An update on JD.com

In light of recent events, many have asked me for an update on JD.com. I remain massively long (though less than before due to the decline in share price). Guess I’m a bagholder. I would add to the position but it clearly was too big from the get-go (and still is). Hindsight’s 20/20.

In this post, I will first condense the thesis down to the few things that really matter (in my opinion) for the benefit of those who do not have the time to go through the very long post and subsequent posts (here, here, and here). Then I will discuss competitive developments – largely from Alibaba. Finally, I will detail my thinking regarding Richard’s scandal.

If you have approached me via twitter/email, you would already be somewhat familiar to my thinking and thus can skip to the portions you are unfamiliar with.

As always, I could be partially, substantially, or totally wrong, and would appreciate if my (very) smart readers convince me of my folly.

The few things that really matter 

In my view, JD is a compelling long as the consensus is under-appreciating the company’s structural cost advantages and exceptional reinvestment economics while overstating the competitive threat from Alibaba.

Lowest-Cost Operator: Due to its position as the largest retailer in China, JD has a highly efficient cost structure with OpEx/sales of 12% on a normalized basis as compared to peers (Suning, Gome, Sun Art, Lianhua) who operate at 15%+ OpEx/sales. This allows the company to price competitively and reinvest volume rebates from vendors into discounts/rebates/coupons for consumers in a virtuous cycle. The passing on of volume rebates from suppliers to consumers is easily verified by querying Baidu/Sogou/etc and casual usage of the JD app, website, and mini program.

Apart from returning value to consumers, the company has also reinvested in logistics and R&D. These investments are real and verifiable – the former by comparing the company’s gross PP&E with GLP’s logistics portfolio in China and assessing the reasonableness of daily orders fulfilled per fulfillment personnel, the latter by tracking the company’s progress in automated warehouses/vehicles and job postings on LinkedIn.

Continued reinvestment in logistics has resulted in JD having ~20% lower per-order fulfillment costs relative to third party express couriers (e.g. ZTO, YTO, STO, Yunda) despite having less than half the order volume.

While JD is not profitable on a consolidated basis – this is largely due to its reinvestment in discounts, rebates, logistics, R&D, and FMCG – the company is actually profitable on an unit economics basis even in low gross margin electronics, in my view.

Assuming a 5% gross margin on a RMB1,000 smartphone (conservative, given other electronics generally have much higher ASP) results in RMB50 in gross profit dollars. OpEx per order is roughly RMB24 and hence operating profit dollars is RMB26.

(Although some believe JD needs to gain substantial share in high gross margin categories such as apparel, the company actually earns a similar amount of operating profit dollars from apparel as compared to electronics, assuming RMB100 apparel ASP, 50% gross margin, and OpEx per order of RMB24.)

Exceptional Reinvestment Economics: The company has a long runway for growth given low household income and consumption relative to GDP in China (30%-40%) as compared to other developed countries (50%-70%), continued market share gains from C2C and less efficient offline retailers, its current market share of ~3.4% of total China retail sales, and a relatively immature customer base (8 annual orders on average vs 30 orders of the 2008 cohort). Reinvestment in growth is capital-light given zero-cost financing due to long payable days – the company actually has negative invested capital unless one capitalizes operating leases.

Alibaba Fears Overblown: Many are of the opinion Alibaba will be able to crush JD due to its dominant market position, large profits, and technological superiority. However, despite having much lower GMV as compared to BABA, JD has 2x-3x the promotional capacity based on estimated vendor rebates (which BABA does not have much of due to it being almost a pure marketplace player) and reported marketing expense. Further, the benefits of incremental economics of scale to JD is much greater as compared to BABA merchants due to the sub-scale nature of the latter. Moreover, while BABA is highly profitable, these profits have been “encumbered” due to the apparently large (and rapidly increasing) funding requirements of its equity investees. Finally, while BABA is definitely way ahead of JD in terms of personalization, there is no structural barrier restricting JD from improving on personalization as the technology does not appear unique given many other tech giants have been able to implement personalization, though with varying degrees of success.

Compelling Valuation: After one nets off $7 per share in net cash, investments in equity investees, and other investments securities, and $13 per share from the company’s 81.4% stake in JD Logistics and ~36% stake in JD Finance based on both of their most recent financing rounds, the implied value of JD Mall is ~$7 per share as of time of writing.

Applying a normalized 1P and 3P EBIT margins of 4% and 30% on FY17 1P and 3P sales (excluding estimated external JD Logistics sales) results in consolidated EBIT of $2.20 per share. This implies JD Mall is being valued at ~3x EV/2017 normalized EBIT, which is stunningly cheap, in my view. Discounting finance and logistics at 50% (after all, these are VC valuations) would lead to JD Mall being implicitly valued at ~$13.50 per share or ~6x EV/2017 normalized EBIT.

Normalized 1P EBIT margins are in-line with the average of top 250 global retailers and legacy EBIT margins of Suning, Gome, Sun Art, and Lianhua prior to competition from JD. Management has guided to 3%-5% 1P net margins, implying 4%-6.6% EBIT margins. Amazon earned 5%-6.5% EBIT margins from ’03-’05 when it chose to slow its growth. Wal-Mart currently earns 5.6% EBIT margins on its US business while Target is doing 6%-7%. Normalized 3P EBIT margins are in-line with marketplace peers such as Alibaba.

Risks: Due to its large accounts payable, JD is at risk of large capital calls if its 1P sales contract sharply – negative working capital is lovely until sales decline. This is mitigated by current high growth rates of ~30%, a diversified customer base, and net cash, investments in equity investees, and other investment securities roughly covering the entire accounts payable balance. The relatively low growth rates of the company’s offline competitors, roughly break-even profitability, and much larger payables balances places them at greater risk – a recession should therefore severely weaken JD’s competition.

Risks stemming from the VIE structure are mitigated by a history of investor-friendly dealings (Tiger, JD Finance), aligned interests (Richard controls both JD and the VIE), China’s implicit acceptance of the VIE structure per draft CDR regulations, and capital markets deterrence (presumably China would not want to do anything to impair their ability to access capital markets).

Competitive developments

Regarding Alibaba, the major issue is delivery timing – it appears BABA has been matching JD’s delivery timings – JD’s advantage in delivery speed has been touted as one of its key differentiators. However, this seems to be limited to tier 1 cities – based on data from State Post Bureau, third party couriers had average delivery timings of ~50 hours in top 30 cities (out of a sample of 50).

Data from Alibaba’s recent investor day shows delivery timings declining from 9 days to 2.8 days from ’13 to ’17 which is materially longer than the timings reported by SPB – I suspect SF Express is dragging the average timings lower. In addition, the aforementioned data from BABA is only for the first 100m orders, suggesting delivery timings are deflated due to pre-Singles Day preparation and the absence of data beyond first 100m orders (translation: it probably got way longer).

Over the long-term however, I expect BABA to be able to match JD on delivery timings in the entirety of China – there are just limits to how fast you can deliver something to a customer. However, I am doubtful Alibaba would be able to match JD’s fulfillment cost per order due to Cainiao partners outsourcing first-/last-mile to franchisees which are mostly mom-and-pop. In order for BABA to match JD’s fulfillment cost per order, it would need to consolidate the entire delivery process into a single entity to maximize economies of scale, in my view. This is occurring (see ZTO buying out network franchisees) but should take decades if we use the consolidation of US trucking as a guide.

Of course, consumers only care about delivery speed – they hardly spend time wondering about the cost of the infrastructure required in order to deliver their orders. While BABA matching JD delivery speeds is a negative, it seems like a slight negative to me – it does not seem reasonable to assume JD competitive position would be materially impaired just because BABA has similar delivery speeds; modeling a slowdown in growth is probably warranted, however.

It is also interesting to note the apparent rise in Pinduoduo cannot be seen in the sales figures of third-party express couriers, suggesting either Pinduoduo is materially inflating GMV numbers or Taobao is losing significant share to Pinduoduo. Pinduoduo’s rise at Taobao’s expense also hurts Tmall because Taobao funnels massive traffic to Tmall (which explains why Tmall has relatively low MAUs).

Richard’s scandal 

In my view, it is highly likely Richard will be found innocent. Studies on false rape accusations in the US, New Zealand, and UK show a wide range of numbers on the rate of false accusations – anywhere from 2% to 90% based on data compiled by a 2006 study by Rumney. Using the FBI rate of 5% suggests the base rate of false rape accusations is pretty low. But relative to false allegations of other crimes (murder, burglary, theft, larceny, etc), false rape accusations are a stark outlier and roughly in-line with false robbery accusations. False rape accusations are 5 times more common as compared to other crimes.

While the rate of false rape accusations is low on an absolute basis, there are problems with all studies on this subject given the enormous amount of subjectivity involved – how do you rigorously define a false accusation?

In my view, looking at non-conviction rates is more objective – BOJ data shows conviction rates which are tiny – not because of the lower risk of conviction, but because non-conviction implies a rigorous investigative process has already occurred and thus if the accused was not convicted, it strongly suggests a lack of sufficient evidence (and thus substantially increases likelihood of the accusation being false in the first place). That being said, non-conviction rates are far from the perfect metric, and I am certainly no lawyer.

The key factor to note, in my view, is that the police allowed Richard to return to China despite him being an obvious flight risk. I believe this strongly suggests Richard is innocent, especially given the recent popularity of the #metoo movement.

Leaving aside the question whether Richard will be found innocent or guilty, I believe the possibility of a sustained, large-scale consumer backlash is the more pressing issue – in fact, the most important issue. There could be a consumer backlash irrespective of whether Richard is found innocent or guilty depending on consumer reactions.

I have been tracking traffic trends over the past few weeks and it does not appear there is any sustained, large-scale consumer backlash. Note that higher/lower traffic does not necessarily equate to better/worse as traffic is rather seasonal day to day. Instead, sustained highs/lows are good/bad.

The worst-case scenario seems to be if Richard is forced to step down as chairman/CEO. In my view, there would be limited impact to the business in such a scenario. I do not think Richard is vital to running operations day-to-day – he appears more concerned with long-term strategy. Sidney, the CFO, seems to know the retail business very well whereas the logistics and finance subsidiaries have separate CEOs.

The consensus seems to believe the company does not have a deep management bench. I think this is a misperception caused largely by Richard’s prominence in English language media; if you only read Chinese media, you’ll likely come away with the perception that the company has a deep management bench as other senior executives are regularly featured.

JD is also much less risky compared to the 2007-2014 period where Richard’s strategic direction was needed. Going forward, it seems to me JD is all about execution – which they don’t really need Richard for. Given his large ownership stake, he’ll likely be kept as a consultant either formally or informally in the event he steps down.

Regarding the corporate governance risk which stems from Richard’s supervoting shares and the requirement he be present in order for the quorum to be satisfied, I believe this risk is overstated. In my view, if it comes down to it, Richard will likely step down quietly given his self-interest (he owns shares) and his history of treating stakeholders (employees, suppliers, shareholders) fairly.

Advertisements

Long: SBS Transit Ltd (SGX:S61)

I am long shares of SBS Transit Ltd in significant size. I have had the position on for a while but have only now decided to do a blog post. I am at an loss attempting to falsify the thesis and am hoping others could potentially disconfirm it (and thus save me money if I’m wrong!).

There are four parts to the business – the operation of bus contracts, the operation of rail contracts, advertising/rental stemming from bus/train stations, and lease income stemming from the leasing of buses to the LTA.

In my view, shares are extremely mispriced as the market ignores the company’s transition from an asset-heavy model to an asset-light one as a result of government action and significantly under-appreciates the growth potential stemming from the relatively new Downtown Line.

Furthermore, the business has been significantly de-risked as fare revenue, fuel costs, and wage inflation are passed-through to the LTA under the new bus/rail contracting model. In addition, capital requirements going forward are likely to be minimal as a result of the new contracting model.

While the new contracting model does increase competition in the industry, this is insulated by relatively long-term contracts for both bus/rail operations.

In my opinion, the largest downside risk to the business is the loss of bus contracts. The company recently lost the Loyang/Bulim bus contracts but won the Seletar/Bukit Merah contracts.

If we make the draconian assumption that the company loses all of its bus contracts, SBS would lose roughly $30m in EBIT (based on implied per-bus ad revenues losses from Loyang/Bulim, or $10.3k/bus; total = ~2,900 buses, and conservatively assuming 100% margin) compared to $162m in FCF, based on my estimates.

The majority of this hypothetical lost EBIT would stem from the advertising/rental portion of the bus/rail operations, given bus operations are likely operating at a loss, while rail operations are probably slightly profitable. The company’s highest margin business is its advertising/rental business.

There has been a rather large increase in accounts receivables and other receivables and prepayments, but this appears to be largely explained by the new contracting model – consumers typically pay cash or credit whereas the LTA has latitude to take relatively longer to pay up.

According to my estimates, lease income from leasing buses to the LTA is ~$67m annually. Based on net book value of buses as of FY17, this lease income will disappear in 10 years. Discounting the total at the 10-year Singapore government bond rate of roughly 2.50% (appropriate given LTA is essentially the government) implies $590m in total lease income, on a present-value basis, or $1.90 per share. There are no associated incremental costs with lease income given its essentially just signing a contract with the LTA.

Current share price is $2.60. 311m shares outstanding. Debt is $0.58 per share. Cash is minimal. EV is thus $3.18 per share. Adjusting for the remaining lease income payments, EV would be $1.28.

Due to the new contracting model, IR has advised there will be no major CapEx or incremental working capital investment for the foreseeable future. FY17 OpCF before changes in working capital and interest expense is $162m or $0.52 per share.

Considering IR guidance for working capital and CapEx, FCF is probably around the same figure. Subtract the lease income (assuming 100% profit margin), this figure drops to $0.31 per share. Shares thus trade at ~4x EV/FCF.

Even if there is some unexpected CapEx, and if we assume FY16/17 as a guide and ding ~$30m or $0.10 per share. FCF would be ~$65m or $0.21 per share, shares thus trade at ~6x EV/FCF.

The company’s earnings should increase drastically over the next few years as the new Downtown Line 3 ramps up given high operating leverage and incremental advertising/rental opportunities. So ~6x EV/FCF suggests tremendous margin of safety, particularly given the current earnings are likely materially depressed given the DTL3 ramp.

Insuring Against Multiple Crises: Chinese Banking and Australian/Canadian Housing

I am short, through options, the AUD and CAD relative to the USD in anticipation of a severe depreciation of the AUD and CAD driven by a desertion of wholesale funding for the Australian and Canadian banks and/or a major currency adjustment in response to a RMB devaluation relative to the USD.

In my view, there are two main ways a substantial depreciation of the AUD and CAD relative to the USD can occur – 1) the bursting of their respective domestic housing bubbles, and/or 2) a China-related crisis.

The bursting of their domestic housing bubbles would result in vastly higher net charge-offs for Australian and Canadian banks which should elevate solvency concerns and thus cause wholesale funding to vanish. While I expect Australian banks to experience a liquidity crisis, I do not expect the same for Canadian banks due to the latter’s relatively lower usage of wholesale funding.

Potential China-related crises include a hard landing and/or recapitalization of its banking system and/or major RMB devaluation leading to a severe decline in demand for commodities and drying up capital flows which would likely result in the adjustment of the AUD and CAD relative to the USD.

There are fairly substantial overlaps between the two outcomes. One example is a major RMB devaluation leading to the drying up or reduction of capital flows to Australia and Canada should effectively lead to a run on wholesale funding for the Australian and Canadian banks, which itself should burst their respective domestic housing bubbles. That being said, a severe depreciation in the AUD and CAD can occur independent of a China-related crisis.

I do not believe the scenarios listed above are likely to be the median outcomes, but I do believe these potential outcomes are far from remote. However, put options on the AUDUSD and CADUSD have extremely low implied volatility suggests the market views these potential outcomes as remote. Therein lies the mispricing.

My position is tiny, which raises the question – why even bother? However, the risk/reward is outrageously asymmetric, which makes insuring against these outcomes very attractive, in my view.

China

All figures for this section are in USD.

In response to the 2008 financial crisis, China embarked on a $4 trillion stimulus program in order to maintain local employment and thus social stability. The government chose to finance this stimulus through the banking system, effectively expanding credit by ungodly amounts.

For a variety of reasons largely irrelevant to the thesis, China has continued said credit expansion at very fast rates to this day. Continued credit expansion eventually leads to diminishing returns, and returns have diminished dramatically over the past decade.

It is quite impossible to accurately measure the “return on investment” on this credit expansion as there is always noise with any metric. One example is to look at absolute credit (defined as total social financing because bank loans understate actual credit outstanding as it excludes large credit sources such as entrusted/trust loans, financing through corporate bond issuance, bankers’ acceptances, etc) growth relative to absolute GDP growth.

On this metric, the absolute credit:GDP ratio (or credit productivity) was 1:1 from 2003-2008, 2:1 from 2009-2010, and has rapidly increased in recent years to 4:1 in 2015, 6:1 in 2016. The figure is closer to 2.2:1 in 2017 as TSF growth slowed considerably.

However, these are likely conservative estimates of “credit productivity” for three main reasons – 1) TSF is understated as it excludes local government debt swaps, peer-to-peer lending, asset-backed securities, etc, 2) GDP is likely overstated by an indeterminate amount (e.g. Liaoning) and 3) the metric GDP itself is an imperfect measurement of productivity as it is indifferent to projects with varying ROIs; projects with negative ROI would still be considered productive using this metric despite the fact they are economically non-productive.

Because much of the country’s GDP growth has been driven by fixed asset investment, the incremental capital output ratio (“ICOR”) is probably a more accurate (but still imperfect) metric to consider. ICOR hovered between 2-4 in the 2000s but has skyrocketed to >15 in the past decade, declining to ~12 in 2017 as fixed asset investment slowed considerably. Most countries have ICORs in the low-single digits.

Likewise, ICOR is also imperfect as it is plagued by similar issues as credit productivity – particularly issues related to GDP. Regardless, it seems reasonable to conclude the ROI on credit expansion has been atrocious over the past decade, based on the above metrics. Atrocious ROI on massive credit expansion has fueled potentially the largest amount of unrecognized non-performing loans globally and historically.

The market appears to severely underestimate the actual level of non-performing loans in the Chinese banking system. According to Natixis, the percentage of non-performing loans for listed banks has increased to ~1.30% from around 1% from 2014 to 2017 while the percentage of loan write-offs has grown from ~40bps to ~60bps over the same period – which vastly understates actual NPLs, in my opinion.

Even if we include special mention loans, the percentage of “adjusted NPLs” would increase to ~5.3% in 2017, implying ~$966b in adjusted NPLs (Natixis estimates 48% of the ~$38tr in on-balance sheet Chinese banking system assets are loans) as compared to ~$237b in non-adjusted NPLs.

The adjusted NPL figure is significantly below my expectations due to two main reasons – 1) there are substantial loan-like assets on the balance sheet of the Chinese banking system which are not captured by the loan figures, and 2) it does not take in account the off-balance sheet (i.e. shadow banking) assets of the system.

Natixis estimates that ~9% of on-balance sheet banking system assets, or ~$3.4tr are “investment receivables” as of 2017. A large part of these investment receivables comprise of structured loans and trust beneficiary rights (“TBRs”). TBRs are essentially non-consolidated loans sold to trust companies, sometimes financed through the issuance of wealth management products, where the bank retains credit risk but is required to set aside significantly lower amounts of capital vs a normal loan – hence getting around loan/deposit limits.

There are many estimates for the amount of shadow banking assets in China ranging from $9tr (PBoC) to $37tr (Daily Telegraph). An apparent leak of the PBoC annual financial stability report in 2017 ballparks the figure at roughly $30tr to $40tr. Thus the lower bound would be $9tr and the upper bound would be $40tr.

In sum, total banking system assets (on- and off-balance sheet) are roughly $50.4tr to $81.4tr.

A 2006 BIS paper assessing China’s banking system reform estimates NPL percentages of 20%-30%+ over the 1997-2003 period following the Asian Financial Crisis and estimates 20.5% cash recovery rates on NPLs disposed off by China’s state-backed AMCs.

Using these figures, actual NPLs in the Chinese banking system could amount to roughly ~$10.1tr to ~$24.4tr with cash recoveries of ~$2.07tr to ~$5tr, implying ~$8.03tr to ~$19.4tr in NPLs net of recoveries.

Natixis forecasts a 0.91% ROA for the on-balance sheet Chinese banking system in 2017, which represents a decline from 0.98% in 2016, and 1.10% from 2015. Natixis also estimates non-adjusted NPL provision coverage of 180%, implying ~$889b in loan loss provisions.

Conservatively assuming similar ROA on total (both on- and off-balance sheet) banking system assets suggests ~$459b to ~$741b in net income, or ~$581b to ~$938b in pre-tax income, using Natixis’ 21% tax rate estimate for Chinese banks.

As a result, total pre-tax, pre-provision income for the Chinese banking system is roughly ~$1.47tr to ~$1.83tr. Natixis estimates 13.2% assets/equity on on-balance sheet banking assets for FY17 implying ~7.57% equity/assets or roughly ~$2.88tr of Chinese banking system equity.

Assuming the Chinese banking system requires a 8% equity/assets to be considered adequately capitalized and that total banking system assets are treated as risk-weighted assets, the system would need to recapitalize to ~$4.03tr to ~$6.5tr of equity after adjusting for NPLs net of recoveries.

With a starting point of ~$2.88tr of equity, the equity “shortfall” would be roughly ~$9.18tr to ~$23.02tr after NPLs net of recoveries, suggesting at current pre-tax, pre-provision profitability, the Chinese banking system would take ~6.2 years to ~12.6 years to recapitalize.

Suffice to say, these are extremely rough numbers given these metrics are not perfect and do not represent the only way for the Chinese banking system to recapitalize.

For example, while system pre-tax, pre-provision profitability might decline as underwriting standards tighten, it could also improve as competition (especially from the shadow banking system) is wiped out.

In addition, the system could rebuild equity through other means – following the AFC of ’97, the system was recapitalized through a combination of 1) government liquidity injections, 2) NPLs transfers to state-backed AMCs, 3) PBoC injections through FX reserves, 4) equity injections from Singapore’s state-controlled investment fund, Temasek Holdings, and of course, 5) pre-tax, pre-provision profitability. With relatively more open financial markets as compared to the 1997-2003 period, other potential sources include raising equity from the capital markets and debt/equity swaps.

While the market appears to be relying on the PBoC FX reserves as the ultimate backstop for the Chinese banking system, the consensus fails to consider whether the PBoC warchest is “unencumbered” or even sufficient.

According to the IMF, reserves/short-term debt exceeding 100% is adequate, reserves/broad money (typically M2) of 5% is typical whereas 20% is the upper range of prudent. China has roughly $3.1tr of FX reserves as of March 2018, short-term debt of roughly ~$1tr as of Dec 2017, and M2 of ~$27.6tr as of Dec 2017.

If we are conservative and utilize the more flattering metric (in this case, short-term debt) to calculate the reserve “headroom”, China would have “unencumbered” FX reserves of ~$2.1tr.

If we assume a 3-year recapitalization, the Chinese banking system would have generated ~$4.4tr to ~$5.5tr of pre-tax, pre-provision profits, which would reduce the equity “shortfall” figure to ~$4.78tr to ~$17.5tr, with the upper range of the figure being more likely; note the lower bound is underpinned by the assumption that the PBoC’s official report of shadow banking assets being only $9tr is accurate, as opposed to its leaked report of $30tr to $40tr, and other estimates in the mid-$30tr range.

This range would hence represent the amount China has to raise from sources other than pre-tax, pre-provision income and state-backed AMCs. This range also far exceeds China’s “unencumbered” FX reserves, or even its FX reserves, for that matter.

Similarly to the AFC ’97, I expect China to recapitalize its banking system with a combination of money printing, government liquidity injections, and PBoC FX injections. A significant devaluation of the CNY would increase the PBoC’s FX reserves in CNY terms. Regardless of the combination, a severe CNY devaluation appears to be the common outcome.

Australia has substantial exposure to the Chinese economy as China accounts for 1/3 of Australia’s share of trade exports (mainly commodities) and roughly 18% of its services exports. Australia is thus tied to the RMB.

Canada has lesser exposure to the Chinese economy in terms of share of exports, but both Canada and Australia have benefited over the years from Chinese capital flows which have manifested themselves largely in their respective housing bubbles. The reversal or drying up of these capital flows would effectively lead to a run on the Australian and Canadian dollar.

Australia

All figures in this section are in AUD.

Australia is an outlier economy in one prominent aspect – the country hasn’t had a recession (as technically defined) in almost 3 decades. Australia is also home to one of the most profitable banking oligopolies in the world. However, Australia is also home to one of the largest housing bubbles in history.

Australian house prices to income has exceeded 6.0x in 2017, as compared to the 4.8x the US peaked at in 2005. According to the RBA, total household debt to income has increased ~30% over the past 5 years to ~190%, after being roughly flat for a decade. As a reference, during the peak of the global financial crisis, US household debt to income was roughly ~130%. While the RBA observes total household debt mortgage repayments have been steady as a share of income for years, this neglects to consider the fact interest rates have declined substantially over the past few years.

Although the RBA has encouraged households to switch over from interest-only loans to amortising loans in an attempt to reduce risk, this will massively increase monthly payments for mortgage holders – the RBA estimates roughly ~30% of all national mortgage debt outstanding will be subject to the reset over the next four years, increasing monthly repayments by 30% to 40% for 1.5m borrowers; ~$120b billion of loans will convert to amortising loans annually from 2018-2021. There has also been a decline of interest-only mortgages from 60% of mortgage issuance four years ago to 30%.

Despite the share of new loan approvals with loan-value ratio >90% declining significantly since 2009, the share of 80 < LVR <= 90 loan approvals has offset this decline, suggesting a material portion of new loans are still being issued at high LVRs.

While the RBA expects this wave of mortgage resets to be manageable, this appears unlikely given the massive growth in household debt to income ratios as noted above. The RBA seems to think that borrowers who cannot meet the increased repayments would be able to refinance their mortgage, a view which is at odds with a scenario where housing prices decline – which is expected to occur due to these mortgage resets. It is somewhat ironic that the RBA’s measures to deflate the housing bubble and avoid a crisis could in fact be the catalyst for the crisis.

Furthermore, there are reports of substantial and growing amounts of liar loans being issued in recent years, which suggests the outstanding mortgage book could be lower quality than perceived.

Because of the extreme concentration of Australian banking, I will focus only on the big four banks (ANZ, CBA, NAB, and WBC), which account for over 80% of Australian banking.

According to KPMG, the big four Australian banks had equity of ~$235.4b as of FY17. Pre-tax profit was ~$43.7b. Provisions to credit risk-weighted assets were 0.79%, 0.73%, 0.86%, and 0.76% while credit RWA were $336.8b, $377.3b, $325.9b, and $349.3b for ANZ, CBA, NAB, and WBC respectively. Thus big four provisions were ~$10.86b and hence total pre-tax, pre-provision profit was ~$54.6b.

Loans make up ~69% of the Australian banking system assets. ~42% of system assets are housing loans and thus housing loans make up ~60% of the system’s loan book. This is similar to the U.S. commercial banks’ ~63% exposure to mortgage-related assets during 2008.

Total net loan charge-offs to total loans held by US banks doubled to 1.32% in 2008 on an annual basis, peaking at 2.70% in 2010 and before declining sharply, suggesting a 3-year recapitalization period. On a cumulative basis, this suggests US banks charged off 6.56% of their 2008 loan book. Assuming ~69% of big four assets are loans, total loans would be ~$2.42tr. 6.56% charge-offs would imply ~$158.7b in charge-offs. With pre-tax, pre-provision profit of ~$54.6b, the Australian banking system would recapitalize in just under 3 years.

While this is a relatively rapid recapitalization period for the Australian banking system relative to my estimates for the Chinese banking system, it is comparable to the roughly 3-year period for the US, suggesting the situation is still very severe in absolute terms.

According to KPMG, the big four Australian banks have deposit to loan ratios ranging from 71%-80% as of FY17, suggesting a significant portion of their funding is wholesale in nature. The RBA estimates that the Australian banking system sources roughly ~20% of its funding from short-term wholesale debt which typically have maturities of 6 months or less.

While this figure is materially lower than the 30%+ during the financial crisis, it remains substantial and presents significant refinancing risks. Funding through securitization has tailed off to a negligible amount.

In my view, the market is vastly underestimating the potential net charge-offs in the Australian banking system and when this is eventually recognized, there will be a run on wholesale funding. Illiquidity concerns due to the run on wholesale funding could also result in a self-reinforcing vicious cycle.

Unlike other countries, Australian banks only offer fixed rate loans up to 5 years, while the majority of loans are variable rate. This also explains why despite dramatic growth in household debt relative to income, mortgage repayments have remained steady. As a result, raising interest rates in order to defend the AUD would likely backfire, as it would exacerbate the mortgage crisis, similar to the situation in Britain prior to exiting the ECU.

In addition, Australia has ~$1.6tr in external debt amounting to ~116% of nominal GDP in 2017 compared to ~$76.5b in FX reserves, which suggests it has limited firepower to defend its currency. Even if the central bank stepped in to intervene, it would be self-defeating as it would worsen the country’s external debt situation, which would result in pressure on the AUD anyway.

Canada

All figures in this section are CAD.

Canadian house prices to income is nearing 7x, which also far exceeds the 4.8x peak in the US prior to the bursting of the bubble. According to Statistics Canada, Canadian households are levered ~170% to disposable income, also in excess of the ~130% of US household debt to income during the global financial crisis.

A BoC report found that by 2Q17, the originations of high-ratio (i.e. with LTV > 80%) mortgages had fallen 17% vs the previous year, while the proportion of households with high loan-to-incomes (i.e. LTI >450%) among new borrowers fell from 19% to 7%. However, households with high LTI accounted for 22% of low-ratio (LTV <80%) mortgages in 2016, up from 16% in 2014. These households also tend to have a larger than average mortgage, making up 32% of the value of all low-ratio mortgages in 2016, up from 9% in 2014. Thus, while the quality of new mortgages appears to be improving on the margin, the fact remains that the outstanding mortgage book remains of dubious quality.

The BoC recently raised its benchmark interest rate by 1.25% and suggested further rate hikes over 2018 were likely. It also estimated that 47% of mortgages outstanding will reset within 1 year, and 31% would reset in 1 to 3 years. In addition, the OSFI recently unveiled final changes to the stress test requirements on uninsured mortgages, which would require buyers to prove they can afford payments based on the greater of the BoC’s 5-year benchmark rate or their contract mortgage rate + 200bps. Substantial refinancing risks could cause the housing bubble to unravel.

Similarly to Australia, I will focus only on the big six Canadian banks due to the concentrated nature of the banking industry in Canada. According to PwC, the big six Canadian banks system had ~$324b in equity as of FY17. Pre-tax profits were ~$54.2b. Provisions were ~$7.5b. Hence pre-tax, pre-provision profits were ~$61.7b. Their exposure to housing loans ranges from 30% to 60% of their loan books.

Using the 6.56% cumulative charge-offs on its 2008 loan book the US commercial banks experienced during the crisis on the big six’s roughly $2.5tr loan book would imply ~$164b in net charge-offs. As a result, the Canadian banking system would take under 3 years to recapitalize.

Loan to deposit ratios at the big six Canadian banks are at ~80% as of FY17. As a result, the Canadian banking system should not be vulnerable to a liquidity crisis stemming from a run on wholesale funding.

However, wholesale funding remains substantial at roughly ~$1tr relative to the ~$4.88tr in total assets and ~$4.55tr in total liabilities. Although the big six Canadian banks have materially increased their portion of wholesale funding with maturities exceeding one year, the majority of their wholesale funding is remains short-dated, according to a BoC review.

Similarly to Australia, the market is vastly underestimating the potential net charge-offs. Unlike Australia, when this is eventually recognized, I do not expect the big six Canadian banks to experience turn illiquid as a result of a run on wholesale funding as its loans are more than fully covered by deposits; wholesale funding could disappear and the system would still remain liquid. However, the run on wholesale funding would result in severe downward pressure on the CAD.

As with Australia, most mortgages in Canada are fixed rate for a 5-year tenure. As noted above, significant portions of mortgages are due to reset over the next year and 1 to 3 years, which makes raising interest rates to defend the currency counterproductive.

Further, Canada has external debt amounting to ~114.6% of nominal GDP in 2017 compared to $106.6b of FX reserves, suggesting limited firepower to defend the CAD as utilization of FX reserves to shore up the CAD would further worsen the external debt situation and result in further downward pressure on the CAD.

United States

The US is a country with massive unfunded liabilities. The net present value of its Social Security and Medicare liabilities as of September 2017 is $49tr, according to the US government’s 2017 financial report. However, this figure is based on a 75-year projection period. If one uses infinite horizon accounting, unfunded liabilities swell to ~$210tr.

It is hard to see the US ever fully funding these liabilities without the government resorting to money-printing or the country massively increasing its productivity presumably through some sort of step-change in technology and innovation.

Nevertheless, these unfunded liabilities are long-tailed in nature and hence is unlikely to result in massive money-printing over a short span of time resulting in a sharp devaluation of the USD; the money-printing would likely take place over decades.

In addition, US growth appears satisfactory to the Fed in the near-term, which should enable further rate hikes and progress in unwinding its balance sheet. These actions should result in tightening liquidity conditions for the USD, placing upward pressure on the currency.

Conclusion

In my view, shorting the AUD and CAD relative to USD is an exceedingly cheap insurance policy against outcomes the consensus views as remote whereas analysis suggests these outcomes are far likelier than remote, though they are unlikely to be the median.

While the potential outcomes are disastrous, I do not believe them to be ultimately cataclysmic. If the events discussed materialize, asset prices globally would likely be exceptionally cheap thus presenting the best investment opportunities in a long time if one has the resources to capitalize on such a situation.

Exited: Platform Specialty Products (NYSE:PAH) – Falsifying the thesis

I recently fully exited my long position in Platform Specialty Products (NYSE:PAH) (“Platform” or “the Company”). This post details how I got interested in the company, my original thesis and variant view, as well as where I (think) my variant view has probably been falsified.

My PAH articles on Seeking Alpha covered a significant (though not comprehensive) portion of my research. I will not rehash them entirely, but stick to the main points, as well as include the rest of my research.

As some of you may know, PAH was conceived by Martin Franklin of Jarden fame. The idea was that Franklin could potentially replicate his Jarden’s success in specialty chemicals by consolidating the industry. This was also Pershing Square’s stated thesis.

I largely disagreed with this thesis because I believed (and still believe) it was overwhelmingly reliant on the continued health of the capital markets; if I had agreed with this thesis, I would have went long in the $20s. Ackman also recently mentioned he made a mistake in believing the “roll-up” thesis was anything but ephemeral. Instead, I researched the company and waited for a lower price. Shares soon fell to the $10 range.

I was of the opinion that shares fell for the following reasons:

  1. the company was over-levered, largely due to two guidance cuts increasing net debt/adjusted EBITDA levels,
  2. that the ‘outsider’ roll-up thesis had been busted, hence resulting in indiscriminate selling by event-driven hedge funds and other ‘outsider’-focused investors,
  3. weakness in end-markets, particularly ag, driven by high channel inventories and low commodity prices,
  4. that its Series B preferred – which had characteristics which made the name ‘death spiral preferred’ more apt – would result in a negatively reflexive outcome (under the share purchase agreement, the Arysta seller had capped their downside at ~$27/share as PAH agreed to pony up a variable amount of cash to a maximum of $600m to make the seller whole as shares fell below that level),
  5. concern over significant non-GAAP adjustments, spring-loading, and other accounting issues

As a result, I thought the above reasons had created a compelling opportunity. The main points of my long thesis were as follows:

  1. While leverage was very high at first sight (exceeding 7x net debt/adj EBITDA, slightly higher if we include unfunded pensions/capitalized operating leases), the amount of leverage was not inappropriate considering the stability of the underlying businesses. More importantly, there was significant headroom (~40% based on the net first lien debt/LTM adj EBITDA maintenance covenant) provided by the company’s lenders, and the company did not require incremental debt financing, which would have likely triggered the incurrence covenant.
  2. The guidance cuts were not due to deterioration in the underlying businesses, but instead due to 1) strength of the USD vs the BRL, and 2) a discretionary change in distribution strategy for the North American agrochemical business unit. Thus, there was little reason to expect further adj EBITDA deterioration.
  3. Consensus estimates (which mainly tracked management guidance) were far too low as the sell-side appeared unaware of incremental profitability stemming from the anniversary-ing of the distribution strategy change, potentially significant tax savings to be realized through restructuring of legal entities, and a normalization of corporate costs. If realized, these factors would significantly boost free cash flow and allow the company to deleverage at a rapid pace, potentially resulting in over 100% upside.
  4. Non-GAAP adjustments were reasonable, restructuring expenses were not ordinary expenses being placed in the one-off bucket, and there was limited, if any, spring-loading. Further, acquisition accounting and depreciation & amortization policies were also not out of line. There were also no one-time gains included in adj EBITDA.

Note my original thesis was largely researched over the 2016 to early 2017 period, and hence the stale numbers. In the section where I falsify my original thesis, the numbers are updated.

Leverage Levels Appropriate Considering Business Quality

Management commentary in 2Q’15 warned the continued rapid devaluation of the Brazilian Real might cause the firm to cut full-year guidance. The continued depreciation of the Real resulted in two guidance cuts; one in August (from $670m to $635m @ the midpoint, or ~5%), and another in October ($635m to $560m @ the midpoint, or ~12%). Aforementioned numbers are pre-Alent/OMG acquisitions.

Pro forma for all acquisitions, this led to net debt/adj EBITDA skyrocketing to ~7.3x incl. the Series B preferred. This also highlighted the mismatch between the Company’s sales exposure (~80% of sales are outside US) and its USD-denominated debt (which was ~70% of total debt at year-end 2015). Additionally, this decline in EBITDA also brought PAH dangerously close to breaching its incurrence covenants (~10% headroom to <4.5x gross first lien and 15%-20% headroom to <6.75x-7.00x total net leverage).

Guidance Cuts Not Driven By Deterioration In Underlying Businesses

Of the ~17% adj EBITDA guidance cut, ~5% (or $40m) was due to a change in the pre-season distribution strategy for NA ag which should return at enhanced margins as the cycle turns, while the remaining stemmed from Real depreciation. I was of the opinion that FX risks, particularly the Real, had been significantly overstated by the market as a result of a misunderstanding of how the working capital cycle works in ag.

Seasonally, the growing/harvesting seasons in Brazil dictates that working capital would accumulate for much of H1 ’15 (the first five months or so) and begin to release in the later months of the year. This is the reason why DSO metrics in the ag space tend to be ~120-150 days. The depreciation in the Real was largely concentrated in H2 ’15 – the USDBRL exchange rate jumped from 3 to 4 during those few months.

Therefore, the flaw in the bear argument with regards to pricing is as follows – it was not that PAH could not increase prices in a major way to offset currency depreciation. Instead, it was the mechanics of the ag working capital cycle which prevented them from doing so.

As of 2Q-3Q’16, much of the Real-related reduction in EBITDA has been recovered as the Company’s implemented “catch-up” pricing initiatives for much of 1H’16. However, this improvement has been disguised due to increased integration-related corporate costs.

In my view, a repeat of the Real depreciation was highly unlikely given the confluence of events required, 1) 30%+ depreciation, 2) occurring largely in the second-half of the year, and 3) occurs over a period of 2-3 months making PAH unable to adjust pricing quick enough. Apart from the Real, the Company’s sales exposures are mainly in hard currencies, which are relatively less volatile than the Real and hence should present even less of a risk.

Mismatch In Sales Exposure & Currency Of Debt Denomination Correcting

In addition, the Company has also begun shifting its USD-denominated debt into EUR-denominated debt, reducing the risks associated with the mismatch between sales exposures and currency denomination of debt. On 7 December ’16, PAH repriced its debt, essentially shifted $425m in USD debt to EUR debt and also resulted in ~$15m in interest savings. The fact banks, in addition to the above, did not rewrite covenants to tighten up covenant leverage ratios also suggests significant lender support, which has persisted as the Company continues to reprice debt at lower interest rates.

Stable Underlying Businesses

The Company’s products are a small portion of its customers’ costs, but are essential to the performance/quality of their end-products. PAH’s businesses have the #1 or #2 market share position in each of their markets, suggesting dominant competitive positions. While anyone can open a chemical formulation plant, genuine replication would be costly. Entrants are guaranteed years of losses with low likelihood of ever achieving profitability.

In AgroSolutions (“Ag”), significant R&D budgets, technical & regulatory expertise, as well as a track record is required to convince large discovery-focused ag companies (Monsanto, Syngenta, etc) to license AIs. Long product registration approval processes, required maintenance of said registrations, sufficient scale to leverage retail distribution, end-user needs for support from technical personnel, the difficulty of convincing risk-averse farmers to switch products, and the need for continuous incremental innovation due to evolution, serve as further barriers to entry.

In Performance Solutions (“PFSN”), significant R&D budgets, long-term collaborative relationships with risk-averse OEMs, and end-user needs for on-site technical personnel make for a highly defendable business model.

~6% of farmer costs are spent on specialty chemicals, whereas $0.90 is spent per phone, and $30-$50 per car, illustrating the low costs of specialty chemicals relative to total costs of the end-product.

As an example, within PFSN, PAH provides specialty chemicals to subcontractors of auto OEMs to prevent car corrosion, something which is highly prized given corrosion can result in recalls; warranty costs are the death knell for the auto OEMs.

The Company’s products are developed in collaboration with OEMs; they often award specification to PAH, which mandates usage of said products by the OEM supply chain. Minimum usage of the Company’s products is required, as customer production lines shut down without said products.

Within Ag, PAH provides a variety of pesticides for usage throughout the entire crop life-cycle. It is also unique due to its ownership of a market-leading bio-solutions portfolio. Products are often combined and sold as a complete solution. Customer lock-in is unsurprising given high benefits/costs.

Ag focuses on specialty crops (~30% of their crop exposure, likely generating disproportionately high margins) such as fruits & vegetables, which offer farmers much higher gross margins as compared to row crops (corn, soybean, etc); wheat gross margins range between 10%-30% while bananas enjoy 70%-90%, depending on price assumptions. While the Company does have ~32% exposure to row crops, this exposure is located primarily in LatAm where strong growth tends to insulate cyclicality. The remaining exposure are to diversified crops, which are individually cyclical but collectively stable. Main competitor FMC is run by shareholder-oriented management, suggesting rational competition in Ag.

The above dynamics equates to strong predictability in cash flows and exceptionally high resilience to down-cycles in the segments’ respective end-markets.

Significant Headroom to Maintenance Covenants

Bears cite the Company’s high debt load (~7.3x net debt/adj EBITDA incl. the Series B preferred) as unsustainable and suggest PAH could violate its covenants, hence triggering a vicious cycle where lenders will likely force stricter covenants, mandate larger compulsory principal payments, or a large equity offering.

Despite the high leverage in absolute terms, the situation is not as dire as bears seem to believe. The Series B preferred amount is not included in the covenant EBITDA calculation and has since been settled through a small equity offering which caused ~12% net dilution (263 o/s – 22.1 + 5.5 + 48.7 over 263).

In addition, the Company’s covenants are split into two types – maintenance and incurrence covenants. Maintenance covenants subjects Platform to a quarterly test as to whether it has a net first lien leverage ratio of less than 6.25x. As of 3Q ’16, the Company has EBITDA ~40% headroom (or ~$384m in EBITDA) to maintenance covenants, adjusted for the Series B settlement.

Considering the resiliency of the business, tripping maintenance covenants seemed highly improbable. Note of the larger legacy assets acquired by PAH (Arysta, Alent, and MacDermid), the largest drop in adj EBITDA was ~25% peak to trough during the financial crisis, implying significant margin for error – especially considering the current down-cycle in ag.

Incurrence covenants are only tested when the borrower takes an affirmative action, such as when the borrower wishes to take out more debt. This difference is notable as it means that the Company can exceed the incurrence covenant limit without resulting in a violation if it did not take an affirmative action.

The issue now turned to whether the Company would need to take on more debt. The Series B has already been settled, so the most likely consumer of cash would be working capital – particularly in the ag business. Management mentioned that peak working capital would consume ~$200m. Considering the Company’s current cash balance of ~$264m (3Q’16 balance adj’d for Series B settlement of $460m), significant projected 4Q’16 WC release (mgmt. thinks >$100m, probably a highly conservative estimate as FY15 release was ~$250m [albeit FY15 release was likely too high]), in addition to a $500m unused revolving credit facility, year-end ’16 liquidity is substantial, cumulatively amounting to >$800m. As a result, it seemed highly unlikely PAH would need to incur more debt, and thus incurrence covenants are a red herring and maintenance covenants should be the prime focus here.

Consensus Estimates Were Far Too Low 

Per the Company’s revised guidance in 3Q ’16, pre-WC FCF is $180m-$200m. I believed these numbers are highly depressed and do not take into account a) interest expense reductions due to debt repricings, b) elevated corporate costs, c) a strategy change in NA ag distribution, d) unrealized synergies, e) the current ag down-cycle, and f) inefficient tax structure.

In April 2017, the Company completed its third repricing, with aggregate interest savings of $48m from all three repricings, or ~$36m after-tax.

Corporate costs have been elevated to the tune of ~$50m as PAH invested in infrastructure and disproportionately used third-parties to support a larger business. Management expects most of these costs to disappear over time, but not all; I estimated ~$35m goes away, or ~$26m after-tax.

The Company’s distribution strategy change which limited pre-season selling at lower margins reduced EBITDA by ~$40m; its return would portend incremental EBITDA in excess of $40m – I estimated $40m, or $30m after-tax.

As of 3Q’16, $73m in synergies remain unrealized, which would contribute ~$55m after-tax. Peak ag EBITDA excluding corporate costs was $446m in 2013 while the trough was $382m in 2015. Excluding the pre-season change, trough EBITDA would be $422m. Mid-cycle EBITDA would thus be $436m, or an incremental $12m in EBITDA, which is $9m after-tax.

Subtracting ~$340m in annualized D&A and ~$312m in interest expense (adjusted for loan repricings) from mid-point ’16 EBITDA guidance of $758m, I estimated the Company should pay ~$26.5m in cash taxes instead of the $100m-$125m it is guiding to. This would imply incremental ~$86m of cash flow at the mid-point.

According to IR, normalized tax rate would be in the 20s; I used 25% for the above calculations. The Company’s track record in outperforming synergy guidance suggests high likelihood of realization. IR also mentioned they will beat synergy guidance. Management mentioned legal entity rationalization, which would resolve the inefficient tax structure, would take place over 2-4 years.

In total, the above items should raise levered FCF to $424m-$440m on a normalized basis or an incremental $242m, implying shares were purchased at ~8x normalized earning power on my average cost (~$11.50), and that there was a wide gap between the consensus and my view of cash generation capacity. This significant boost in free cash flow would allow for quicker deleveraging.

Non-GAAP/Acquisition Accounting Adjustments Are Reasonable 

The largest add-back to adjusted EBITDA is amortization expense. Thus, PAH could overstate adj EBITDA if it allocates an outsized amount to acquired intangibles. The Company has consistently allocated ~50% of the purchase price to identifiable intangible assets for its ag acquisitions (Arysta, Agriphar, CAS). FMC allocated ~55% of the purchase price to intangibles for its acquisition of Cheminova. The Company has consistently allocated 30%-40% of its purchase price to intangibles for its PFSN acquisitions (Alent and OM Group). In acquiring Rohm and Haas, Dow (PFSN comp) assigned ~34% of its purchase price to intangibles.

The MacDermid acquisition appears to be an outlier, with ~68% of the purchase price being allocated to intangibles. This could be due to the specifics of the acquisition. As the Company’s acquired intangibles allocation generally does not appear out of line relative to acquisitions in similar industries by peers, it does not seem PAH is overstating amortization expense.

Further, it is important to determine whether it made economic sense to disregard these amortization charges.

I believed it did. Platform has ~$2.3b in intangible assets – ~$800m in customer lists, ~$1.5b in developed technology, and a de minimis amount (~$18m) for trade-names and non-compete agreements. The majority (~$1.6b) of these intangible assets came from the Arysta acquisition and thus that will be our focus.

Prior to being acquired by Platform, Arysta filed an F-1 in preparation for its IPO. Note 15 to its financial statements shows that the company had ~$600m in identifiable (ex-goodwill) intangible assets (~$500m in product registration rights and ~$100m in software). When Arysta was acquired by Platform, its identifiable intangible assets were written up to ~$1.6b (pg 12 of 10-Q) and were required to be amortized over 12-20 years as set forth by purchase accounting rules.

Assets, whether tangible or intangible eventually wear out without continual reinvestment. Customer lists, trade-names and non-compete agreements are not the focus here – they are a product of normal business operations. Instead, product registration rights (renamed to “developed technology” by Platform post-acquisition) is the line item that investors should focus on.

Before products can be introduced into the ag markets, they must be registered with the relevant regulatory authorities. Product registration rights also include costs to combine/improve existing AIs. Clearly, amortization of developed technology is an ongoing business cost.

PAH has continued to maintain and occasionally increase its investment in product registrations, which it includes in its CapEx guidance. Furthermore, the Company has slightly increased its R&D spending over the years.

Amortization policies have changed as the Company embarked on its acquisitions, but has remained mostly unchanged since 2014; developed technology amortization has changed from 7-10 years to 5-14 years from 2013 to 2014 and remained the same thereafter, but this change actually resulted in a more conservative policy – weighted average useful life for developed technology was ~10 years in 2013 but increased to 12 years in 2014 following the change; the 2016 figure is 11.6 years.

PAH’s amortization policies is also conservative relative to peers – Dow has a weighted average useful amortization life of 15 years for intangibles, FMC is roughly 18 years, while PAH’s is at 14 years. Hence, it does not appear the Company is adjusting depreciation and amortization policies to shift said expenses to future periods.

As a result, I concluded the amortization add-back for PAH was reasonable.

Changing depreciation and amortization policies related to useful life can also lead to overstated income. The Company has not changed its depreciation policies since inception. Its depreciation policies are also relatively conservative as compared to peers – FMC depreciates land improvements at 20 years, buildings at 20-40, and machinery at 3-18, Dow does not disclose, while PAH depreciates buildings and improvements at 5-20, and machinery, equipment and fixtures at 3-15.

In addition, the Company could overstate income by inappropriately writing down assets and writing up liabilities of the acquired companies which would lead to excess income in future periods as the asset is monetized and the liability write-up is reversed. A high-level check would be to look at whether there were large goodwill changes in the year subsequent to the acquisition.

Initial purchase price allocation shows $993m of goodwill for the MacDermid acquisition, with the eventual number being $990m, initial goodwill of ~$179m for Agriphar, eventually ~$169m, initial goodwill of $281m for CAS, eventually ~$269m, initial goodwill of ~$1,697m (restated to ~$1,770m) for Arysta, eventually ~$1,798m, initial goodwill of ~$1,142m for Alent, eventually ~$1,155m, initial goodwill of ~$116m for OM Group, eventually ~$121m, initial goodwill of ~$67m for OM Malaysia, eventually $79m.

Another check would be to look at whether the Company steps up inventory to fair-value following its acquisitions. In all of its acquisitions, PAH has stepped up its inventory and thus incurred manufacturer’s profit in inventory purchase accounting costs, decreasing income. In contrast, if the Company had wrote down its inventory following the acquisition, it would inflate reported gross profits.

(Although inventory step-up is an add-back to arrive at adjusted EBITDA and hence a larger number would inflate adjusted EBITDA, the magnitude of the adjustment has declined sharply (from $77m in FY15 to $12m in FY16) as PAH turns over its inventory, as one would expect.)

Couple the fact initial and eventual goodwill has not systematically increased by large amounts and the Company has historically stepped up its inventory, it does not seem PAH is overstating income.

Restructuring/Acquisition/Integration Expenses Are Not Ordinary Expenses Being Placed In “One-Off” Bucket 

Cumulatively, the Company indicated the potential for $150m in cost synergies in aggregate over all its acquisitions, excluding MacDermid (MacDermid was the first acquisition and with no pre-existing business, there was no synergies to speak of).

In order to determine whether this synergy target was not used as a rationale to put ordinary expenses in the one-off restructuring business, I looked at it from multiple angles.

First, $150m in cost synergies (proxy for restructuring expenses, given management mentioned 1:1 ratio of synergy to expense) were estimated over $2.9b in sales ($3.6b – ~$700m estimate of MacDermid sales), representing ~5% of sales. This seemed reasonable given most specialty chemical M&A had similar or even more aggressive synergy estimates as a proportion of sales.

FMC Corporation, which acquired Cheminova in early 2015, is probably the closest comp and hence the best example. FMC estimated US$120m synergy potential over 3 years, on 2014 Cheminova sales of DKK$6.8b (or ~US$1.1b), representing 11% of sales. As a result, PAH synergy estimates appeared very reasonable.

Secondly, the Company’s restructuring cash charges largely matched provisions over time, evident by limited restructuring liability ($1.6m, $1.1m, and zero in FY14, FY15, and FY16 respectively) being held on the balance sheet.

Thirdly, PAH’s adjusted EBITDA margins were in-line with peers such as FMC in the low-20% range. If restructuring expenses were instead ordinary expenses and not one-offs, PAH adjusted EBITDA margins (which, among other things, add-backs restructuring expenses), would have been materially higher than its peers.

For the above three reasons, I concluded that it is highly unlikely restructuring expenses were ordinary expenses in disguise being placed in the “one-off” bucket in order to inflate adjusted EBITDA. Acquisition and integration costs have also declined sharply from $93m in FY15 to $33m in FY16, suggesting a similar conclusion.

Spring-Loading

A common method of acquisitive companies to flatter their financial results is to engage in spring-loading. Basically, spring-loading involves holding back revenues prior the closing of the acquisition in order to inflate revenues following the close as these inflated revenues would show up in reported numbers and thus provide the illusion of strong financial results.

If there was any spring-loading going on at PAH, I figured it would most likely occur in relation to the Arysta and Alent acquisitions, simply because these were the two largest acquisitions (acquired for $3.5b and $1.75b respectively) and hence could have provided the most spring-loading ‘ammunition’, so to speak.

Arysta

The Arysta acquisition was completed on 13 Feb 2015, 44 days into a 90-day quarter. This was the only acquisition during said quarter. The 1Q15 10-Q disclosed that Arysta contributed $178m in revenue through the 13 Feb 2015 to 31 Mar 2015 period. It also disclosed pro forma revenues of $622m, as if the Arysta business was owned throughout the entire quarter.

Reported 1Q15 revenues were $535m, implying Arysta sales throughout 1 Jan 2015 to 13 Feb 2015 were $87m (622 – 535). This figure represents ~33% of total Arysta revenues generated during the quarter (87/[87+178]). As it was 44 days into a 90-day quarter, we would expect ~49% of sales of be generated throughout the 1 Jan 2015 to 13 Feb 2015 period. On this basis, it appears the company spring-loaded the Arysta acquisition.

However, I ultimately decided they did not engage in any spring-loading for Arysta because agrochemicals is a seasonal business as it follows the planting season for farmers. Based on Arysta’s geographical exposures, one would expect sales to be lowest in the beginning of the year, ramping up throughout. Hence, Arysta would have very low sales in January, which would generally accelerate through the 4th quarter.

For this reason, I decided generating ~33% of total quarter sales 44 days into a 90-day quarter was reasonable. I would come to a different conclusion if Arysta did not have the seasonal dynamics it has. Moreover, I figured if the company really wanted to spring-load the Arysta acquisition, the differential would be much larger than ~16% (49-33).

Alent

The Alent acquisition was completed on 1 Dec 2015, 31 days prior to the end of the fiscal year. This was one of two acquisitions completed during 4Q15, the other being the $237m acquisition of certain OM Group businesses.

The 2015 10-K disclosed that Alent generated $71m in revenues over the 1 Dec 2015 to 31 Dec 2015 period. The full-year 2015 presentation slides disclosed that Alent generated pro forma (as if Platform owned the business throughout the FY) revenues of $919m. In other words, Alent’s December sales represented ~8% of full-year sales (71/919), which is what one would expect if there was no spring-loading going on and limited seasonality; 31 days in December represents ~8% of the days of a full calendar year (31/365).

In addition, the lack of spring-loading of the Alent acquisition also increased my confidence that there was unlikely to be any substantial spring-loading occurring with the Arysta acquisition.

To further support my view that there was no spring-loading going on, I looked for an absence of sales falling off the cliff following the year of acquisitions for respective segments, where there were no acquisitions for said segments in that following year. I also checked for an absence of days sales outstanding and days in inventory drastically changing, the opposite of which would suggest shenanigans to prop up financials.

There were no ag acquisitions in 2016. 2016 ag net sales fell by 1% y/y, and on a constant currency basis it rose 1%. While segment DSO increased, it was a slight increase. DII fell by a slight amount.

There were no PFSN acquisitions in 2017. 9M17 PFSN net sales grew 6% y/y, 7% in constant-currency. While both segment DSO and DII increased, the increase was slight.

One-Time Gains

PAH has excluded one-time gains from multiple sources that would have inflated adjusted EBITDA if it were not excluded. These one-time gains include 1) legal settlements, 2) settlement of Series B preferred, 3) amendment of Series B preferred, among others.

Other Adjustments

The remaining add-backs – FX losses on foreign-denominated debt, losses on FX forwards utilized to hedge the purchase price of Alent, fair value changes on preferred stock redemption liability, fair value changes on the earn-out, debt refinancing costs – are clearly warranted as they are of the non-operating nature.

Falsifying The Original Thesis 

In my view, the points relating to business quality and most of the accounting factors has probably been proved to be correct. However, I believe my view that consensus estimates were far too low most likely had been incorrect and the assertion PAH was not over-levered could potentially be misguided. In addition, there is an accounting development, separate from the items I identified above, that is a major cause for concern.

The two big reasons I believed the consensus was not pricing in were the savings from an improved tax structure and the anniversary of the pre-season distribution strategy change for the North American ag unit. Collectively, I estimated these would add just shy of $130m in incremental free cash flow, which is roughly half of the guided 2017 free cash flow generation, disregarding changes in working capital.

There were two things, should I observe them, would tell me I was wrong with respect to the above two factors:

  1. Lack of detail on the tax restructuring, especially in management’s public presentations, or its slow implementation
  2. the absence of the anticipated return of $40m of lost EBITDA due to the change in distribution strategy by year-end 2017, or its delay.

While the observation of these points would not necessarily falsify my views, it could suggest a much longer time horizon to realization, thus reducing their value on a net present value basis.

To understand why I believe the two aforementioned points has been falsified, some background is required.

The Company has mentioned multiple times during 2016 and 2017 that it has meaningful opportunity to reduce its cash tax rate over the medium term. While there has been progress in this area, the improvement is very slight; similar cash taxes despite marginally higher profitability.

The previous CFO, Sanjiv Khattri, briefly described the opportunity in the 4Q16 call – that there was significant local profitability in Brazil, China, and Mexico, which could be reduced through tax restructuring. He left PAH shortly after – which wasn’t a good sign. His post was taken up by a former Jarden executive, which made me accept the rumored explanation; because Franklin pushes his team very hard, Khattri could not cope with the stress it entailed.

Regarding the distribution strategy change, management explicitly mentioned on the preliminary 3Q15 call they expected these revenues and EBITDA to return fully by the end of 2017. Since this call, there has been little mention of the topic.

Concurrent with the abandoning of the sale of its ag unit in late August this year, the Company announced it was planning to split the firm into two by 2018, eventually settling on a mid-2018 time-frame. The plan was quite clear – the Company intends to raise equity by IPO’ing a portion of Arysta and using the proceeds to de-lever.

Because incentives drive human behavior, I expected the Company to pull out all the stops by year-end 2017 in order to improve market sentiment and increase its share price in an attempt to sell shares of Arysta at as high of a price as possible, thus limiting equity dilution. This is paramount, especially considering the bad optics following the failed sale of the ag unit to a strategic player or private equity.

However, PAH merely reaffirmed full-year 2017 guidance in 3Q17. I expected the Company to raise full-year guidance in anticipation that the ag unit would recover the lost revenues and EBITDA it had incurred in 2015, as expected at that time. I also expected more concrete details on the tax restructuring such as timeline and magnitude of savings.

Management clearly attempted to improve market sentiment, stating that the ag unit expected ~$1.3b in peak potential sales from its pipeline through 2025 as well as $100m in incremental cost savings over 5 years.

The fact management did not raise full-year 2017 guidance expecting lost ag revenues and EBITDA in 2015 would return suggests 1) they lost these sales to the competition, or 2) the sales and EBITDA returned, but was largely offset by organic declines in the business, or 3) they expect them to return on an extended timeline.

The fact management did not provide more concrete details on the tax restructuring suggests 1) the tax savings could not be realized, or 2) the tax savings could be realized, but it would occur over a longer timeline.

Regardless, this suggested to me I was most likely wrong regarding the assertion that consensus estimates were far too low. As a result, the potential for rapid deleveraging through these FCF boosters was gone, leaving the Company to de-lever primarily through EBITDA growth and not through debt pay-down due to the significant working capital requirements in ag as the business grows.

In addition, I initially expected management to run the business as-is instead of IPO’ing a portion of ag, given their (and my) belief that the Company was not over-levered. The fact management is choosing to go down the IPO route potentially falsifies that belief. Alternatively, it suggested to me management was listening to the Street (which generally bemoaned PAH’s leverage levels) too much.

Furthermore, management mentioned they expected to not be taxed on the proceeds of the Arysta IPO. Taxes are levied if you sell an asset at a higher price than your cost basis, and the fact management did not expect any tax on the IPO potentially implied they did not have high expectations. It also implies bad capital allocation or that they overpaid for Arysta (which still implies bad capital allocation).

The accounting development which concerned me was the usage of accounts receivable factoring both with and without recourse, along with financial guarantees.

While proceeds from financial guarantees and factoring related to the sale of receivables with recourse is properly accounted as financing activities in the cash flow statement, and the sale of receivables without recourse is properly accounted as operating activities in the cash flow statement, they nonetheless bring forward future operating cash flows to the present period. Moreover, this also has the effect of depressing the DSO metric as the receivables amount is smaller than it would be if not for said arrangements.

The Company factored ~$105m and ~$167m of receivables without recourse and ~$71m and ~$38m with recourse in FY15 and FY16 respectively, totaling ~$176m and ~$205m in total factoring in the respective years. PAH had ~$1.02b and ~$1.05b in net accounts receivables and ~$3.58b in pro forma net sales representing consolidated DSOs (using management definition) of ~103 and ~106 of in the respective years, assuming a 360-day year.

Without factoring its receivables, PAH would have had ~$1.2b and ~$1.25b of net accounts receivables, representing DSOs of ~121 and ~126 in FY15 and FY16 respectively. If we assume ag had similar DSOs for FY14 and FY15 (i.e. 134 days) – as the ag acquisitions had not been consummated in 2014 – and that the ratio of PFSN and ag DSOs were similar for 2014 and 2015 (i.e. consolidated DSO accounts for ~48% of the sum of both segment DSOs), consolidated DSO would be ~100 days in FY14, a year where no factoring transactions had occurred.

This suggests consolidated DSOs, excluding the impact of factoring, had substantially increased from FY14 to FY15 and FY16, a clear red flag – one I should have caught much earlier but did not. It thus appears the Company factored receivables in order to conceal the large increases in DSOs. 

Furthermore, the assumption that ag had similar DSOs for FY14 and FY15 is a conservative one because FY15 ag DSO includes the impact of factoring. I am also not cherry-picking FY15; FY16 consolidated DSO accounts for ~50% of the sum of both segment DSOs).

Management is targeting 3.5x net debt/adj EBITDA and 4.25x (mid-point) net debt/adj EBITDA leverage levels for ag and PFSN respectively, by year-end 2019. Assuming a 45/55 split between ag EBITDA and PFSN, this implies $369m and $451m in ag and PFSN adj EBITDA respectively at the mid-point of FY17 guidance.

Using the target leverage levels, the implied equity raise from the Arysta IPO is ~$1.8b (~$5b in net debt – [$369m * 3.5 + $451m * 4.25]). PAH aborted the sale of the ag unit to private equity and strategics because it failed to meet valuation expectations of more than $4.5b. If we conservatively use $4.5b as a base case valuation for Arysta, this suggests 40% of Arysta shares would be sold in the IPO.

As a result, shareholders of PAH would then own a company generating ~$672m in adj EBITDA ($369m * 0.6 + $451m), after accounting for the non-controlling interest of Arysta.

Assuming all IPO proceeds are allocated to principal debt payments, net debt would fall to ~$3.3b (3Q17 net debt is ~$5.1b) and gross debt would be ~$3.7b (3Q17 cash is ~$400m), costing ~$222m in interest expense (assuming a 6% weighted average interest rate, in-line with its current average; $330m in FY17E interest expense on $5.5b of gross debt).

Keeping cash taxes and net capex steady per FY17 guidance ($140m at mid-point and ~$100m respectively) – a conservative assumption for cash taxes given lower interest expense means higher profitability and thus higher cash taxes, PAH would generate $210m in levered free cash flow. After pension contributions, it would be ~$190m.

With ~300m in diluted shares outstanding and current trading prices of roughly $9.50/share, PAH does not have any upside, assuming a 15x multiple. It however has limited downside.

While it is true if we utilize management’s adj EBITDA growth targets through 2019, shares would see significant upside, these targets are not guaranteed to be hit, especially considering the current down-cycle in ag, and tepid end-markets for PFSN. A 15x multiple also suggests said targets are likely priced in. Moreover, the growth in DSOs excluding impact from factoring is a significant cause for concern and could potentially suggest a portion of adj EBITDA is unsustainable. Shares could be good value if management somehow manages to get a very good price for Arysta, which would limit equity dilution. It is however not a bet I am comfortable making.

As a result, I exited my long position in PAH and (luckily) broke even. Better to be lucky than good, I guess. Given the incoming IPO of Arysta, PAH could potentially develop into a special situations play, and I look forward to analyzing it once more following the offering.

MercadoLibre: Finance Company Masquerading As An E-commerce Company

From an e-commerce point of view, the prevailing perception regards MercadoLibre (NASDAQ:MELI) as LatAm’s Alibaba. Like Alibaba, MELI primarily operates marketplaces, has no direct-sales business, outsources delivery to third-party couriers, and thus earns very good margins. I agree with the prevailing perception, though not for the aforementioned reasons. Instead, I believe MELI’s accounting is just like BABA’s – misleading.

MercadoLibre is highly dependent on consumer credit financing. While the company has experienced strong top-line and GMV growth, an overwhelming portion of said growth was credit-driven.

For the years 2014, 2015, and 2016, installment payments represented 53.9%, 58.5%, and 55.7% of total payment volume respectively. In addition, 79%, 79.3%, and 77.2% of payment volume were funded using credit cards in the respective years. Hence, the company’s business appears to be largely dependent on consumer credit financing.

While bulls may downplay this credit dependence based on the premise credit card and installment payments are typical in LatAm due to the nature of the banking system, such an assertion is inappropriate given provisions for accounts receivable are extremely high, nearing 30% of gross AR in FY16. AR write-offs peaked at 62% in FY12 and was last reported in FY14 (I wonder why…) where the figure was 38%. These data points suggest the Company’s growth has been significantly driven by low-quality credit, a very real risk.

A significant portion of the company’s revenues are driven by factoring transactions. MELI regularly factors its credit card receivables to financial institutions. Aggregate gains included in net revenues resulting from factoring transactions, net of fees, amounted to ~$71m, ~$96m, and ~$120m for FY14, FY15, and FY16 respectively, representing ~13%, ~15%, and ~14% of total revenues and ~39%, ~37%, and ~34% of non-marketplace revenues in the respective years.

Because the amount factored is at management’s discretion, MELI has a mechanism to materially increase EBIT during a slow quarter if need be. The EBIT comparison is appropriate given said transactions were recorded net of fees. These transactions accounted for ~59%, ~69%, and ~66% of reported EBIT in FY14, FY15, and FY16 respectively.

The company has minor operations in Venezuela, which contributed 10.4%, 6.2%, and 4.4% of revenues in FY14, FY15, and FY16 respectively. MELI has benefited from Venezuela hyperinflation as a result of its unwillingness to deconsolidate its Venezuelan operations.

Essentially, MELI has been translating Venezuelan revenues to USD at massively overvalued exchange rates. Although the company has been slowly adjusting the exchange rate at which Venezuelan revenues are translated back to USD, these adjustments lag reality.

MELI measured its Venezuelan financials at the SIMADI exchange rate of 198.70 BsF per USD as of December 2015. As of December 2016, the company re-measured said financials at 674 BsF per USD. However, press reports in late 2016 and early 2017 suggests a black-market rate of 1,500 to 3,000 BsF per USD, suggesting MELI’s Venezuelan financials were overvalued by roughly 2.2x to 4.3x.

MELI recorded $37.2m in Venezuelan revenue in 2016, a decline of $3.3m or 8.1% from the $40.5 recorded in 2015, negatively impacting overall revenue growth by 0.5%. If 2016 Venezuelan revenue were recorded at the black-market rate of 1,500 to 3,000 BsF per USD, the overall revenue growth of 29.6% would have been negatively impacted by 3.5% to 5%. If MELI followed other multinationals such as Clorox and P&G in deconsolidating their Venezuelan operations altogether, the negative impact would be 6.2%. Thus, the acceleration from 17.1% revenue growth in FY15 to 29.6% in FY16 is less impressive than it seems.

Earnings quality have been declining over the past few years. In FY16, 19% of pre-tax earnings stemmed from interest income and other financial gains, as compared to 9% in FY11, suggesting an increased proportion of the company’s earning capacity has not stemmed from its e-commerce operations.

In addition, interest income and other financial gains seem to stem from high-inflation jurisdictions, given interest income and other financial gains as a % of cash & short-/long-term investments have generally ranged from 4%-6% over the past 6 years. With US government securities way below these rates, it is likely MELI has turned to riskier jurisdictions to park its cash. This is further supported by the fact the Company disclosed ~$259m in sovereign and corporate debt securities as part of its short-/long-term investments.

Furthermore, the company capitalized ~$14.5m and ~$20.7m in website development costs in FY15 and FY16 respectively, representing 9.6% and 11.1% of pre-tax earnings. Factoring of credit card receivables, as mentioned above, contributed to ~41%, ~45%, and ~45% of pre-tax earnings in FY14, FY15, and FY16 respectively.

(It is interesting to note the Company has provisions of ~30% of gross AR but provisions for credit card receivables are generally sub-1%. This disparity could be explained by management’s intention to factor the overwhelming majority of its credit card receivables, which would not necessitate substantial provisions for credit losses – the financial institution eats the credit losses instead!.)

In short, while bulls may attribute the Company’s strong profitability to its e-commerce operations, over half of its profitability is actually driven by 100% EBIT margin factoring of credit card receivables and significant investments in high-inflation jurisdictions.

Hence, MELI appears to be a finance company masquerading as an e-commerce company. Finance companies are generally valued within a range of 0.5x-2.0x tangible book – MELI trades at somewhere around 50x tangible book.

ZTO Express: Related party subsidization, overstated cash flow, overpaying for acquisitions

ZTO Express (NYSE:ZTO) has incredible profitability. It is the most profitable express delivery company in China by operating margins. 2016 EBIT margins are ~28%. The ttm figure is almost 30%. Contrast this to its peers such as SF Express, YTO Express, Yunda Express, STO Express, which had 2016 EBIT margins of ~6.4%, ~10%, ~22%, and ~17% respectively.

To be fair, one of these is not like the others – SF Express fully inhouses the delivery process whereas the rest, including ZTO, are responsible for sorting/packing and line-haul transportation, with the first-mile pick-up and last-mile delivery portions of the delivery process being outsourced to third-party franchisees.

So it is quite clear why ZTO is more profitable than SF Express, but it is unclear why ZTO is more profitable than YTO, Yunda, and STO, especially considering YTO is ~1.7x the size of ZTO while STO is roughly the same size as ZTO, with Yunda being slightly smaller, in terms of 2016 revenue. Presumably, YTO would be the most profitable of the lot.

There is an explanation for ZTO’s stunning profitability, but it is not the explanation the sell-side or longs would be comfortable with. In my view, it appears ZTO’s profitability is largely subsidized by related parties.

ZTO has a special arrangement with Tonglu Tongze, a transportation operator, which works exclusively for ZTO. Tonglu Tongze is also the company’s largest related party. The company outsources part of its line-haul needs to Tonglu.

Curiously, despite certain of ZTO’s employees beneficially owning 65.1% of Tonglu, the fact Tonglu works exclusively for ZTO, where three mid-level ZTO managers are nominee shareholders of Tonglu, where historically ZTO employees served as management of Tonglu gratis, and ZTO having participated significantly in the design of Tonglu during its formation, as disclosed in the 20-F, ZTO’s management do not deem the company the primary beneficiary of Tonglu and hence Tonglu remains unconsolidated in the Company’s financials.

Crucially, ZTO appears to be under-paying Tonglu, which suggests ZTO is over-earning. ZTO delivered ~4.5b parcels in FY16 and garnered ~$9.4b in express delivery revenue, which implies revenue per parcel of ~$2.08. ZTO paid Tonglu ~$853m in fees that year.

Given the fact Tonglu works exclusively for ZTO and has a truck fleet which is ~44% the size of ZTO’s (1,270 of Tonglu vs 2,900 of ZTO), and that Tonglu’s truck fleet is ~31% of the sum of ZTO’s and Tonglu’s fleet, it seems appropriate to assume Tonglu delivered ~31% of the total, or ~1.4b parcels. In other words, Tonglu was paid ~$0.62 per parcel, which is roughly a third of what ZTO received per parcel.

While some may argue the delta between the fees paid and revenue received per-parcel is due to Tonglu handling line-haul transport and ZTO handling line-haul in addition to sorting/packing, this rebuttal is invalid given sorting/packing costs only account for ~10% of the fulfillment expense per parcel while line-haul accounts for ~37%, according to various sell-side reports. Even ZTO’s own line-haul transportation costs per parcel were ~$0.83, suggesting Tonglu is massively unprofitable.

In other words, sorting/packing revenue should account for ~21% of ZTO’s revenue per parcel, with the remainder going to line-haul. As a result, line-haul revenue per parcel at ZTO would be ~$1.64 per parcel. If ZTO paid Tonglu a comparable fee per parcel, ZTO’s FY16 EBIT margins would decline to ~14% from ~28%.

In addition, the company’s cash flow appears overstated. ZTO generated $2.5b in net cash from operations in FY16. While total working capital benefit was merely ~$38m, this figure severely understates the benefit derived from increased current liabilities. Specifically, ZTO includes payables related to property and equipment – presumably a capital expense – in other current liabilities.

Payables related to property and equipment increased by ~$505m in FY16. Because said metric in included in other current liabilities, it effectively transforms what should have been a cash outflow from investing activities into a cash inflow from operating activities. The increase in payables related to property and equipment accounted for ~20% of net cash from operations in FY16.

ZTO also appears to be overpaying for its acquisitions. Over the past few years, the company has been acquiring its network partners. In October 2015, ZTO acquired 16 network partners for a total consideration of ~$1.34b, largely paid with stock.

(Network partners are basically first/last-mile franchisees which have no competitive advantage given a hugely fragmented industry where every player accounts for a minuscule portion of the market.)

This appears excessive, given said network partners generated only ~$62m in net income basis, as implied by the delta between reported net income and pro forma net income, excluding the deemed disposal gain and adjusting for net income generated since the acquisition date included in reported net income.

In other words, it appears the Company paid ~22x FY15 net income, a rather hefty multiple for a commoditized business. This overpayment is compounded by the fact the company pays largely in stock, not cash.

A commoditized business, presumably earning returns roughly equal to its cost of capital, reinvesting at returns roughly equal to its cost of capital (the result of the absence of durable competitive advantages), deserves at most a 10x multiple, assuming a 10% cost of capital.

These acquisitions also allowed for a re-measurement of the Company’s pre-existing equity interests in said partners, enabling the Company to record a gain on deemed disposal of equity method investments of ~$224m, representing ~17% of FY15 net income. As an aside, other operating income included in operating expense accounted for ~2.5% of FY15 net income.

Finally, ZTO has significant VIE exposure.

Due to PRC legal restrictions on foreign ownership in companies that provide mail delivery services in China, shareholders of ZTO do not own the key PRC operating subsidiaries such as ZTO Express but instead own the offshore holding company ZTO Express (Cayman).

The offshore holding company has entered into contractual arrangements with the domestic PRC company in order to gain effective control – though such control is inferior to actual ownership and the PRC subsidiaries are treated as variable interest entities, which are majority-owned (73.8% interest) by ZTO management.

The risks of the lack of actual control is material. An infamous case was GigaMedia, where the owners of the VIE essentially breached the contractual arrangements and GigaMedia shareholders lost control of major assets.

The degree of VIE exposure depends on the amount and type of assets held by the VIE. In the case of ZTO, there is extremely high VIE exposure given the VIE contributed 100% of the Company’s consolidated revenues and accounted for 41% of total assets as of FY16.

While the risks of significant VIE exposure is mitigated by the fact ZTO management controls both the Company and the VIE along with the fact they own 53.4% of ZTO – which implies an alignment of interests, the fact remains the risk is present.

In conclusion, ZTO may appear to be a significant beneficiary of the secular trends in China e-commerce, but a closer look suggests the company has many red flags, with the major ones being – strong profitability largely subsidized by an unconsolidated related party, materially overstated cash flows, poor capital allocation, and significant VIE risks.

Autobytel – peak auto, shifting earnings, under-reserving, significant cost capitalization

Autobytel (NASDAQ:ABTL) recently reported 2Q17 earnings and it wasn’t pretty. 2017 guidance was lowered due to lower conversion rates and anticipated auto sales decline.

As of the time of writing, shares trade for $7.66 or roughly $100m in market cap. Trailing free cash flow is about $20m. So on reported figures shares look dirt cheap at a ~5x multiple.

However, trailing free cash flow has been inflated by certain items which are probably not sustainable – 1) share-based compensation of ~$4m and 2) ~$4m of accounts receivable are the largest ones. Payables and accrued expenses are mostly reversals of increases. Adjusting for 1) and 2) and you get roughly $12m in free cash flow.

Alternatively we can look at things on a pre-working capital basis but post-share based comp – free cash flow would then be roughly $15m representing a ~6.7x multiple.

However, there are a couple of reasons why the above figures are not representative of a normalized level. One reason is obvious – we are close to a peak in the auto cycle, and I have no strong opinion regarding a mid-cycle free cash flow figure other than “way lower than the trailing figure”.

The other reason is non-obvious. Based on its historical record, the company appears to have shifted revenue and expenses from one period to the next, may be under-reserving its accounts receivables, and may be improperly capitalizing operating expenses.

Analysis of accounts receivables and accounts payable suggests the company has been rearranging revenue and expenses from one period to another in order to show consistent multi-year profit growth.

Accounts receivable days has increased from 51.5 to 78.4 from ’08 to ’16, representing a massive ~52% increase and a 9-year peak. Interestingly, receivable days were at their trough during the ’08 recession year. Assuming the average AR days for ’11 and ’12 is representative of a normalized figure, the Company deflated AR by $6.6m from ’09 to ’13 while inflating AR by $5.9m from ’14 to ’16.

Accounts payable days have increased from 25.4 to 36.1 from ’08 to ’16, representing a massive increase of ~42%. Assuming the average AP days for ’08 and ’09 are representative of a normalized figure, the Company inflated AP by $1.5m from ’09 to ’13 while deflating AP by $2m from ’14 to ’16.

It thus seems the company deflated earnings by $8.1m from ’09 to ’13 in order to later inflate earnings by $7.9m from ’14 to ’16. The latter figure represented 40% of EBIT and 67% of net income over the ’14 to ’16 period.

earnings management.png

Clearly, the above is driven by my assumptions that ’11-’12 are representative of normalized AR days and ’08-’09 are representative of normalized AP days.

Here’s my reasoning – ’11-’12 was generally a decent period for auto volumes and should not include a lot of uncollectible accounts and so AR days are not artificially high.

For AP days, ’08-’09 was a recessionary period for auto volumes, and hence should suggest artificially high AP days. AP days in excess of this artificial highs should thus pose a huge red flag.

The business also hasn’t changed much over the years, so there doesn’t seem to be any reason why the dynamics of the company’s accruals should change drastically.

Provisions as a percentage of gross accounts receivables were ~11% in ’08 and ’09 but have declined to a mere ~3% in ’16. While ~11% provisions against gross receivables is likely the result of the recessionary conditions experienced during those years, the Company’s current receivables reserving policies are highly optimistic even compared to recovery periods – reserves against receivables were ~6.4% in ’10.

The Company’s potential under-reserving of receivables is worsened by the fact despite significant increases in receivables days, which suggests an increased likelihood of customer non-payment, reserves against receivables have instead declined significantly.

The Company capitalized $1.7m, $1.5m, and $0.6m of website development costs over the years ’16, ’15, and ’14 respectively. Worryingly, the significance of these capitalized costs have grown over said years. Because these costs were capitalized and not incurred as an operating expense, ABTL’s EBIT is over-stated by 28%, 19%, and 10% over the respective years mentioned.

In a nutshell, ABTL looks dirt cheap at first sight, but if you look under the hood, things aren’t what they seem. As always, I may be partially or totally wrong, and I welcome people to convince me as to why I am wrong.