Corporate Restructuring and Turnaround completo

L

correlation.

In conclusion if we want to estimate the rating, then the PD of a company and use only one

!"#$

indicator we have to know that the most reliable one is . However, we know

#%$&'&($ &*+&%(&(

that we cannot compute a PD with just one indicator, because of the scheme above. So

when we look at Modigliani miller or Hamada transformation we are talking about very

simplified models, loss in accuracy.

EXECISE WITH FINANCIAL MAPPING APPROACH: (tables in slides)

We have credit classes, for each of them we have the average level of each indicator, then we

have a company and for this company for each indicator we have a value. Then we have to read

the value in the table and just to say, if I apply only 1 indicator, what is the rating. For each

indicator we have a default frequency. Then I perform an average simple or weighted to obtain

the final rate. Benefit is that it is very simple method but not to much accurate. Not so good to

take decision to take decision in a credit committee of a bank

2) Credit model

Estimate a credit model is a more reliable approach. In order to estimate it we have to perform 5

steps at least:

- Sampling: if we don’t have enough data we cannot have a functioning model. Imagine

there are two databases, good companies and bad companies (wrt backward analysis). We

have to work with a well-balanced sample composed with same number of good and bad

companies. In this way we can more easly understand which is the best indicator and

which is the accuracy ratio of these indicators. On average then PD will be hence 50% (not

reflect real world, hence we need calibration, performed in last step).

- Univariate analysis: define what is the accuracy ratio of each single indicator and what are

the best indicators. I have 1000 companies, 500 good and 500 bad. Eg, we want to say if

ROE is a good predictor of default. So I take 1000 companies, divide those companies by

ROE, make a ranking, and than I divide these companies in 10 deciles and I plot the average

ROI in each decile and the default frequencies in each decile. So plot the account indicator

with default frequencies. If I come out that ROI is a good indicator, because the shape of

the plot has a monotonic trend (slides). We can see on the slides a lot of indicators and

their explanatory power. Debt service coverage ratio is the most used

- Multivariate analysis: then we select the best indicators try to put them together. It is very

important is this phase that there is not too high correlation between indicators we

choose, statistical problem of multi regression approach. Putting these indicators together

we can obtain several models, with weights for each variable. We can use analytical

approach or put data into machine learning approach and the software will give us the

result.

- Best model selection

- Calibration: Since we assumed a sample with PD of 50%, and this is not a real value have to

calibrate this PD. So we have the model outcome divided by 50% multiplied by expected

default probability (we obtain a looking forward default probability) (in the past it was

used average default frequency, backward looking).

Who knows how much is looking forward default probability? Nobody, so we have tools to make

a judgment. Take info of equity market, credit market, default frequencies coming from different

sources, and put together all information try to understand what is the relation between

macroeconomic variables and default frequencies.

It results several models (choose the weight to assign). In machine language (software) you

provide a strong model to separate bad companies from good companies, studying all possible

variables. The price you pay with software is that all the process is a black box, you have only the

final result from the input. With the traditional approach I can see the algebra calculations. Issue

of regulators are trying to solve is that of Black box. At least this approach works, test it with

backtest. In backtest we have the rating classes, the default probability and then we can say what

is the expected default probability before the test (see the curve) and then we compare results ex

post.

Public rating definition procedure takes 3/6 months of time, due diligence, analysis, and have a

cost to have a public rating for the first time. Which is the accuracy? Around 95% on external

evidences. Then if we want to have a quantitative approach you lose accuracy. The combination

between qualitative and quantitative approach can improve the the accuracy, but take 1-2 weeks

time. But if we have a peer-to-peer lending platform you can apply only an auto approach, lot of

numbers.

Very simplified model to obtain immediately a default probability, based on some variables.

Excel example: The real valuable data is the database (sample).

Excel example 2: pdf document, the research relies for CF based valuation

WEEK 4 – CR & Turnaround

Terminate PD estimation for companies and some Financial tools in order to understand if crisis is

coming.

Just to remember, there are some differences in default definition if we refer to:

- fixed income market/rating agencies: a missed cupon/principle payment, bankruptcy

procedure, restructuring of the debt (public information).

- lending market according to Basel 3: 90 days pass due on commercial credit lines,

overdraft and loans. After default we have NPL (private information, deriving by CCR).

We known how estimate PD starting by just two accounting indicators (excel). Now we move to

Moody’s investor services – hermes example (PDF); according to this methodology we have

three kind of items to be analyzed:

- Sector profile( weight 12,5%): composed by sector volatility (7,5%), easy to understand,

more volatility means more risk for the entire industry where the company operates.

Sector outlook (5%), is the forecasting about sector volatility. We can say from the weights

that history is more important than future previsions. We can see the sector profile

judgments in the table, from AA to CCC, with the description. We choose the best

description of the industry in a way that we can apply our rating with specific outlier.

- Business profile (17,5%): total weight split between competitive position (10%), and

concentration risk (7,5%). And then there are the descriptions of the item corresponding to

the different ratings. So understand in which rating columns these items are better

described.

- Financial profile (70%): based on financial accounting metrics, we have 5 dimensions to

take in consideration. The first is the Size, means revenues in terms of millions of euros;

the bigger the size of the company the lower the default frequency. Profitability, measured

by two indicators, ROCE (return on capital employed), means EBIT/(NFD+Equity). Then

there is the leverage, to evaluate it we have two kind of indicators, the first one is

,#-"#.#$#&( 456

and the second . Then Capitalization, two indicators, equity ratio and

!/0123 ,#-"#.#$#&( !/0123

leverage ratio. Finally we have the Coverage ratio, means . If we want to

0%$&'&($ &*+&%(&(

construct a model based on only one variable the best one to use is the latter. According to

the level of this indicators, looking at the table, it corresponds a score. According to this

approach the last indicator, the powerful, has only 5% weight. The Leverage have the

higher weight, we have to keep in mind the reference sector, because every sector has

references value for this index (utilities or PA makes sense a lev of 6,5x but in

manufacturing, textile the maximum acceptable level is between 5-6). To have info about

industry risk we can go to bank of Italy website and look for “Tassi di decadimento”. Most

dangerous sector is real Estate (Astaldi situation).

So we have 70% financial profile and 30% sector and business profile. When we arrive to this

estimation the final step of adjustment, basing on the subjectivity of the analyst, and you can

correct the final result between a range +1% – 3% according your feeling to the credit valuation.

We can try to apply this methodology to whatever company we want.

rd

3 tool: rating model provided by Fondo Centrale di Garanzia (with CCR)

In this model we take in account the CCR importance in assess PD of a company. In CCR we can see

the committed credit lines and the usage of credit lines. From that we can compute the:

- Credit reserve = committed credit lines – credit usage (drawn credit lines), key indicator.

Drawn credit line is the net debt in balance sheet. But it is important to know the

committed credit lines or credit reserves of company, see how much money can have

company coming from the banking system.

- Overdraft: drawn credit line > committed credit line. Overdraft persistent for more than 90

days situation of default.

This info provided by CCR is totally reliable. Information in CCR are known by all banks, if more

than 90 days past due (default) banks will immediately cancel accorded credit lines. At this point a

the only way for a company is look for any asset to monetize and repay the banking system.

Liquidity contingent plan: when study a business plan important to ask in evaluate to give credit

or not to a company. Maybe build a liquidity reserve. Treasury flexibility valuation is the key part

of rating assessment in this model. We don’t read it from accounting metrics, but only having

access to CCR. To avoid liquidity shortfall, ask for new incremental credit lines or asset sale.

Remember the credit channel asymmetries, if you work as a bank you have access to CCR, so you

can understend if there are some past due situations. However if you work only on the bond credit

channel you don’t have access to CCR, so you have to protect yourself inserting in the contract

accounting covenants and you work with delayed information (BS is a delayed info). Huge

difference between the two markets.

If we give a look to FCG rating model we have two legs:

- Accounting module: very simple, there are items coming from the BS and IS identified by a

2&"#$# - "'&7&

code. Then there is the definition of variables used in the model, like and we

5-$$8'-$9

can clearly understand what are the accounting metrics underlying this approach. You can

evaluate not only limited companies, but also SME. Once defined the variables there is the

formula to calculate the score:

Constant factor plus sum of multiplication between variable and its coefficient. See

variables and coefficients in very transparent way, there is the list of variables used in the

scoring, the values of constant and coefficient. According to the final score we define a

class of rating.

- CCR module or behavioral data: same methodology above it is applied to CCR data. This

data express the “committed credit lines” (accordato) and “drawn credit lines” (utilizzato).

“Rischi a scadenza” means “term loan” (is the facility of rischi a scadenza). Then we have

also the NPLs &