once heard a story about Arthur Schlesinger bitching to a journalist about how newspapers were getting Kennedy all wrong, and a reporter countered “so that’s why you footnoted so many newspapers in the Age of Jackson, huh?” Its probably apocryphal, but it gets to the point I’m trying to make. The newspapers weren’t any more accurate back then, but they could be exploited to support arguments you want to make about the 1830s (with selective reading). Selective data – I’m willing to wager – could similarly be employed to “prove” that “atheism” is on the rise, too (But I don’t have time to check).
And if I’m the one that made an assumption about larger inferences than I promise to heed Felix Unger’s advice in the future: never assume, as it makes an a** out of u and me.
December 19, 2008 at 6:22 pm
Michael Turner
If you divide the number of dancing angels by the area of the heads of the pins . . .
Yes, please stop being exceptionally obnoxious, PorJ. That’s my job here. Clearly. Just ask ari.
(I only say this because I noticed unlabeled axes in your “Make-work” as well.)
Adding axis labels has been left as an exercise for other make-workers. At least until we have new axis powers to fight.
I have no fundamental problem with a war on January, but I think a war on February would be a wiser allocation of resources.
We have always been at war with February. It has greater stimulus value. But if you need to open a new makework front somewhere, how about a march against March?
I have to wonder sometimes, how hard can it be to make useful work, for lots of people, quickly, for its economic stimulus value? I mean, if you went to Google and Yahoo and others, and asked them, “if you could employ five million people for free, how would you use them to improve search engine results?”, I bet they’d come up with some good ideas very quickly.
December 19, 2008 at 6:26 pm
Cosma
[normalizing by "the" and by total articles]
Lord, I believe; thou hast helped my unbelief.
December 19, 2008 at 6:31 pm
Cosma
I mean, if you went to Google and Yahoo and others, and asked them, “if you could employ five million people for free, how would you use them to improve search engine results?”, I bet they’d come up with some good ideas very quickly.
My wife does some research work for Yahoo, and it turns out that for all of the search companies, a big issue is the speed and expense of getting “editorial” judgments. That is, they actually train and pay people to go over (highly anonymized) search logs and decide whether someone who searched for “swimsuit photo” and got Eric found what they were looking for. They use this to evaluate search effectiveness, among other things. If you made getting editorial judgments dirt cheap, there are lots of things they could do right away, and I’m sure many more would occur to them.
December 19, 2008 at 6:42 pm
andrew
Here‘s “God” and “January” compared for the Congressional Record back to December 2000.
December 20, 2008 at 4:07 am
touhy
in Blagojevich’s favor (whoa, there’s a phrase you don’t and won’t hear often), he did not invoke God during his news conference. He didn’t claim that God was on his side. Only Truth. And Kipling.
December 20, 2008 at 7:13 am
jazzbumpa
PorJ -
“If you are honestly just looking at the (relative) use of the word ‘God’ in the New York Times over time, and are not asking us to draw conclusions, . . ”
Looks to me like that is exactly it. I don’t see the example, or the comments driving to any particular point.
Isn’t this an exploratory example – Hmmm, look at at this? Isn’t it interesting? If we examine it further what might we find, what might it mean? That sort of thing.
Anyway, drink plenty of fluids, get lots of rest, and feel better soon.
December 20, 2008 at 10:00 am
Prof Burgos
What’s lacking here is a sense of the undertaking as a qualitative or quantitative measure.
If quantitative, the variable (mentions of “God”) is meaningless. What is it was in the context of (“I hate God,” “God is dead,” “there is no God?”). The simple count would, obviously, miss the substantive meaning.
If qualitative, then a word count is the wrong way to go. You’d need the electronic files and would load them into something like MP2.2 or one of the other concordancing/qualitative content programs, and do an analysis (god w/2 “good” OR “praise” or…) to try to get the substance within which it was mentioned.
During the 1940 (I believe) presidential campaign, the Chicago Tribune — ranked as the worst paper by professionals at the time, owing to its extreme partisanship — ran editorial cartoons above the fold, center columns, on the front page, and in one it declared that “We (meaning Republicans) do battle for the Lord” by opposing a third Roosevelt term. Now THAT is “god-talk.”
December 20, 2008 at 10:46 am
eric
If quantitative, the variable (mentions of “God”) is meaningless
No, it’s not. We know there’s a secular increase of God-talk over the latter c20 (albeit with weird periodic/cyclic variations). That’s not meaningless; meaningless would be if there were no pattern at all.
You’re right, we don’t know if this is an increase of use or of mention. But even an increase of mention is meaningful.
My hunch would be that some of that is sports section articles in which an athlete says “I just thank God for the win”; some of that is disaster articles in which some survivor says “Thank God we’re okay”; some of that is thumb-suckers on the importance of God in modern American life; some of it may indeed be advertisements mentioning God—and almost none of it is the newspaper itself invoking God.
That’s still not meaningless; it’s an index of the portrayal of modern culture in the New York Times.
Most important, though, this is not an isolated trend. It occurs more or less at the same time as an increase of “God” in the presidential addresses—which are almost certainly all invoking, and not merely mentioning, God.
So now we have two nearly concurrent trends: let’s say in all probability an increase in presidents invoking God and an increase of the NYT mentioning God. We don’t know whether one leads or lags, but we have a reasonably good idea that the two things are happening simultaneously.
Saying that in 1901 people were quite likely to say “lord” or “the almighty” means only that we shouldn’t look at the level of “God” in 1901 and the level of “God” in 2005 and say “a-ha! Slightly more God now than then!” But we can still note the direction of the trend as meaningful. There was a fair bit of “God” in the last year of McKinley’s presidency, then it declined, then more recently it has increased.
What this isn’t is conclusive. It’s suggestive:
(a) There are some interesting discontinuities and odd blips, like the drop in WWI and the immediate rise thereafter—that’s weird.
(b) It’s possible, I suppose, that as the NYT reprints presidential addresses, this is driven entirely by the shifts in presidential rhetoric, but I bet it’s not simply the same texts—there’s probably presidential influence on other texts.
(c) There are some weird periods or cycles over the years where there’s in increase in the trend.
So the true meaning of “God” is in the details. But a quick sketch of what’s going on isn’t meaningless at all, it’s an excellent guide to where you might want to look to find out more.
December 20, 2008 at 2:04 pm
Matt W
And again, my purpose in asking was to give some kind of baseline comparison for the graph on presidential use of “God” in the previous post, and I thought of the New York Times because that was the nearest thing to a constant corpus of political speech that I could think of. It may not be the most rigorous historical exercise because I am a philosopher goofing around on a blog rather than a historian trying to prove something (I mean trying to prove anything, not trying to prove some particular thesis).
Also, what the fuck is up with putting asterisks in the word “ass”? Are we in kindergarten here?
December 20, 2008 at 2:31 pm
Gassalasca
I don’t know if anyone posted this link, but this is super awesome:
http://chir.ag/projects/preztags/
December 20, 2008 at 5:38 pm
jazzbumpa
Matt -
I don’t know about actually being in kindergarten, but I certainly feel that the five and under set is my peer group.
January 13, 2009 at 8:06 am
Normalization | Cosmic Variance | Discover Magazine
[...] with you meaning-laden word by the number with that neutral word. Here, from Eric Rauchway at Edge of the American West, is the ratio of appearances of the word “God” to the word “January” in the [...]
January 13, 2009 at 10:33 am
therehman
What’s interesting to me is the alternate increase and decrease by at least two magnitudes between 1947-75.
In other words there is a predominant frequency here and it would be very interesting to see what are the other phenomenons that have similar frequency (like natural disasters, etc)
January 13, 2009 at 10:38 am
therehman
What’s interesting to me is the alternate increase and decrease by at least two magnitudes, more so between 1947-75 but nonetheless through out the century.
In other words there is a predominant frequency here and it would be very interesting to see what are the other phenomenons that have similar frequency (like natural disasters, etc)