This note was originally published at 8am on December 09, 2014 for Hedgeye subscribers.
“It solved practical questions that were unanswerable by any other means.”
-Sharon Bertsch Mcgrayne
That’s how Sharon Bertsch Mcgrayne summarizes using a Bayesian approach to problem solving in an important book for your risk management library – The Theory That Would Not Die.
“Although Bayes’ Rule drew the attention of the greatest statisticians of the 20th century, some of them vilified both the method and its adherents, crushed it, and declared it dead… In discovering its value for science, many supporters underwent a near-religious conversion yet had to conceal their use of Bayes’ Rule and pretend they employed something else.”
Much like during Bayes’ revolutionary times (18th century), where “mathematics was split along religious and political lines” (pg 4), today we are at the crossroad for the same in economic analysis. If you tell me what someone’s political and/or financial motivation is, I can usually predict their answers. If you’re Bayesian in your approach, your answers are far less certain.
Back to the Global Macro Grind…
We hosted an astute group of RIAs (Registered Investment Advisors) at HedgeyeHQ after the close yesterday where I gave a teach-in on our #process. One of the main examples I like to use is Bayes trying to locate a billiard ball’s location (blindly) on a table…
Each incremental throw gives you more information on where the ball is not. After multiple throws, you can narrow the probability of the ball’s location to a probable range. I call this the Risk Range. And I have no business telling the table where the ball has to be.
What happens if there are multiple competitors at the table, all racing to figure out where the ball is at the same time? That’s what makes a market. Whether you like my competitive style or not, I fully intend on coming out of the room with the ball (read Diary of a Hedge Fund Manager for details!).
If you want to beat your competitors in this game, not only do you need to play your game (read: #process), but you need to understand where the other players at the table are positioned and why. One really simple way to look at this from a consensus hedge fund positioning perspective is through CFTC Non-Commercial net LONG/SHORT futures/options.
I cite the rate of change in this Bayesian information as frequently as I can, but whether I do in a timely fashion or not doesn’t mean that the information ceases to exist. Yesterday’s rip (higher) in the Long Bond (TLT +1.2%) and drop in the SP500 (SPY -0.73%) can be (partly, but importantly) explained by the Consensus Macro’s net positioning as of Friday’s close:
In other words, into the “everything is awesome” jobs cheerleading report:
“So”, no matter what you think consensus is… that’s what it was - and the next Bayesian probabilities to weigh have everything to do with why consensus is positioned that way. Isolating why on both GROWTH and INFLATION, here are a few A/B tests (toss the ball):
A) On GROWTH, they must think US growth, employment, wages etc. are fixing to achieve some sort of “escape velocity”…
B) Or they realize that even if they are wrong on growth… that the Fed, BOJ, and ECB bails them out anyway
While its perverse, that B) scenario is credible. It’s the levered-long heads you win, tails you win bet – throw the ball anywhere on the table (other than the Russell 2000 and Energy stocks) and you win, right? How about we roll the queue ball on INFLATION?
A) As #deflation expectations accelerate, consensus must think the Fed is going to raise rates into that? …
B) Or that consumption growth is going to be so strong that the Fed will dismiss #deflation as a risk and hike anyway
The problem with what we call the #Quad1 scenario: A) US growth accelerating B) on inflation decelerating is that it’s a theory, not yet a reported reality. And that’s the thing about theories – they are for people who are smarter than me that don’t need to keep taking more reps with the uncertainty ball. They just need a survey confirming their pre-determined belief.
What if McDonalds (MCD) reporting a down -4.6% year-over-year same store sales number for November has something to do with the non-Wall Street economy that still has no wage growth? What if the recent Retail Sales, real personal consumption growth, and jobless claims numbers reported by the government are right (they’ve been slowing)?
Well, that will make Friday’s Consensus Macro position in SPY vs TLT wrong… and it might remind some of us that practical questions are unanswerable by any other means than by embracing the uncertainty of each risk management day.
Our immediate-term Global Macro Risk Ranges are now:
UST 10yr Yield 2.16-2.32%
WTI Oil 61.27-68.02
Best of luck out there today,
Keith R. McCullough
Chief Executive Officer
Takeaway: Elevated Attrition + Peaking Penetration = Declining Users in 2015.
This is a summary of the salient points behind our thesis. As usual, we will be publishing follow-up notes with incremental analysis, and hosting a call to run through the detail. In the interim, let us know if you have any questions or would like to discuss in more detail.
We have previously identified that P has historical retention issues, which we detail in the chart below. Since 1Q11, P has added more than 160M registered accounts, yet only grew active users by 45M, suggesting total churn of at least 115M accounts, or 72% of its gross account gains during this period. The bigger issue is that P has already exhausted much of its TAM, particularly the low-hanging fruit.
Back in August, we ran a survey of 20K US internet users to determine P's actual penetration levels. Our survey results suggested that P has penetrated roughly 55% of US adult internet users. That may sound like a ton of runway, but we estimate that roughly 2/3 of P’s remaining adult TAM is over the age of 45, roughly half is over age 55. Further, we also estimate that P has likely penetrated over 70% of the teenage population. In short, new user growth will prove more challenging from here.
We estimate that P needs to sustain a run-rate of gross new quarterly account adds of 11M-13M to maintain its active user base given our estimate of mid to high-teens quarterly churn rate. Even If that run-rate was possible over the long-run, and P could penetrate every internet user in the US, we estimate that P would exhaust its unpenetrated TAM within 7-10 quarters
In a more likely scenario, we expect gross new account adds to slow given the high concentration of older users within P's unpenetrated TAM, which should lead to y/y declines in user and/or listener hour growth sometime in 2015. See the link below supporting detail and charts on our survey results & TAM analysis.
P: User Penetration Survey (N=20,000)
08/28/14 04:12 PM EDT
Outside of listener hours, advertising revenue per user (ARPU) is driven by increasing ad load and ad rates; we're most concerned about the former. P’s ad load has steadily risen throughout its history. However, we suspect that it’s these increases in ad load that are exacerbating its retention issues, particularly its most recent one in late 2013.
Back in August 2013, P increased its max ad load per listener hour from 4 to 6, but the bigger issue was that P altered its ad feed from 1 every 15 minutes to 2 every 20. Collectively that translates to as much as a 33%-100% increase in ad load depending on how long the user’s session lasts. Further, the altered ad feed conditioned the user to expect back-to-back ads, which we suspect may cause some users to end their session prematurely upon hearing the first ad.
That said, it shouldn’t be a surprise that in 4Q13, P saw its sharpest deceleration in net user growth in its reported history (from 25% to 14% y/y growth). We suspect that surge in ad load in 8/13 led to accelerated churn in the following quarter. Moving forward, we have no reason to expect anything different when/if P increases ad load again: rising ad load will push the user away, especially with a growing wave of options for streaming music online.
The one big positive for P is ad rates, specifically its push into the local advertising market, which the company suggests carries rates as high as 2.5x the national average. However, we do not believe P has the geo-targeting ability to command rates that high (P’s geo-targeting based on the zipcode provided by the user during the registration process). Regardless, consensus estimates already imply a sharp acceleration in ad pricing; even if the sell-side doesn’t realize it (we’re likely alone in our declining user view).
In the table below, we’re running a scenario analysis on 2015 revenues; flexing ad-supported listener hours against Ad RPMs (proxy for ARPU). We have included P’s 2014 YTD performance on both fronts for perspective. We caution not to read too much into its YTD listener hour growth since that is inflated by the removal of P's mobile listening cap from 2013 (see chart below). In 2015, we don't believe listener hour growth will exceed double-digits.
That means that consensus estimates for 2015 are unattainable outside of a considerable acceleration in ARPU. As mentioned above, increasing ad load will come at the expense of listener hours and/or user retention. So the question is whether its local ad push can produce accelerating pricing growth vs. 2014, and if so, by how much…all things considered, we suspect it won’t be enough.
We believe the stock could easily trade in the $13-$16 range (~15%-30% downside) based on 0.5x-1.0x turns of P/S multiple compression on our 2015 estimate of $1.1B. Note that P recently closed at $16.90 last Tuesday
But the better question is how much do you pay for a company that starts losing users before it generates positive FCF? Or what happens to the stock the first time P prints declining users? Our price target range may be too optimistic.
Let us know if you have any questions, or would like to discuss in more detail.
Hesham Shaaban, CFA
Enter your email address to receive our newsletter of 5 trending market topics. VIEW SAMPLE
By joining our email marketing list you agree to receive marketing emails from Hedgeye. You may unsubscribe at any time by clicking the unsubscribe link in one of the emails.
Takeaway: This morning's Existing Home Sales print looks disappointing at first take, but less so when you look at the EHS/PHS dynamic since March.
Our Hedgeye Housing Compendium table (below) aspires to present the state of the housing market in a visually-friendly format that takes about 30 seconds to consume.
Today's Focus: November Existing Home Sales
Existing Home Sales dropped -6.1% sequentially from 5.26MM in October to 4.93MM in November, the largest sequential decline since January and the first month below 5MM units SAAR since May.
However, given the similar temporal dynamic last year, from a rate of change perspective, EHS grew +2.1% YoY, down from the +2.3% reported in October but just the second month of positive YoY growth in the last 13 months.
A few things to consider:
Soft but Not Surprising: In the October EHS review we highlighted the emergent divergence between Pending and Existing Home Sales. Specifically, off the March trough, EHS were up +14.6% while PHS was up just 11.5%. Further, PHS for October were down -1.1% sequentially, taking the total gain from trough down to +10.5%.
Given that the two series are invariably tethered, unless we saw sequential strength and/or a significant positive revision to the PHS data, it was unlikely EHS would show similar strength in November. Inclusive of November, the gain since trough for EHS is now +7.4% (vs. +10.5% for PHS). Should PHS for November (released 12/31) come in flat to better sequentially, we'd expect rebound strength in EHS over the next couple months.
GSE Regulatory Changes: As we highlighted alongside last week's Purchase application data, the regulatory change for Fannie Mae which reduces the minimum down payment requirement to 3% from 5% took effect on December 13th. To the extent the change weighed on November or early December demand remains to be seen. We suspect the impact was probably modest - and seasonal distortions in the peri-holiday period will complicate delineating any specific factor in broader demand trends - but we'll get a look at the first week of potential impact in Wednesday's MBA release.
Seasonality: The seasonal factor for EHS in the fourth quarter has been getting progressively smaller (ie. getting closer to 1.0) since 2006 which is to say that organic trends, and not seasonal adjustments, have become an increasingly predominate driver of the reported sales figures.
The October and November seasonal factors (defined here as NSA/SA) have been notable. The October seasonal factor was well above the trailing 5Y average and above 1.0 for the first time since 2008. Had the scalar been in-line with the average, seasonally-adjusted EHS in October would have been +5.43MM (vs. 5.26MM reported).
Conversely, the seasonal factor for November (0.857) was well below the trailing average (0.905), helping support what would have been an even worse sequential decline.
Normalizing the last two months of data yields a two month average of ~5.05MM, which compares to the TTM average of 4.88MM. So ongoing, albeit modest, improvement remains the broader trend.
About Existing Home Sales:
The National Association of Realtors’ Existing Home Sales index measures the number of closed resales of homes, townhomes, condominiums, and co-ops. Existing home sales do not take into account the sale of newly constructed homes. Existing home sales account for 85-95% of all home sales (new home sales account for the remainder). Therefore, increases in existing home sales tend to signify increasing consumer confidence in the market. Additionally, Existing Home Sales is a lagging series, as it measures the closing of homes that were pending home sales between 1 and 2 months earlier.
The NAR’s Existing Home Sales index is published between the 20th and the 22nd of each month. The index covers data from the prior month.
Joshua Steiner, CFA
Christian B. Drake
Editor's note: This is a brief excerpt from Hedgeye morning research. Click here to learn more about how you can become a subscriber to the fastest-growing independent research firm in America.
* * * * * * *
Things that crash…bounce.
Over in the Gong Show that has become Putin's Russia, the RTSI is up +4.2% this morning after being up big on Friday. That said, Russia is still down a monster -42% year-to-date (which means only +72%, from here, to get whoever owned it up there back to breakeven!).
There is massive resistance for the Russian Trading System in the 885-912 range (no support to 623).
No, it's not looking good for Mother Russia.
Daily Trading Ranges is designed to help you understand where you’re buying and selling within the risk range and help you make better sales at the top end of the range and purchases at the low end.