BanruptcyPrediction

Bankruptcy Prediction

Contents

Case Brief 2

Background 2

Accounting Analytics 4

Data and Additional Resources 7

Case Brief

Your team has just been engaged by the central office of a large audit firm. A recently promoted partner with an interest in data analytics has approached you to work on quantifying the risk related to bankruptcy. Your team has been approached due to your expertise in determining how publicly available accounting and other information can quantify bankruptcy risk and your expertise in visualization that will help communicate potential trends in bankruptcy risk.

Background

Why is bankruptcy risk important to measure? Forecasting a debtors’ ability to repay its financial obligations is an important activity for creditors who desire to be repaid. Due to the “dead weight” costs of financial distress it also has significant impact on equity investors. Some of these costs include increased interest rates, decreased productivity due to poor employee morale, lost investment opportunities due to capital limitations, and increased legal costs. Because these costs begin to be incurred prior to actual financial distress, for providers of capital it is important to be able to predict distress as early as possible. Most bankruptcy prediction models have taken this view and tried to extend the forecast period to multiple years. Counter to this long-term focus auditors have generally focused on a single year forecast of financial distress to help with “going concern” opinion decisions.

Bankruptcy risk and going concern: One of the basic assumptions in the preparation of financial reports is that the business is a “going concern”. This assumption means that the business is expected to be in operation for at least the following year. The valuation of all the assets and liabilities of the business in the financial statements is based on this assumption that they will continue to be operated rather than liquidated. If there is considerable risk that the business will not be in operation at the end of the following year the auditor should issue a “going concern” opinion. Auditors are reluctant to do this due to the dramatic impact it can have on a firm. However, if the auditor does not issue a going concern opinion and the business encounters financial difficulties within the next year the auditor faces risks from being held responsible by stakeholders in the firm.

Under the going concern basis of accounting financial statements are prepared with the assumption that the business will continue its operations for a reasonable period. In preparation of the financial statements, the company’s management may be required to perform an evaluation of the company’s ability to continue as a going concern (explicitly required by both FASB and GASB).

The auditors are required to evaluate a management assertion that the going concern basis of accounting is appropriate. As set forth in AU-C 570.10, the objectives of the auditor are:

“a. To obtain sufficient appropriate audit evidence regarding, and to conclude on, the appropriateness of management's use of the going concern basis of accounting, when relevant, in the preparation of the financial statements

b. To conclude, based on the audit evidence obtained, whether substantial doubt about an entity's ability to continue as a going concern for a reasonable period of time exists

c. To evaluate the possible financial statement effects, including the adequacy of disclosure regarding the entity's ability to continue as a going concern for a reasonable period of time”

In assessing bankruptcy risks the auditor should discuss managements evaluation with management and determine if conditions or events that raise substantial doubt have been identified along with managements’ plan to address them. The auditor evaluation of management’s evaluation should cover the same period of time as that used by management (as required under accounting standards) and include consideration of whether management’s evaluation includes all relevant information of which the auditor is aware. The auditor should also inquire regarding knowledge of conditions or events beyond the period of management’s evaluation that may have an effect.

How does the Auditor deal with going concern issues? The possibility of bankruptcy is a big deal, and auditors have the option to acknowledge their concerns about the inability of a company to be a going concern in the audit report. Depending on the basis assessed by managers (a going concern or not a going concern) and the auditors judgment there are differing implications for the auditor’s report.

1. Going Concern basis used by managers and the auditor judgment is that it is inappropriate:

• In this situation the auditor should express an adverse opinion.

2. Going Concern basis used by managers and auditor judgment is that it is appropriate, but conditions and events have been identified that question the ability of the firm as a going concern:

• If the auditor concludes that there is substantial doubt about the entity’s ability to continue as a going concern then the auditor should include an “emphasis of matter” paragraph in the report. In this paragraph the expression of the existence of substantial doubt should NOT be conditional.

• If there is not adequate disclosure in the financial statements about the ability of the entity to continue then the auditor should issue either a qualified or adverse opinion.

For example, the emphasis of matter example when management’s plans are not sufficient to remove going concern doubt can be worded as:

The accompanying financial statements have been prepared assuming that the Company will continue as a going concern. As discussed in Note X to the financial statements, the Company has suffered recurring losses from operations, has a net capital deficiency, and has stated that substantial doubt exists about the Company's ability to continue as a going concern. Management's evaluation of the events and conditions and management's plans regarding these matters are also described in Note X. The financial statements do not include any adjustments that might result from the outcome of this uncertainty. Our opinion is not modified with respect to this matter.

Alternatively, an emphasis of matter example when plans are sufficient to remove doubt could be worded as:

As discussed in Note X to the financial statements, the Company has suffered recurring losses from operations and has a net capital deficiency. Management's evaluation of the events and conditions and management's plans regarding these matters are also described in Note X. Our opinion is not modified with respect to this matter.

Accounting Analytics

Bankruptcy is clearly a big issue. Accounting analytics have sought to quantify bankruptcy risk for a long time. This is perhaps the accounting analytic that has the longest historical precedent we will cover in our course with studies on this topic dating back to at least 1932 – really highlighting how important understanding bankruptcy risk is!

Early evidence: In one of the earliest studies on Bankruptcy Prediction, Paul J. Fitzpatrick presented data for a matched set of 40 firms, 20 that had declared bankruptcy and 20 that did not (“A Comparison of Ratios of Successful Industrial Enterprises with Those of Failed Companies”, Certified Public Accountant, across 3 issues October, November and December of 1932). The articles presented and discussed differences in 13 accounting ratios for the matched firms but did not present any statistical analysis. In 1966, William Beaver provided a statistical analysis by applying t-tests to evaluate the importance of individual accounting ratios within a pair-matched sample (“Financial Ratios as Predictors of Failure”, Journal of Accounting Research, 1966 Volume 4). Beaver found that several ratios provided information regarding the probability of insolvency up to 5 years prior to failure. A t-test is a statistical test, an inferential analytic, that assesses whether the difference of the averages, or means, of two different samples is meaningful. Having an idea of which ratios help predict bankruptcy is a nice start on the traits that predict bankruptcy, but as some firms will have all the traits and others will only have one, we need to be able to consider all the important ratios at the same time. This is where we turn to multivariate regression analysis, another inferential analytic, that allows us to better predict bankruptcy.

The Z-score:

In 1968 Edward Altman published one of the most influential works on bankruptcy prediction putting forth a model that is still used today (“Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy”, Journal of Finance, 1968). The Altman Z-score is a linear combination of five common financial ratios. The coefficients on the ratios and zones of discrimination were determined using a matched sample of 66 public manufacturing firms. Later variations published by Altman extended the model to non-manufacturing and private firms. The original Z-Score assigns the weights to five ratios as follows.

Original Z-score:

Z = 1.2 x WC/TA + 1.4 x RE/TA + 3.3 x EBIT/TA + 0.6 x MVEquity/BVLiab. + 1.0 x Sales/TA

Where:

WC is working capital, TA is total assets, RE is retained earnings, MVEquity is the market value of equity, BVLiab is the book (accounting) value of liabilities, and Sales is total revenue.

Calculating the ratios and applying the weights (the numbers in the equation above) can be interpreted as follows:

• If Z > 2.99 then safe

• If 1.81 < Z < 2.99 then gray zone

• If Z < 1.81 then distress predicted

Non-manufacturing Z-Score:

Z = 6.56 x WC/TA + 3.26 x RE/TA + 6.72 x EBIT/TA + 0.6 x BVEquity/BVLiab.

This version of the Z-score for non-manufacturing firms can be interpreted similarly as:

• If Z > 2.6 then safe

• If 1.1 < Z < 2.6 then gray zone

• If Z < 1.1 then distress predicted

In subsequent research, Accounting Professor Jim Ohlson, improved upon the Altman Z-score (with the Ohlson O-score) by using 9 factors and a sample of 2,000 firms to estimate the coefficients with logit regressions. Logit regressions are a form of regression that seek to explain a binary outcome such as bankrupt or not (i.e., 1=yes, 0=no) given a set of explanatory variables (i.e, the various ratios expected to predict bankruptcy).

While slightly less accurate, the Altman Z-score is easier to understand and provides the same intuition as the Ohlson model, or O-score. In recent years more data intensive methodologies have been used to predict bankruptcy. These methods have included survival rate models, neural networks, and option value implied volatilities. These models often include non-financial statement data such as numbers of negative press releases, payment experiences, etc. (my favorite was “A Rule Based Model for Bankruptcy Prediction Based on an Improved Genetic Ant Colony Algorithm”, Zhang, Wang and Ji, Mathematical Problems in Engineering 2013). These complex models are interesting, and machine learning models can potentially make this prediction better. Our focus, however, will be on the changes in bankruptcy risk over the last few years, with an interest in contemporaneous bankruptcy risk. We will use Altman’s Z-score to determine this risk.

Data and Additional Resources

The following data and resources are available in the case supplement:

1. A dataset that has the raw variables needed to calculate the Z-score called “ZScoreInputs.csv”