For a better browsing experience, please upgrade your browser.


As we head toward Election Day, Republicans continue to hold the edge. The GOP has a 68.3 percent chance of taking back the Senate in the latest FiveThirtyEight Senate forecast. The polling out Friday shows why Republicans have that lead but also why Democrats are still in the game.

SENATEUPDATEIowa is key. Democrats probably need to win there to have a chance of holding the chamber. Four Iowa polls were released in the past day, and they tell a consistent story: This is a close race, but one where Republican Joni Ernst holds a slight edge over Democrat Bruce Braley. Ernst led by 2 percentage points per CNN, 1 percentage point per Fox News and Rasmussen Reports, and was tied according to Ipsos.

The public polls in Iowa have been incredibly consistent. In October, the median nonpartisan sponsored poll there has shown Ernst with a 1-point lead. She has led in 12 of the polls taken during the month, trailed in two and tied in three more.

Braley has a shot to win, but you’d rather be Ernst; she’s a 64 percent favorite.

Iowa’s mirror image is North Carolina — a very close race featuring a very consistent lead, this time for the Democrat. If Republican Thom Tillis wins in North Carolina, it’s likely to be a long election night for Democrats. On Friday, CNN found Democratic Sen. Kay Hagan up 2 percentage points; Fox News had her up 1 point.

The new polls match the old polls. The median nonpartisan sponsored poll in North Carolina in October has given Hagan a 1.5 percentage-point lead. She has led in 12 polls, trailed in four (though one of those was by Vox Populi Communications, a pollster with a heavy Republican house effect) and been tied in two more. Hagan, with a 70 percent chance of winning, looks in slightly better shape than Ernst, but not by much.

The model is much more uncertain about Kansas than Iowa or North Carolina. Friday’s Fox News poll had independent Greg Orman ahead of Republican Sen. Pat Roberts in Kansas, 45 percent to 44 percent. Our model has Orman with a 53 percent chance of winning. Kansas is by far the closest race.

One of the ways the model measures the margin of error for a forecast is by looking at the disagreement between the polls and the “fundamentals,” a measure that includes a state’s partisan history and the ideology of the candidates (a liberal candidate is less likely to win in a conservative state and vice versa). And the polls and fundamentals don’t agree in Kansas. The median of the 14 polls taken in October shows Orman up by 1.5 percentage points. The fundamentals project Roberts to win by 26.3 points. That gap, and the fact that 10 percent of voters are still undecided in Kansas, translate into a lot of uncertainty.

A different type of uncertainty is found in Alaska. Rasmussen Reports’ latest poll has Republican Dan Sullivan leading Democratic Sen. Mark Begich, 47 percent to 42 percent. FiveThirtyEight’s model has the race considerably closer: Sullivan ahead by 2.4 percentage points and with a 71 percent chance to win.

As my colleague Nate Silver and I pointed out, the model is unsure about Alaska due to a lack of data and a history of bad polling. Only seven nonpartisan sponsored polls were conducted in Alaska in October. Five showed Sullivan ahead by 3 percentage points or more. The other two had Begich up by 7 percentage points or more. Only one of these seven polls was conducted by a pollster who rates higher than a C+ in the FiveThirtyEight Pollster Ratings, and that poll (by CNN) was taken at the beginning of the month.

Sullivan is flying first class. Begich is in coach. But the weather is bad, and the pilot just came on the intercom to say there won’t be any beverage service and announce the in-flight movie will be “Batman and Robin.”

Finally, one race that shouldn’t keep us up late is Virginia. Democratic Sen. Mark Warner is very likely to beat Republican Ed Gillespie. Yes, two polls released Friday showed Warner up by only 4 and 7 percentage points. The one that had him up 4 points comes from the aforementioned Vox Populi Communications (heavy Republican house effect). Warner has led in every survey taken of the race, and Vox Populi Communications’ 4-point margin represents his smallest lead. FiveThirtyEight’s model has Warner with a 99 percent chance of winning.

Overall, Republicans are leading in the races they need to be leading in. Although, as Friday’s polling demonstrates, they aren’t leading in enough states by a wide enough margin to feel overly confident about taking control of the Senate.

Check out FiveThirtyEight’s latest Senate forecast.


Comments Add Comment

It was one of the biggest moments in Cleveland sports history. (Seriously!) LeBron James’ first game back as a member of the Cavaliers, wearing that familiar No. 23 jersey. James returned with great fanfare Thursday night, leading his new teammates onto the floor against the New York Knicks. Those same Knicks were fresh off being routed by the Chicago Bulls in their 2014-15 season debut, and the Cavaliers — expected to be the best team in the Eastern Conference (not just by us, either) — were at home, in front of one of the most fired-up crowds imaginable.

Piece of cake for the Cavs, right?

Not quite. After a hot start (they led the Knicks by 11 early in the second quarter), Cleveland struggled to make shots and to keep New York from doing the same, eventually losing 95-90 in a game they were favored to win by 13.

It’s only one game, the first of 82 in what is effectively the longest regular-season in sports. But it’s fair to ask what went wrong, and what it means for the Cavs.

In the game, Cleveland had an offensive rating (points scored per 100 possessions) of 108.3 and a defensive rating (points allowed per 100 possessions) of 114.3. Just as a point of comparison, last season’s Knicks scored 109.3 points per 100 possessions and allowed 110.6, but the Real Plus Minus (RPM) ratings of New York’s roster this season would imply a decline of 2 points per 100 possessions on offense from last season, and a decline of 1.5 points/100 on defense. In essence, an average team would have scored around 112 points/100 and allowed 107 points/100 against the Knicks, giving the Cavaliers a ratings shortfall of 4 points on offense and 7 points on defense.

Cleveland was expected to be an offensive juggernaut going into the season, so in a sense its performance at that end was the big disappointment. But it’s difficult to imagine a LeBron James-led team not being an offensive powerhouse; a James-led team hasn’t ranked outside the NBA’s Top 8 in offensive rating since 2007-08. (And that’s not even considering the teammates he now has in Kevin Love and Kyrie Irving, among others.) The bigger concern is that Cleveland’s defensive struggles against New York give credence to the most reasonable fears people had about the Cavaliers over the offseason — that their defense would not be up to par.

Anderson Varejao is an elite defender by plus/minus metrics, and he logged starter’s minutes for the Cavaliers last night. But although James has a strong defensive reputation as well — he’s made the NBA’s All-Defensive squad in some form (first or second team) every year since 2008-09 — his RPM defense ranked just 24th among small forwards a year ago. Meanwhile, scouts for years have questioned Love’s defense, even though his defensive RPM is surprisingly positive. And Cleveland’s other top minute-earners Thursday night — Irving, Dion Waiters, and Tristan Thompson — are defensive disasters whether you look at the numbers or just watch the games.

If ever there was an area where Cleveland’s new super team would struggle, it would be on defense, and Thursday night’s outing against the Knicks only reinforces that fear.

Cavaliers coach David Blatt is new to the NBA, and his defensive system is still a work in progress. Two of his team’s biggest issues Thursday night were transition defense and defending spot-up shooters, both areas that can be improved with better defensive cohesion, communication and awareness.

It’s unlikely the Cavaliers will play as badly on defense all season as they did against New York. But in a Bayesian sense, it does give us one new data point to reinforce doubts about the defensive quality of the group Cleveland has assembled.

Comments Add Comment

Two states — Oregon and Alaska — and Washington, D.C., are voting Tuesday on ballot measures to legalize marijuana. Florida is voting to legalize the use of marijuana for medical purposes. On the heels of successful legalization efforts in Colorado and Washington state in 2012, activists are hoping to pick up at least one win in a tougher year — there are midterm elections rather than a presidential election.

The voting blocs most sympathetic to marijuana legalization — younger voters and people of color — turn out less for these off-year elections than for presidential fights. Still, Florida and Alaska have high-profile statewide offices up for grabs in tight races — governor and senator, respectively — so it’s possible there may be some turnout advantage.

But realistically, the Florida ballot measure doesn’t have much hope because it’s a constitutional amendment that requires 60 percent of the vote. “Otherwise,” said Brian Vicente, a lawyer who worked on the legalization effort in Colorado in 2012, “it would be a slam dunk.”

Vicente said he’d be thrilled with two victories in 2014 but sees the climb as uphill. In Colorado, the 2012 effort was helped by the Obama turnout machine.

As for Oregon and Alaska — which have medical cannabis and are voting on recreational legality — there’s a solid possibility of wins.

Polls in Oregon, perhaps the best pickup opportunity, have been tight. A SurveyUSA poll conducted for Portland’s KATU television station in late September found 44 percent of likely voters breaking for legalization and 40 percent against. And an Elway Research poll released Tuesday found 44 percent for and 46 percent opposed.

In Alaska — which, in fairness, is a pain to poll — the pro-legalization campaign is down 10 percentage points, according to a Dittman Research poll from Oct. 8.

Leading the effort to defeat the legalization ballot measure in Alaska is former state Democratic chairwoman Deborah Williams of the group Big Marijuana Big Mistake. “The main thing we had to do was counter the inevitability argument,” she said.

Williams’s group opposes Colorado-style commercial availability in Alaska, which has made medical marijuana available. Although the organization hasn’t run any commercials — Alaska airtime is particularly expensive because of the close Senate race, she said — it has tried to reach voters through media and in-person interaction.

Taylor Bickford, the director of Alaska operations at the political communications firm Strategies 360, is working on the pro-legalization campaign. Bickford said the campaign is working on targeting young voters: “Young people are overwhelmingly likely to vote yes, and so we are focusing on GOTV [get out the vote] with that group.”

Bickford said the persuadable voters are the middle aged. Indeed, Vicente added, that group comprises the main swing voters in marijuana legalization efforts, but he specifically mentioned women ages 30 to 50. The members of this group are “soft supporters” of legalization, he said, in that they’ll generally support the measure unless hit with substantial negative messaging. And the campaigns of 2012 and their results seem to back this up. Win moms, the logic goes, and you’ll legalize weed.

Just look at this ad from the 2012 Washington state campaign, which approximates the platonic ideal of this argument:

The exit polls bear out the idea that white women are bellwethers on marijuana. Here’s the data from the 2012 Fox News exit polls in Oregon, Washington and Colorado showing support for legalization overall and among white women specifically.


The only state where legalization didn’t win a majority with white women in 2012 was also the state where marijuana legalization failed.

Comments Add Comment

Voter access has been a hot topic recently. Earlier this month, the Supreme Court upheld a Texas law passed in 2011 that requires voters to present photo identification when casting their ballots. It may seem like voter access has only become an issue in the past two years, but the numbers show that the current debate is part of a longer trend toward more restrictive laws.


Some states have had ballot initiatives requiring voters to present some form of ID, and those initiatives were subsequently voted down. Many proposals didn’t even make it that far. So will this trend continue? Or is the tide changing? The answer is … complicated.

According to the National Conference of State Legislatures, 24 states considered for Tuesday’s election legislation to implement stricter voter ID standards. Of those, 12 states’ proposals garnered enough support to make it on the ballot. On the flip-side, 11 states considered proposals to loosen restrictions or repeal them altogether — five made it on the ballot. The NCSL has detailed data since 2012. Here’s a look at what we know so far:


The number of states with proposals to loosen voter ID requirements seems to be experiencing a slight upward tick, while the percentage of those that made it on the ballot is growing — 22 percent in 2012, up to 46 percent this year.

Then there are the courts, where results are mixed. The Supreme Courts in Pennsylvania and Arkansas have struck down more restrictive voter ID laws this year. While the U.S. Supreme Court has held that such laws in Indiana and Texas are constitutional. A battle in Wisconsin was put on hold this month, when the U.S. Supreme Court blocked the law from being implemented for this year’s election but left room for its implementation in the future.

As the number of states with some kind of voter ID requirement grows, there will (theoretically) be a decrease in proposed legislation for stricter standards. The real question is whether the increase in proposals for loosened standards will continue, and whether those proposals will be implemented.

Comments Add Comment


The GOP’s chances of winning the Senate are 68.5 percent, according to the FiveThirtyEight forecast, its highest figure of the year.

Among the 20 new polls released Thursday — it looks like there will be no Great Poll Shortage after all — two were principally responsible for Republican gains. The first was in Kentucky, where a SurveyUSA poll for the Louisville Courier-Journal had Republican incumbent Mitch McConnell ahead 48-43 over Democrat Alison Lundergan Grimes. The poll represented a shift from SurveyUSA’s previous two polls, which had a 1 percentage-point lead for McConnell and a 2-point lead for Grimes. With SurveyUSA (a highly rated pollster) now more in line with other polls of the state, we have a clearer story in Kentucky. It’s one that probably ends in a victory for McConnell, whose chances of winning are up to 87 percent.

The other poll was in Arkansas, which hasn’t been surveyed as often as other key Senate races. That poll, from the University of Arkansas, found Republican Tom Cotton up by 13 percentage points over Democratic Sen. Mark Pryor. No other poll of the state has shown Cotton with a double-digit lead, but he hasn’t trailed in a nonpartisan poll since Sept. 22. When the choice is between polls that show a candidate with a small lead and polls that show him with a large lead, he’s usually in good shape just a few days before Election Day.

Speaking of which, it’s not too early to look ahead to election night (along with our partners at ABC News, we’ll be covering everything; we hope you’ll join us). The number you’ll be hearing about all night is six — as in, Republicans need to net six seats from Democrats to win control of the Senate.

As we’ve pointed out before, the “net” part of that phrase is key. Republicans will have to win more than six Democratic-held seats if they lose a couple of their own. Their incumbent in Kansas, Sen. Pat Roberts, is only even-money to win re-election (although there’s a chance independent Greg Orman, even if he wins, could caucus with Republicans). The GOP candidate in Georgia, David Perdue, is ahead by only about 1 percentage point against Democrat Michelle Nunn, and that race could go to a runoff. McConnell is likely, but not certain, to survive.

More often than not, in fact, Republicans’ magic number will not be six seats. Instead, in more than 60 percent of our simulations the FiveThirtyEight model ran Thursday night, Democrats won at least one GOP-held seat. And about 20 percent of the time, they won at least two Republican seats. So the GOP’s magic number could be seven, eight or even nine seats instead.


Fortunately for Republicans, they have an abundance of Democratic targets. Running in order from the most likely to least likely pickups:

1. Montana (99.8 percent chance of a GOP win)

2. South Dakota (99 percent)

3. West Virginia (98 percent)

South Dakota looked competitive for a fleeting moment. But otherwise these races, red states where Democratic incumbents have retired, have been in the Republican column all year. The GOP is almost certain to win them, giving it a three-seat head start.

4. Arkansas (90 percent chance)

Pryor trails by 5 or 6 percentage points in the polling average. Few candidates have won a Senate election when trailing by that much so late in the race.

5. Louisiana (77 percent chance)

The Democratic incumbent, Mary Landrieu, also trails by about 5 percentage points — but those are in polls of the Louisiana’s probable Dec. 6 runoff against the Republican Bill Cassidy. She’s ahead in polls of Tuesday’s vote, when there are multiple Republican candidates on the ballot, although she’ll need a minor miracle to get to 50 percent and avert a runoff. Still, because of the runoff dynamic, this race can’t quite be placed into the same category as Arkansas yet.

6. Colorado (75 percent chance)

The polls of Colorado have been subject to more debate than those in any other state. My colleague Harry Enten had a counter-critique last week of those who claim the Colorado polls are “skewed” against Democrats, but let me add just one other simple-minded observation: There are a heck of a lot of polls in Colorado. There have been 23 in October representing 18 distinct polling firms. Excluding polls from firms designated as partisan by Huffington Post Pollster, there have been 14 leads for Republican Cory Gardner compared to just 2 (each by a single point) for Democratic incumbent Mark Udall.

Udall can win. But his chances are inferior to Democrat Michael Bennet’s in 2010, who trailed by a slightly narrower margin in the polling average. And unlike in 2010, when Republicans nominated a problematic candidate in Ken Buck, the GOP has a fairly good one this year in Gardner. (The FiveThirtyEight fundamentals calculation, which doesn’t depend on polls, slightly favored Bennet in Colorado in 2010 but slightly favors Gardner this year.)

7. Iowa (67 percent chance)

True, there’s not always a bright line between states where the polls would have to be wrong for the trailing candidate to prevail and those where it’s not as clear whom the polls have ahead. But Harry and I tend to think of Colorado as falling just toward one side of that line and Iowa just toward the other side.

A handful of recent polls in Iowa have the race tied or Democrat Bruce Braley up by a percentage point. More have Republican Joni Ernst ahead, but they’re not showing as large a lead for her as polls of Colorado are sometimes showing for Gardner.

8. Alaska (67 percent chance)

Alaska is a more emphatic example of this distinction: It’s not clear who’s leading in the polls. Our polling average has Republican Dan Sullivan slightly ahead, and the fundamentals of the race and the state’s history of Republicans outperforming their polls also point his way. This race isn’t really “too close to call,” it’s “too uncertain to call.”

9. North Carolina (32 percent chance)

One backup option for Republicans is North Carolina, where GOP candidate Thom Tillis has gradually drawn closer in the polls against Democratic Sen. Kay Hagan. It still might be a case of too little, too late — especially in a state with lots of early voting — but Tillis’s chances of winning are at their highest point since early September.

10. New Hampshire (17 percent chance)

The notoriously volatile University of New Hampshire poll gave Democrat Jeanne Shaheen an 8-point lead Thursday. She’s unlikely to win by that much — but if you want to discard it as an outlier, you should probably throw out the handful of polls that show Republican Scott Brown ahead, too. The consensus points to a narrow Shaheen lead. Brown has a chance, but our model shows it will mostly be in cases where Republicans are having a strong night nationally and New Hampshire is just one among many Democratic states to fall.

This is about it in terms of viable opportunities for the GOP. There are a few states (Michigan, Minnesota, New Mexico, Virginia) where our model gives Republicans between a 1 and a 4 percent chance of a pickup — but those would be once-a-decade type upsets, like John Engler winning Michigan’s gubernatorial race in 1990 when down by double digits in some polls. Republicans have plenty of more realistic options.

So, what about that tantalizing distinction I teased at before? Would the polls have to be “wrong” for Democrats to hold the Senate?

In my view, we’re not quite at that point yet. There is a reasonably clear consensus among nonpartisan pollsters about which candidate leads in Republicans’ easiest six pickups: Montana, South Dakota, West Virginia, Arkansas, Louisiana and Colorado. Indeed, the Republicans have about an 80 percent chance of winning at least six Democratic-held seats (not necessarily these ones exactly).


Again, the GOP’s magic number might not be six seats; the polling picture is murky in Georgia and Kansas. They might need seven or eight instead, which would require the party winning a state like Iowa or Alaska, where things are more uncertain.

Some of Democrats’ roughly 30 percent chance of holding the Senate represents these cases: where they win squeakers in a number of the closest races and Republicans fall just short of a majority. The other portion reflects cases where there is more of an across-the-board polling failure. This is plausible — sometimes almost all of the competitive races break in the same direction. But an anti-Republican “skew” is just as plausible, in which case Republicans could finish with 54 or 55 seats.

Which states to watch over this final weekend? I’d point to three: Alaska, Iowa and Kansas. Any polling at all in Alaska would be helpful. Iowa, depending on the final few polls there, could wind up anywhere from a true tossup to a case more like Colorado. In Kansas, Roberts’s position is improved from a few weeks ago, but it isn’t clear whether he’s gaining ground or has stalled out. In most of the other states, the possibility of a runoff limits how much the polls can tell us, or we have so much polling that no one further poll is going to move the needle that much.

Check out FiveThirtyEight’s latest Senate forecast.


Comments Add Comment

If you really want to scare someone this Halloween, you should think about dressing up as a microphone — 21 percent of Americans say they are afraid of “public speaking/performance.” Alternatively, you could just spend Friday evening forcing fellow partygoers to engage in conversation, since 13 percent fear talking with strangers, according to a survey of 9,282 adults cited in a 2008 study.


Interestingly, when similar questions were posed by the World Health Organization in 20 countries (including the U.S.) in 2011, researchers found that the anxieties charted above were more prevalent in developed countries than in developing ones — with the exception of using a public bathroom.

But it’s hard to find a costume that conveys a fear like “important exam/interview” so I looked at another study, published in 2007, which looked at objects as well as social situations. This survey was even bigger — it asked 43,093 U.S. adults about 12 “specific phobias” as defined by the Diagnostic and Statistical Manual of Mental Disorders. Respondents were asked if they felt nervous, frightened or tried to avoid any of the items listed, when they thought about or were confronted with the items.

The most commonly reported phobia was “insects, snakes, birds or other animals” (a phobia for 4.7 percent of respondents) followed by “heights: e.g. tall buildings, bridges or mountains” (reported by 4.5 percent). It’s also worth noting that there can be overlap — most respondents who reported one phobia also listed a second, third or more from the list.


These are phobias, a type of anxiety disorder, which the American Psychological Association describes as “intense and irrational fear,” so the reported percentages are far lower than the fears described in the first chart above. It also means that while dressing up as a cockroach dentist would be a highly effective Halloween costume, it could also be an extremely cruel one.

A YouGov poll from March of 989 adults’ fears offers some more reasonable costume choices. Snakes, spiders and mice all frighten Americans, especially women, who were more likely to report a fear of every item listed (perhaps because they are more willing to admit feeling fearful).

This all suggests that people are choosing their Halloween costumes very badly if their aim is to scare others. As Quoctrung Bui at NPR shows, for the past five years “witches” have been the most popular adult Halloween costume. True, none of the surveys mentioned here asked adults if they were frightened of witches (or pirates, or vampires, for that matter) but clowns do come up. The 11th-most popular Halloween costume in 2011 was feared by just 13 percent of respondents in the YouGov survey — after dogs, blood and darkness. So before you do the clichéd messy lipstick and wig routine this Friday, remember the data suggests you’re more slapstick than scary.

Comments Add Comment

Dear Mona,

What percentage of women have breast implants?

Marisa, 25, New York

Dear Marisa,

After receiving her breast implants in an experimental surgery, Esmeralda, a dog of unknown breed, chewed at the stitches until they had to be removed. The operation was still deemed a success, and the surgeons, Frank Gerow and Thomas Cronin, began searching for a way to test the procedure on a human female. Not long after, Gerow offered his new silicone products to Timmie Jean Lindsey, a 29-year-old mother of six who agreed to the procedure on the condition that Gerow would also “fix” what really bothered her — her “Dumbo” ears. And so the world’s first breast implant surgery was performed at Jefferson Davis hospital in Houston, Texas, in 1962.

MONALindsey received her breast implants more than 50 years ago, but national statistics on the procedure have only been collected on a yearly basis since the late 1990s. And those statistics only measure the number of procedures that take place; they say nothing about the women receiving those implants, which makes an estimate here all the more shaky. Still, though, I estimate that almost 4 percent of women in America, or one in every 26, has breast implants.

Here’s how I got to that number.

I looked at the number of breast augmentations that have taken place each year since 1997, according to the American Society for Aesthetic Plastic Surgery (ASAPS). Unlike other professional organizations for certified surgeons, ASAPS members specialize in cosmetic surgery, so these numbers, collected from surveys of surgeons, should exclude reconstructive surgeries for women who have lost one or both breasts after mastectomy surgery.


I simply added up all of those procedures (a total of 4,798,349 since 1997) and divided it by the current U.S. female adult population (124,501,374) and arrived at the estimate that 4 percent of American women alive today have had breast implants.

Again, 4 percent is a very rough estimate, for a number of reasons:

  • I assumed that each procedure for “breast augmentation” involved implants. In reality though, a fraction of these will use a woman’s own fat rather than implants to increase the size of the breast. In “fat transfer breast augmentation,” liposuction is used to move fat from one part of the body to the breasts (breast-lift surgery is counted as a separate category). That means that 4 percent is an overestimate.
  • I assumed that one procedure equates to one woman. That means this is an overestimate because in reality some of these procedures will be revision surgeries (i.e. intended to fix or improve the initial result).
  • I assumed that none of the women who received breast implants since 1997 has since died. That means that this is an overestimate.
  • I assumed that each woman kept her breast implants. Removal of breast implants counts as reconstructive surgery, so I contacted the American Society of Plastic Surgeons (ASPS) which, unlike the confusingly similar ASAPS, doesn’t just collect statistics on cosmetic surgery. ASPS statistics show that in 2013 alone there were 23,770 breast implant removals. That means this is an overestimate.
  • I assumed that these procedures were performed on U.S. women. In reality, some of those operations would have been performed on women who have subsequently left the United States — and some women currently living here will have received their breast implant surgery abroad. There is scant information about the scale, direction of travel or most common operations performed in cosmetic surgery tourism, so I can’t say whether that makes this an overestimate or an underestimate.
  • My total doesn’t include women who got implants before 1997. That means Timmie Jean Lindsey (now an 82-year-old great-grandmother) and thousands that came after her aren’t counted. Even as early as 1989, a survey of 40,000 American households estimated that there were 815,700 women with breast implants in the U.S. That means this is a considerable underestimate – which I can only hope outweighs the factors above.

There’s another trend here, though. For every two women who have had breast implants since 1997, there was one woman who went under the knife to have her breasts reduced. It’s also easy to overlook the fact that men have breast reductions, too (it may even come as news to some that men have breast tissue). In addition to the numbers listed here, a further 22,939 breast-reduction procedures were carried out on men in 2013 to address gynecomastia, or enlarged breast tissue.


Although breast implants remained the most common surgical cosmetic procedure for women last year, coming up from behind is a big rise in buttock augmentation.


I have no idea how many of the 4 percent of women that I estimate have breast implants also have their bums augmented, or any other cosmetic procedure for that matter. Nor can I tell you much about the number of men who undergo multiple procedures. But there is data on the reasons why people choose to get cosmetic surgery, which (while we’re talking about possible vanity) you can discover in my interview with NPR last week.

Hope the numbers help,


Have a question you would like answered here? Send it to or @DataLab538.

Comments Add Comment

It would have been close. Alex Gordon might have scored, particularly if he’d been in the mindset to do so all along. Or maybe not. I’m sure there will be Zapruder-film-type breakdowns, and I’ll look forward to seeing them. It would have been one hell of a moment: Gordon, 220 pounds, who looks like he could have been a strong safety at the University of Nebraska, bearing down on Buster Posey, the catcher whose season-ending injury in 2011 helped inspire baseball’s home-plate collisions rule.

Game 7 will leave us with that sense of what might have been. Partly because it involved the Kansas City Royals, who were making their first World Series appearance since 1985. But mostly I’m referring to that penultimate play: When Gordon hit what was officially scored as a single and wound up on third base because of defensive miscues by San Francisco Giants outfielders Gregor Blanco and Juan Perez. It seemed to take an eternity — it was actually just 13 seconds — but I was surprised that Gordon wasn’t rounding third base by the time the TV cameras returned to the infield.

Here’s what I know: Gordon should have tried to score even if he was a heavy underdog to make it. It would have been the right move if he was safe even 30 percent of the time.

Between 1969 and 1992 — I’m using this period because it better approximates baseball’s current run-scoring environment than the offensive bubble of the 1990s and aughts — a runner scored from third base with two outs about 27 percent of the time, according to the tables at We should probably round that down a bit in this example. The Royals had Salvador Perez at the plate — a league-average hitter — and the light-hitting Mike Moustakas due up after that.

More importantly, they were facing Madison Bumgarner. That Bumgarner had been so dominant in the World Series is not as relevant as you might think. There’s extremely little evidence for a “hot hand” in pitching: In-game performance tells you next to nothing about how the pitcher will fare in future at-bats. Instead, you should look toward longer-term averages. Still, I feel comfortable asserting that Bumgarner was an above-average pitcher at that moment: Certainly not the first guy you’d want to have on the mound if you were the opponent. So let’s round that 27 percent down to 25 percent.

So, Gordon should have tried to score if he had even a 25 percent chance of being safe?

It’s just a touch more complicated than that. With the Royals down 3-2, Gordon represented the tying run rather than the winning run. If he’s thrown out at home, the game’s over; it forecloses on the possibility of Perez scoring as the winning run, like with a walk-off homer. What was the probability of that? Perez homered in about 3 percent of his plate appearances this season, but he could also have scored in other ways — by doubling, for example, and then scoring on a base hit by Moustakas. We can turn to Tangotiger’s tables again, which suggest that a league-average batter has about a 6 percent chance (I’m rounding down slightly) of eventually scoring from home with two outs.

So, after Gordon holds at third, he has a 25 percent chance of scoring. Six percent of the time, Perez (or pinch-runner Jarrod Dyson?) also scores, and the Royals win outright. The other 19 percent of the time, Gordon is the only Royal to score in the ninth and the game goes to extra innings. If we assume the Royals are even money to prevail in an extra-inning game, their chances of winning at that point are:

6% + (19% * 50%)

That works out to 15.5 percent. Not coincidentally, this matches FanGraphs’ in-game win probability for the Royals (after Gordon held at third) almost exactly.

What if Gordon rounds third and tries to score? If he’s successful even 30 percent of the time, the Royals’ win probability is at least 15 percent — a 30 percent chance of Gordon scoring, multiplied by a 50 percent chance of the Royals winning in extra innings. But it’s slightly higher than that. The 30 percent of the time that Gordon scores, Perez still has his 6 percent chance of scoring the winning run in the ninth. That brings the Royals’ overall win probability up to about 16 percent.

We’re splitting hairs. The point is that if even Gordon had been a 2-to-1 underdog to score, he should have tried.

These decisions can be counterintuitive. Sometimes a strategy that’s successful less than 50 percent of the time — like splitting eights in blackjack — is still the right move because the alternative is even worse. In this case, the alternative involved trying to score against Bumgarner with your catcher at the plate and two outs, and then having to prevail in extra innings.

It would have made for one of the best plays in baseball history. We’re talking about the tying run with two outs in the bottom of the ninth inning in Game 7 of the World Series: Even a sacrifice fly can be thrilling under those circumstances. But this would have been in a league with Bill Mazeroski and Kirk Gibson and Bill Buckner: under serious consideration for the greatest play of all-time. (The play already had a little Buckner in it, with Blanco’s and Perez’s misplays in the outfield.)

Unlike any of those moments, it would have involved an incredibly gutsy decision. It’s an extraordinary play if Gordon scores. It’s an extraordinary play if there’s a collision at home plate — and baseball needs to decide whether to invoke the “Buster Posey Rule.”

And if Gordon were thrown out, it would have been the most extraordinary way to lose a game in the history of baseball.

CORRECTION (Oct. 30, 11:14 a.m.): A previous version of this article misstated the first name of a Kansas City Royals catcher. He is Salvador Perez, not Santiago Perez.

Comments Add Comment


Republicans got favorable news in Georgia on Tuesday, where a Monmouth University poll put their candidate, David Perdue, eight points ahead of the Democrat Michelle Nunn. The poll is something of an outlier relative to the consensus — nonetheless, the last seven surveys we’ve added to our database show a tie or a Perdue lead, reversing a string of polls that put Nunn ahead.

Perdue now leads Nunn 49 percent to 48 percent in the FiveThirtyEight forecast — with the Libertarian Amanda Swafford, whose vote share has declined slightly in the most recent surveys, projected to get 3 percent of the vote. Perdue’s chances have improved of getting to 50 percent of the vote on Nov. 4, which would allow him to avoid a runoff.

Still, a runoff in Georgia remains more likely than not. Our model gives Perdue about a 30 percent chance of winning outright on Nov. 4, while Nunn is down to only about a 10 percent chance of doing so. The rest of the time, the race will be runoff-bound.

As we’ve written before, Georgia is not the only race that might require “overtime.” Our model estimates that while Republicans have a 64 percent chance of winning the Senate eventually, there’s only a 27 percent chance they’ll be able to claim their victory within the first 24 hours or so after polls close on Nov 4. Democrats are even less likely to win a quick victory — they have just a 12 percent chance.

The other 60 percent of the time, it will take days or weeks to sort everything out. The chart below lists the states that are the biggest part of the problem.


Louisiana tops the list. Its Nov. 4 “jungle primary” ballot includes eight candidates; the top two will advance to a Dec. 6 runoff if none wins a majority on Nov. 4.

Previously, we’d been assuming Louisiana was certain to go to a runoff. That was a slightly lazy assumption, so as we’ve done for Georgia, we’re now modeling both the primary and runoff ballot in Louisiana.

The Democratic incumbent, Mary Landrieu, has held the plurality in most polls of the primary ballot, usually with numbers in the low 40s. It’s a long way to get from the low 40s to 50 percent, especially with so many candidates on the ballot, but there are a fair number of undecided voters. With all votes allocated our model projects Landrieu to get about 45 or 46 percent of the vote on Nov. 4. It has Republican Bill Cassidy at 41 or 42 percent, another Republican, Rob Maness, at about 10 percent, and the other candidates at 3 to 4 percent of the vote combined.

It’s still not quite clear where Landrieu would pick up the rest her votes. Her approval ratings are stuck at about 40 percent and her two major challengers are conservative Republicans — there’s not a liberal spoiler in the race. But Louisiana can be a tricky state to poll; on the basis of the intrinsic uncertainty in the contest, the model gives her about a 6 percent chance of somehow getting to 50 percent on Nov. 6. It also gives Cassidy a one percent chance of doing so; he’d need a number of Maness voters to defect to him at the last minute. Still, a runoff is very likely.

Alaska is another huge problem for those hoping for a quick resolution on Nov. 4. In fact, its polling places don’t close until 1 a.m. Eastern time on Nov. 5.

The polls in Alaska haven’t been telling a consistent story lately, but the race seems to have tightened, and the Democratic incumbent Mark Begich’s chances of surviving have improved. An internal poll put out by the Republican Dan Sullivan’s campaign on Tuesday showed him with a 4-point lead, while a poll for the Democratic Senate Majority PAC showed a tied race. Our forecast is somewhere in between, showing a 1- or 2-point lead for Sullivan but with a high degree of uncertainty.

It’s entirely possible that Sullivan could still win by a margin in the mid-to-high single digits — or that Begich could do so. But if the margin is anywhere in between, we might not know the outcome for days or weeks. In 2008, when Begich beat the Republican Ted Stevens by 1.2 percentage points, most news organizations did not declare him the winner until Nov. 19 — 15 days after polls closed. Likewise, in 2010, when the Republican incumbent Lisa Murkowski held her seat as a write-in candidate by 2.5 points over fellow Republican Joe Miller, it took about two weeks to declare a winner.

Not all races take quite that long to count. In Alaska’s senate race in 2004, the Democrat Tony Knowles conceded to Murkowski two days after the election; he eventually lost to her by 3 percentage points. In 2006, Sarah Palin won Alaska’s governorship by 7 percentage points and most news organizations called the race the morning after the election.

Still, our assumption is that either Begich or Sullivan would have to win by at least 5 percentage points for Alaska to be declared in time for the evening news on Nov. 5. Our model says there’s about a 30 percent chance of that, and a 70 percent chance that things will take at least a little longer.

We already discussed Georgia, so let’s skip ahead to Kansas. It presents two problems. First, the independent Greg Orman has about a 50 percent chance of winning. Orman takes center-left policy positions, but he hasn’t said whom he’ll caucus with — in fact, he’s said it might depend on which party controls the majority. His decision might take a long time to sort out, especially since it might not be clear which party has the majority in the first place.

Alternatively, it’s possible the Republican incumbent Pat Roberts will prevail, but by such a close margin that the race will require a recount or otherwise take a long time to call. Roberts trails Orman by just 0.3 percentage points in our latest forecast.

Our model assumes that races go to “overtime” whenever the vote finishes within 0.5 percent. That threshold, 0.5 percent, is a trigger for automatic recounts in some states. In other states, a recount wouldn’t necessarily be required. But mail and absentee ballots can take some days to count. There can be questions about provisional ballots. The lawyers can get involved. Under these circumstances, news organizations would usually hold off for at least a few days before declaring a winner.

Any of the close races could potentially fall into this category. There’s about a 13 percent chance that Iowa will finish within 0.5 percent, a 12 percent chance in North Carolina, an 11 percent chance in Colorado, and so forth. With so many tight races this year, it’s better than even money that at least one state will fall into Recountland.

There are two pieces of hidden good news for Republicans. First, if they’re declared the winner without going to overtime, that likely means they’re headed to a very clear victory. In those simulations where they won without going to overtime, they eventually finished with an average of 54 seats once all states were counted.

Second, Republicans are about 60-40 favorites to eventually win the Senate if the contest does go to overtime. This is largely because of Louisiana, where Cassidy would be a reasonably clear favorite in a potential runoff.

But here’s one way to think about next Tuesday. There’s a 25 or 30 percent chance of a big, sweeping win for Republicans. There’s a 10 or 15 percent chance the Democrats retain the Senate with surprising ease. And there’s a 60 percent chance that we’ll be sweating out the races on a state-by-state basis, possibly for weeks to come.

Check out FiveThirtyEight’s latest Senate forecast.


Comments Add Comment

The Federal Reserve announced on Wednesday that it would bring its nearly 2-year-old bond-buying program, dubbed “QE3,” to an end; around $1.6 trillion was spent to stimulate the U.S. economy over the life of the program, and we’re not sure what it bought us.

In halting QE3, the Fed also gave the clearest sign yet that it’s OK for inflation to fall below its 2 percent target. And it flatly said it’s not so worried about current inflation readings despite evidence that low inflation can slow consumer spending and business investment.

At least one senior member of the Fed’s policy-setting body is concerned: Narayana Kocherlakota, president of the Minneapolis Fed. Kocherlakota dissented in Wednesday’s vote, citing the “continued sluggishness in the inflation outlook and the recent slide in market-based measures of longer-term inflation expectations.”

Kocherlakota’s reference to inflation expectations is important: It’s not just current inflation that worries economists, but beliefs about future inflation, too. We can gauge inflation expectations by looking at the premium investors pay for inflation-protected Treasury bonds. The future rate of inflation, according to these measures, has fallen to around 2.3 percent over the last month from 2.6 percent in August and 2.8 percent at the start of 2014.

Why should you worry that prices won’t increase fast enough? Outright deflation can be poisonous to an economy, as it was for the United States during the Great Depression. If money is worth more tomorrow than it is today, then consumers will hold off spending, reducing demand, sparking further price cuts and thus triggering a cycle of falling prices and stagnating growth.

But even short of deflation, low inflation can wreak havoc, too. In last week’s issue of The Economist, Greg Ip outlined the dangers of low inflation. The chart below, taken from Ip’s briefing, shows just how widespread low inflation has become across rich countries.


The consumer price index is currently running about 1.8 percent higher than its level a year ago — below the Fed’s target of 2 percent. But the Fed doesn’t appear to care (aside from Kocherlakota). In its statement on Wednesday, the Fed’s policy committee noted it thinks “the likelihood of inflation running persistently below 2 percent has diminished somewhat since early this year.”

Let’s hope this risk has receded. But the numbers don’t fully bear that out.

Comments Add Comment

More >

Powered by VIP