Story Behind ISM Errors Reads Like Hammett Novel: Opening LineC. Thompson
As yesterday’s Institute for Supply Management manufacturing index number was being botched, and calmly yet furiously attempting to get the right number to trading floors while markets were, um, fluctuating.
The Top desk can be thought of as the ship’s bridge, plotting a course for the expected news and navigating the day’s unexpected currents that flash across the screens. Most days have an economic indicator or two that get handled by the economy team in Washington, with the Top team picking up their work and applying certain treatments to it. Red headlines, which we sometimes call hot heads or sticky heads (because the red headlines stick to the top of your screen for a short length of time), get re-sent to “flash” at the top of the Top screen.
While one editor is doing that, another is grabbing the first bulletin of the story and applying special codes to it to ensure it gets sent to the right alert systems that you, the customer, have chosen to receive. Meanwhile another editor takes that same bulletin and loads it into the Top page. It can get a little hairy at times, such as when indicators are moving at the same time as something big and unexpected happens.
Or in this case, both.
At 10 a.m. on the dot, as always, the ISM number flashed red across the Bloomberg terminal, showing the reading had fallen to 53.2 from 54.9 in April. This was unexpected: The economists’ forecast was 55.5, and only 15 minutes before the Markit U.S. manufacturing PMI had actually come in a little above forecasts.
Andy Davidson (U.S. Top team leader): “You take it as gospel. We flashed it and accepted it.”
Carlos Torres (U.S. economic indicators team leader): “Accepted it at face value, yep.”
AD: “You expect -- you have a lot of faith that the first number that comes out is going to be a good number. It’s never been a bad number before, and this, according to Ken Fireman, is pretty much unprecedented -- or very, very rare.”
Ken Fireman (managing editor for economy and government in the Americas): “Nobody can remember anybody having to do two corrections within the space of a couple of hours on a major economic indicator like this. Nobody that I spoke to can remember anything like that.”
Chris Nagi (managing editor of U.S. stocks): “There was no real reason to suspect anything was wrong with it. I mean, it was a little out of step with recent good data, but certainly it was in the realm of plausibility. Was it so outlandish that people started to wonder if it was true? I don’t think anyone thought that.”
AD: “We watched the market’s reaction.”
CT: “The people at Stone & McCarthy put out a note on their system 41 minutes later ("SMR: Update - ISM Used Apr Seasonal Factor: Should be 55.4"). Ken Kim, over at Stone & McCarthy, is a guy who’s in charge of writing up their report about the ISM index. This group is into the nitty gritty of the numbers, and they not only look at the seasonally adjusted numbers, but sometimes they look at the unadjusted numbers just to get a sense of what’s going on. He was coming up with a different number than they had just issued.”
CN: “There was a period when the economist who found the problem had tweeted. You don’t know if the person’s right at that point. There was sort of a frenzied effort on the part of people in the newsroom to find out if there was anything to the tweeter’s claim and to figure out what was going on.”
CT: “So he (Kim) gave them (ISM) a call, but no one was returning his call. So at 10:41 he went ahead and put an update on their (Stone & McCarthy’s) system. Then what happened, as I understand it -- I’m not exactly sure about this -- as I understand it, a bond trader saw that Stone & McCarthy update and alerted someone on our bond team to it.”
(Opening Line: Stephen Stanley, chief economist at Pierpont Securities, sent word via the main incoming feed that most companies and agencies use to release news to us by e-mail, alerting us at 11:12 a.m. to the same gist that Kim had spelled out -- ISM used the “more stringent” April adjustment on the May figures. The e-mail was spotted by forex reporter Rachel Evans.)
CT: The bond desk “alerted someone at Bloomberg First Word, then I started getting barraged by messages from people from Bloomberg First Word. So, almost exactly at 11 o’clock, James Holloway (managing editor for Bloomberg First Word fixed-income coverage) sent Chris Wellisz (team leader for U.S. economic policy) a message saying Stone & McCarthy is saying ISM got the number wrong.
‘‘So, I went to the Stone & McCarthy (web)site and read what Ken Kim had written, and as I was reading I realized that, you know, there could be something to it. Because this sort of thing can happen. It sounded more like a technical glitch then the usual conspiracy theories that we hear.
‘‘We started calling the ISM spokesperson, who wasn’t picking up, left a message via voicemail and e-mail with her, then started calling the ISM factory chairman. Victoria Stilwell, one of the indicator reporters, found his cell phone number on line and we started dialing that, and I finally got through to him at about, shortly after 11:15 or so.
‘‘I finally got Brad Holcomb, the factory chairman -- he finally answered. So I told him what was happening in the markets and that someone at Stone & McCarthy was saying they had gotten the number wrong, that it should have been 55.4, and asked whether he had heard about it and whether there was anything to it. And he said they were looking into it and in fact it did look like they made a mistake. And then I asked him if the 55.4 (number) was correct, and he said no, we’re coming up with 56 now. We’re going to be issuing a correct in a little while, but we’re coming up with 56. That was about 20 minutes after 11. At 11:30, we have a prescheduled independent call with Brad Holcomb, and our reporter Tori Stilwell took that call, and during that exchange he reiterated that it was 56.
‘‘He actually went on Bloomberg Radio before even talking to me and, on a Bloomberg Radio interview, said that they had made a mistake and the number was 56.’’
OL: So this was the basis of the first corrected headline.
OL: Three times Holcomb gave the 56 number to various Bloomberg people.
CT: ‘‘Right. Then he picked up my phone call and told me the same thing and then 10 minutes later told the same thing to Tori Stilwell. They had not yet issued a formal correct. This was all verbal communication from him.’’
CN: ‘‘There was a gasp that went up when the headline showing the correction moved. Certainly interest was piqued. There was, uh, healthy journalistic curiosity and an effort was made to nail down what was going on. So that happened.’’
OL: When the number was corrected, the Top desk also took that one on faith?
AD: ‘‘Initially, we did. We figured they didn’t want to get any more egg on their face, that it would be fine -- they would have looked at the numbers carefully, figured out the problem and it would have been fine. We had heard chatter that this wasn’t quite the right number. And it turned out not to be.’’
KF: ‘‘Apparently they applied the wrong seasonal adjustment. They applied the seasonal adjustment that was appropriate for April to a May number, and that, of course, skews your results. In this case, it skewed it into not only less than forecast, but actually what was showing up as a drop from the previous month. So, it really mattered. It was directional. It wasn’t just numerical, it was directional, which is why markets reacted.’’
CN: ‘‘There was animated, midlevel-voice speaking, but nobody was jumping around, saying ‘I can’t believe this’ or anything like that.’’
CT: ‘‘Then other economists started doing their own calculations after we reported that there was a seasonal-adjustment issue. And everyone was coming up with 55.4 instead of 56. So we started getting barraged by e-mails from economists saying they thought the new number was also wrong and that we should go back to ISM and ask them if they thought it (the new, 56 number) was correct.
‘‘So I got on the phone again and tried to track Brad Holcomb down on his cell phone. After about two or three tries, he picked up again. I told him what was happening, and he would neither confirm nor deny that the 56 was wrong. But what he told me was that their research team was crunching the numbers again, just to make sure they got it absolutely accurate, and that they would put a correction out in an hour.
‘‘So then I reiterated, well is the 56 wrong? Is it going to be 55.4? And he did not answer. (Torres chuckles.) He just said we don’t want to make a mistake, we’re crunching the numbers again, we’ll let you know in an hour.’’
KF: ‘‘But Mr. Kim and some of the others had apparently figured it out correctly. It kind of makes your head spin to have that happen. What they said when they explained the second mistake was, basically, we were trying to do this in a hurry and we didn’t apply all the factors that we should have, and that’s why we got it wrong a second time.’’
CT: ‘‘No one was panicking. It started becoming obvious there was something going on. So we just sat here and redialed and redialed and redialed the spokesperson and Brad Holcomb until we got one of them, until Brad finally answered. And we were just trying to see if we could get him to confirm one way or the other.
‘‘In the meantime we started seeing more people saying it should be 55.4. About 10 minutes after my conversation with him (Holcomb), he was nice enough to call back Tori Stilwell and get her on the phone and say yes, we did make a mistake, the new number should be 55.4, not 56. Again, this is all verbal. There was no paper (press release). And that was the last hot head, which was at 12:26 p.m., when we sent out the second correction.’’
CT: ‘‘At that point, there was a lot of confusion. Everyone’s pretty sure that it is 55.4, because that’s what all the economists are coming up with, and they’re finally agreeing to it.’’
KF: ‘‘What made this so striking was, first of all, you had two corrections, not one but two corrections on a first-tier indicator, and secondly that it really did seem to affect markets. The stock market was bouncing around. In terms of even a single correction of this magnitude, there was something if you go back to 1999. The (U.S.) Labor Department applied an outdated seasonal adjustment to a monthly employment report, and ironically enough it was Stone & McCarthy that caught that error as well. But in this case, it was, according to our story, a two-month old revision. It wasn’t something that had any market impact, whereas this obviously did.’’
CN: ‘‘Then there was a kind of gallows humor. I don’t think any journalist takes any pleasure in something like that. Everyone recognizes that there but for the grace of God go I. I think everyone respected that that’s not a good thing for the market, not a good thing for the people involved in it (the faulty numbers) and that taking any schadenfreude-y pleasure in it would be inappropriate.”
AD: “Everybody was surprised that it would happen twice. We kind of knew it was coming because they (ISM) were planning to issue a statement, and they were declining to verify or debunk the number that they had out at that time, which was 56. We were, like, ‘Oh, man.’ It’s just embarrassing. It was a surprise in a way.”
It’s moment like these when nothing beats working in a newsroom.
We asked Nagi if he was a follower of Stone & McCarthy’s Twitter account.
“I was not. I am now.”
By the way, today we’ll get U.S. factory orders at 10 a.m. EDT.