Project: Quantitative Proposal of a Distressed Bond

The following article outlines a task I was asked to tackle some time back as an assessment for a job I was applying for. The following post is my personal approach to selecting a Distressed Bond for investment. To go directly to this result and the financial reasoning of it skip to 4 Results and Summary.

But first, you may wonder — why invested in the debt of a Company that is currently “distressed”? And what qualifies “distress”?

According to Investopedia: Distressed securities are financial instruments issued by a company that is near to — or currently going through — bankruptcy. Distressed securities can include common and preferred shares, bank debt, trade claims, and corporate bonds.

The philosophy of investing in the debt of a distressed Company is that it can successfully pay its debt coupons and when the Company is (assumed to be) liquidated — the debt holders (i.e. us as the investor in this case) can be paid off the full face value of the bond. These bonds should belong to a Company with a high chance of default and it is up to the investor’s investment approach to be able to potentially find a Company that is in this circumstance but may be able to pull through.

This article is reasonably long and verbose so for those who would like to understand my thinking briefly I outline it in the following paragraph. For anyone technically interested in the methods I used please see the references where I share the paper from which I apply a logistic regression model trained on historical data but re-using its coefficients. The remainder of my method is now described:

  • I obtained a dataset from an iShares fund holding high-yield securities. I wanted an expansive list of high-yield bonds/assets with things like maturity and different yields.
  • From here I used my investment thesis to shortlist issuers in key industries that may be poised for recovery in the post-Covid world.
  • I further narrowed the shortlist based on the issue and issuer's YTM, fundamentals (such as CFFO, Net Income, and Shareholder Equity), and whether the issuer data was public or not.
  • I arrived at a shortlist of 8 companies from which I calculated key financial ratios according to [1] and used a pre-trained logistic regression equation to output the probability of distress from the shortlisted companies.
  • I acknowledged that the data used in the paper to obtain the coefficients may not be as relevant to today’s economic conditions, and so discounted the probabilities that were outputted but still used it as a guideline.
  • From the shortlist I decided to choose an issuer who I judged to be the most likely to pay back its creditors based on its fundamental financial ratio data, its overall adherence to my investment thesis, and the output from the logistic regression equation. (For those most interested in my financial findings jump to 4 Results and Summary to see which issuer and security I think is of interest to a debt investor and why).

1 Approach

The following subsections discuss how I tackle the problem at hand — choosing an optimal distressed bond opportunity. It outlines the approach I will take to make my proposition

1. 1 Investment Thesis

First and foremost, I will define an investment thesis for what I am looking for. The investment thesis will serve as a strong guideline while filtering through opportunities so an optimal shortlist can be formed based on my investment outlook.

My Investment Thesis Definition: I am looking for issuers that are facing financial distress due to negative net income and decreasing cash flows but are operationally stable to the point they can cover their interest expense in the short to medium term (6 months to 5 years) with positive cash flow from operations, or with a large total asset base that can be liquidated to cover their liabilities — specifical companies with relatively high shareholder equity and current ratios. I will specifically target companies issuing debt in industries shocked by the coronavirus slowdown but have the potential to bounce back once lockdowns are lifted such as firms in the consumer goods, energy, and transportation industries.

The following graphic from Bloomberg outlines the most bankruptcies filed by sector for each year since 2008. This information is useful in identifying key industries that have been most heavily disrupted economically by the coronavirus slowdown in 2020.

From the above graphic, it is visually clear that the largest increase in bankruptcies from 2019 to 2020 has occurred in the consumer discretionary services sector. This sector includes the travel and transportation sub-industries — which both have good potential to retrieve lost cash flows once lockdown restrictions are eased. For that reason, it will be shortlisted along with the general cyclical consumer discretionary goods sector. The other sector that has increased bankruptcies considerably is the energy sector. This sector includes oil and gas which has taken a hit due to the general slowdown of the wider economy. The long-term outlook on non-renewable industries is uncertain, however for the short to medium term in which I am concerned, the potential for cash flow recovery over the coming 1–5 years is plausible and hence it will also be considered.

1.2 A Good Data Source of Potential Opportunities

For the next step, I require access to a universe of potential distressed companies that are actively issuing, or have issued, high-yield debt. Many articles and blog pages exist outlining their respective author’s shortlist of issuers — one such shortlist is Bloomberg’s US Bankruptcy Tracker [7] — but this will not suffice a large enough pool to screen the best opportunities according to my own investment Thesis.

Access to such data in its entirety is usually not free and comes at a cost, either a financial or a time-based cost. In order to solve this problem, the following solution will be taken; existing high yield bond ETFs are available on the market which must disclose documentation of their holdings. One such ETF is iSharesBroad HYB ETF[5]. This ETF contains 1281 holdings of high yield bonds from various issuers. This will be sufficient for a data source of issues and issuers and is the output of the sourcing stage of this task. The following subsection will describe the methodology I will take in selecting an optimal opportunity once a shortlist has been created (see 2 Shortlist Opportunities for shortlisting details).

1.3 Analysis Methodology

I will utilize academic research in the field of classifying financially distressed companies to generate probabilities of distress for each of the shortlisted issuers. Papers such as [1], [2], and [3] utilised the machine learning techniqueLogistic Regressionin order to classify companies as either distressed or not distressed, and attributed a probability to whether the company is distressed — with distressed being defined as likely to go bankrupt. The independent variables utilized to assert a firm’s distress were financial Ratiosobtainedfrom calculations of the quantities in their respective financial statements (see Appendix for a description of the choice of financial ratios). Since I do not have labeled training data to create my own logistic regression algorithm or the time to retrospectively label which companies have gone bankrupt and retrieve their historical fundamental data, I will make use of the pre-trained model coefficients (see Appendix) in [1] to calculate the probability of distress for the shortlisted companies from the i Shares Broad HYB ETF so to judge which have the greatest likelihood of paying back their creditors given their financial ratios.

The following equation is used in [1] (which also provides the coefficients) to classify whether a security is distressed or not:

The β0 and β1 terms are not included in the paper and hence will not be used. The definitions of the required ratios are in the Appendix. Given that I am using a list of pre-defined high yield issuers who are under some distress as a criterion for being included in the ETF, issuers with the lowest probability for distress will be viewed as most appealing for selection. The usage of this formula is different to that in [1] — in the paper it was used for classifying distressed and non-distressed companies, while here I am working off the assumption they are all distressed and I am looking for the least distressed i.e. the issuers with the lowest output probabilities of distress.

The ratios provided in the above formula were found to be the most effective in predicting financial distress[1], so these financial ratios will also be used for comparative analysis within a smaller subset of shortlisted companies. I note that the ratios used in the paper may not be the most suitable for the current universe of issuers I am looking at (due to changing economic conditions — the model was trained on pre-2008/9 data)however to obtain the optimal predictors would require extended research into the dataset (namely scraping financial ratios for every issuer and verifying how well they have met interest payments historically) which I will refrain from doing due to time constraints and under the assumption that the research previously conducted along with my discretionary analysis of the fundamental data and ratios should suffice

2 Shortlisting Opportunities

Firstly I load the CSV file containing the iShares ETF’s holdings into a pandas dataframe object in a PythonJupyter Notebook environment. From here I splice the dataframe such that I only obtain issues from the energy, Transportation, and Consumer Cyclical Goods industries — according to my defined investment thesis. This provides a list of 100s of issues but I would like to further shorten this down by sorting it according to the YTM % from each issue and removing all issues with YTM less than 8% — this is to isolate the most distressed bonds with the highest potential of return. The number of debt issues in this list is 38.

Next, I further shorten the list to exclude issues that have maturities after 2025. The thesis is that we are looking for companies who are struggling to stay solvent and hence longer maturities are undesirable due to risk of company default. The shortlist now contains 24 debt issues. All non-public companies were also removed since I require access to financial statements to complete my analysis. Also, any repeated companies with two issues in the list had their lower YTM issue discarded. The remaining shortlist contains 18 issues.

From here, I use my personal discretion along with the defined investment thesis to select 10 companies for a final shortlist — my personal discretionary criteria consist of choosing companies with the highest YTMwithin the initial industries I defined, specifically consumer cyclical, transport, travel, and energy.

Finally, the following shortlist of 10 issuers with their respective high yield issues are selected:

3 Analysis Output

The following table contains the required fundamental data for the calculation of the 10 ratios utilised in the equation quoted in 1.3. Navios and Diamond Resorts have been discarded due to a lack of public data on the SEC website. For the remaining issuers, the data is acquired from scraping the SEC Company Search tool [8]:

The values in are Millions ($).

Using the Logistics Regression equation from [1], the probability the Company is in distress is given in the first column below:

The probabilities presented here are a guideline more so than factual truth. I have used them as a quantitative assurance that certain companies are more prone to distress than others. The model in [1] was trained on companies prior to the 2008/9 financial crash. Since then economic circumstances have changed — for example, interest rates have been decreased significantly encouraging companies to take on more debt. However, the model serves as a guideline. From the probability values, every company selected shows at least an 11% chance of distress — the lowest probability of distress is for Laredo Petroleum Inc but this could potentially be much higher.

4 Results and Summary

During my analysis, the following issues were considered for the proposition for the following reasons:

  • Transocean Guardian LTD’s 9.5% issue due to the Company’s current ratio above 2 (2.05) and the shorter time to maturity (Jan 15, 2024) than the other companies.
  • PBF Holding’s 7.25% issue; although the Company has the highest probability of distress this consideration was due to its large asset base (c.$11161M) and the Company’s positive shareholder equity which indicate it has the ability to sell assets to cover its liabilities as suggested in [4].
  • Nabors Industries Inc’s 5.75% issue was considered due to their positive CFFO, relatively low probability of distress from the logistic regression equation, and relatively high and positive CFFO-I/I ratio

The following issues were discarded after my analysis for the following reasons:

  • American Airlines Group’s 3.75% issue due to their large negative CFFO (-$3680M).
  • AMC Entertainment Holdings Inc’s 10.50% issue due to their large negative CFFO (-$771.6M).
  • Calumet Speciality Products Partner’s 11% issue due to their negative CFFO-I/I ratio (c.-0.3) which occurs due to the fact their interest expense is greater than their CFFO.
  • Weatherford International LTD’s 11% issue due to their negative CFFO-I/I ratio (c.-0.04) occurs due to the fact their interest expense is greater than their CFFO.

After considering each opportunity I have settled on proposing Laredo Petroleum Inc’s 9.50% loan note maturing on Jan 15, 2025, as the distressed bond invests in. The reason for this selection is down to the following:

  1. The Company has an impressive Cash Flow from Operations to Interest Expense Ratio (CFFO-I/I)compared to the other companies on the list (c.2.5). This suggests a strong capability to pay back creditors with current cash flows from operations.
  2. 2. Due to the previous point, the logistic regression equation suggests Laredo has the lowest probability of distress/ bankruptcy due to its ability to pay back creditors in the short term — this should be somewhat discounted but it is promising that the technology backs up my discretionary decision.
  3. The Company has positive shareholder equity (c.$141.5M) and hence has a large enough asset base to cover its liabilities and pay back shareholders, and hence will be able to pay back creditors given amass liquidation event.
  4. The Company has a current ratio (CA/CL) of c.1.40 (1.395) and hence can manage to cover short-term liabilities with current assets.
  5. The industry the Company operates in — oil and gas — falls within the initial investment thesis; an economic recovery after Covid lockdowns could enable firms within the industry to recover many lost cash flows

To conclude, my suggestion for a distressed bond to invest in is Laredo Petroleum Inc’s 9.50%bond maturing on Jan 15, 2025 trading at 94.38c on the dollar with a total YTM of 11.31% (at the date of this writing, Feb 2020).

References

1] A. Kamaluddin, ”Monitoring Distressed Companies through Cash Flow Analysis”, Procedia Economics and Finance, vol. 28, pp. 136–144, Dec, 2015.

[2] P. Jantadej, ”Using the combinations of cash flow components to predict financial distress”, Jan, 2006.

[3] C. S. Rodgers, ”Predicting Corporate Bankruptcy Using Multivariant Discriminate Analysis(MDA), Logistic Regression and Operating Cash Flows(OCF) Ratio Analysis: A Cash Flow-BasedApproach”, San Francisco, 2011.

[4] https://www.nasdaq.com/articles/bankruptcy-looms-over-u.s.-energy-industry-from-oil-fields-to-pipelines-2020-04-23

[5] https://www.ishares.com/us/products/239565/ishares-iboxx-high-yield-corporate-bond-etf

[6] https://www.bloomberg.com/news/articles/2020-10-20/u-s-bankruptcy-tracker-debt-war-escalation-adds-to-debtor-woes

[7] https://www.bloomberg.com/news/articles/2021-01-05/u-s-bankruptcy-tracker-pandemic-spurs-most-filings-since-2009[8] https://www.sec.gov/edgar/searchedgar/companysearch.html

Student of Quantitative Finance, Computer Science, and Applied Probability.