🏅 Top traders

*sad Al Quinn noises*

@AlQuinn Thanks for the mana!

299.170

😭

S&P went to the moon today. Those in the know have snooped the inflation report! Join the NO train, or have fun staying poor!

@AlQuinn nice try, but bond yields would have fallen at the same time if the bump in equities was caused by changes in inflation or interest rate expectations. (They did fall a smidge - but not at the same time).

@chrisjbillington Maybe I forgot to take my meds. In fact, I think it will come in at 6.1%.

I don't really understand this market's case for >50% probability here. Trying to work out the math on how CPI-U from Dec 22 of 296.797 gets to 298.016, just some basic base rates:

If we take the whole year's average percent change for CPI numbers month over month, then it gets over at 298.267.

However, inflation has cooled significantly over the last 6 months, and taking the average percent change for the last 6 months means it's significantly under at 296.879.

I don't see any historical justification to see January as a particular hot month to incorporate that into my base rates. I have also looked into oil prices and while they probably won't cause energy prices to pull down CPI-U as much as in December, they're not increasing either. Overall, there is certainly chance but I would put the probability around 20% just based on this analysis.

Posting this comment mostly because a lot of my uncertainty is wrapped up in the chance I'm misinterpreting this data or just woefully misunderstand something obvious (is there a real-money market for inflation expectations? Extrapolating services inflation increasing with no negative energy contribution?). Curious if anyone wants to correct me, @chrisjbillington maybe?

Potentially throwing away alpha here (I'd like to see more discussion on Manifold, shame it's not particularly incentivised!), but I am basing my betting on the Cleveland Fed inflation nowcast, which has 6.44% YoY, or 0.63% for the month, which would give 298.66 for the index.

I admit it does look strange given the recent trajectory of the CPI, and I'm not 1000% sure the figure in the nowcast corresponds exactly to the index resolution here is based on. But I'm pretty sure!

I've been watching the nowcast the last few months and whilst it has misses, a miss of size 0.44% would be substantial compared to the nowcast's track record over the past year, shown below comparing the final nowcast to the subsequent CPI release (not a completely fair comparison since we are not yet seeing what the nowcasts's final value will be before the CPI release in two weeks, but illustrative):

month nowcast actual err

2022-12 6.64 6.45 +0.19

2022-11 7.49 7.11 +0.38

2022-10 8.09 7.75 +0.34

2022-09 8.20 8.20 +0.00

2022-08 8.24 8.26 -0.02

2022-07 8.82 8.52 +0.30

2022-06 8.67 9.06 -0.39

2022-05 8.23 8.58 -0.35

2022-04 8.14 8.26 -0.12

2022-03 8.41 8.54 -0.13

2022-02 7.81 7.87 -0.06

2022-01 7.26 7.48 -0.22

In addition, Trading Economics' forecast is for 6.3%.

Love the response, very interesting. I also am confused why forecasts seem to differ so much from extrapolating based on recent trends, but you're absolutely right I would trust them over my analysis. Also share your hesitance that the resolution criteria here lines up, I am far from an economist so it's possible there is a mismatch in some vocab that's going over my head.

Regardless, very interesting! Appreciate the generosity with alpha. I'll send you some mana once I have a high enough balance to unlock manalinks.

@chrisjbillington I think there is a fallacy to use the YoY inflation estimates to measure forecasting skill, rather than MoM, since YoY will contain 11 months of actual data, greatly lowering the uncertainty. And since we have 11 months of actual data towards resolution of this, all that matters is how a single month (January) resolves.

Using the current 0.65% MoM nowcast, that would need to be wrong by +24bps for this to resolve NO. Of the last 12 months, the nowcast has been wrong 3 times by this amount (and has been wrong by this amount in 3 of the last 6 months). Looking at the absolute value of the error, the nowcast has been at least this wrong 5 out of the last 12 months. We are probably in a domain that is outside of where there model has been calibrated to be more effective.

I estimate this is closer to 50/50 so got a modest amount of NO.

@AlQuinn Hm, well this market is framed in terms of the CPI-U index level, or equivalently, the YoY change in CPI-U. The MoM figures are seasonally adjusted so it is nontrivial to figure out what MoM figure would be needed resolve this market YES. Are you factoring that in to your thinking?

Furthermore the seasonal adjustment factors are changing with the next release (I think the new seasonal factors have been released, or perhaps leaked, in the last few days). But again this won't affect the YoY figure or the actual CPI-U index level.

The YoY Nowcast would have to be wrong by 0.48 percentage points, which has not happened in the past twelve months. CPI weights are also changing with the next release (this is distinct from seasonal adjustment factors), increasing uncertainty, so we might expect a larger error in the nowcast than most releases. So I think this market is priced about right. If the weights were not changing, I would think maybe ~85%–90%.

@chrisjbillington No, that is a good point. Al Quinn never reads the fine print. Here are MoM for both approaches starting in Dec 2021:

Mo. Raw Adj

Dec 0.3% 0.5%

Jan 0.8% 0.6%

Feb 0.9% 0.8%

Mar 1.3% 1.2%

Apr 0.6% 0.3%

May 1.1% 1.0%

Jun 1.4% 1.3%

Jul 0.0% 0.0%

Aug 0.0% 0.1%

Sep 0.2% 0.4%

Oct 0.4% 0.4%

Nov -0.1% 0.1%

Dec -0.3% -0.1%

LY, the Jan seasonal adjustment amount was ~0.2% (0.8% >>> 0.6%). Is this typical? I pulled all of the archived data to find out:

Year JanAdj

2012 -0.2%

2013 -0.3%

2014 -0.3%

2015 -0.2%

2016 -0.2%

2017 0.0%

2018 0.0%

2019 -0.2%

2020 -0.3%

2021 -0.2%

Overall, there is an average ~20 bps of adjustment for the January reported month. So, yes, this is bad for my initial case, since it means the MoM nowcast (which forecasts the adjusted number) would have to be off by +44bps (assuming no material impact from new seasonal adjustment), and this hasn't happened in the last 13 months.

OTOH, focusing on just unadjusted January MoM figures, just 4 out of the last 10 have exceeded the number necessary for this question to resolve YES:

Year JanRawInf

2013 0.30%

2014 0.37%

2015 -0.47% ???

2016 0.17%

2017 0.58% <<

2018 0.54% <<

2019 0.19%

2020 0.39%

2021 0.43% <<

2022 0.84% <<

Another thing to note is the size of the swing from -0.3% in December to the (required for YES) +0.4% in January in unadjusted inflation. Swings of >70bps have happened 2 times in the last 130 months (Jul>Aug 2012 and Jan>Feb 2015; the latter was a 0.9% increase and looks like a data issue since Jan 2015 had anomalously low inflation, as shown above).

Your comments have pulled my % up somewhat (to 70-75%) but now that it's at 85%, might buy more NO to double-down because there is nothing more embarrassing than admitting you're wrong on the internet!

@chrisjbillington Would love to send you mana for this conversation. Where can I send the mana link?

@DanStoyell much appreciated! I can be reached at chrisjbillington at gmail dot com.