Analyzing the Altman Z-Score with Alphalens¶

Let's use Alphalens to explore the forward performance of stocks based on their Altman Z-Score. In previous notebooks we formed equal-sizes quantiles, but in this notebook we'll form bins corresponding to the Altman Z-Scores three "zones":

Distress Zone: Z-Score < 0
Grey Zone: 0 < Z-Score < 3
Safe zone: Z-Score > 3

Defining your own bin edges can be done either in the Alphalens' from_pipeline() function or in the Pipeline definition itself. First, we'll show how to do it in Alphalens, and then we'll show how to do it in Pipeline.

Defining bin edges in Alphalens¶

To define the bin edges in Alphalens, we can start by creating a simple Pipeline that contains the Altman Z-Score factor:

In [1]:

from zipline.pipeline import sharadar, Pipeline
from codeload.fundamental_factors.universe import CommonStocks, BaseUniverse

universe = BaseUniverse() 

altman = sharadar.AltmanZScore('ART', mask=universe)

pipeline = Pipeline(
    columns={
        'altman': altman,
    },
    initial_universe=CommonStocks(),
    screen=universe & altman.notnull()
)

The Alphalens from_pipeline() function contains two, mutually exclusive parameters that control how the factor will be binned: quantiles and bins.

In the previous notebooks, we specified quantiles=n to create n equal-sized bins. Under the hood, this caused Alphalens to pass the factor data to pandas' qcut() function, which calculates the bin edges needed to have the same number of items in each bin.

In this notebook, we will use the bins argument to tell Alphalens where the bin edges should be. Under the hood, this will cause Alphalens to pass the factor data to pandas' cut() function. To split the data into three bins, where the first bin contains everything below 0, the second bin goes from 0 to 3, and the third bin contains everything above 3, we can pass the following bins argument:

bins=[-float("inf"), 0, 3, float("inf")]

Since we want 3 bins, we have to pass 4 bin edges; the length of the list you pass to bins will always be the desired number of bins plus one. The range of the first bin is defined by the first two numbers (here, minus infinity and 0), the range of the second bin is defined by the second and third numbers (0 and 3), and the range of the third bin is defined by the third and fourth numbers (3 and infinity).

In [2]:

import alphalens as al

al.from_pipeline(
    pipeline,
    start_date="1998-02-01",
    end_date="2022-12-30",
    periods=[1, 5, 21],
    factor="altman",
    bins=[-float("inf"), 0, 3, float("inf")],
    segment="Y"
)

Factor Distribution
	min	max	mean	std	count	avg daily count	count %
Altman Quantile
1	-326,034.031	-0.000	-14.856	1,408.404	1,754,284	279.7	9.4%
2	0.000	3.000	1.735	0.777	6,944,924	1107.5	37.2%
3	3.000	264,810.750	14.864	673.918	9,951,467	1586.9	53.4%

Returns Analysis
	1D	21D	5D
Ann. alpha	-0.045	-0.030	-0.027
beta	0.055	-0.023	0.019
Mean Relative Return Top Quantile (bps)	0.107	-0.034	0.056
Mean Relative Return Bottom Quantile (bps)	-1.354	-1.365	-1.476
Mean Spread (bps)	1.461	1.937	1.838

Information Analysis
	1D	21D	5D
IC Mean	0.011	0.018	0.015
IC Std.	0.065	0.082	0.073
Risk-Adjusted IC	0.165	0.216	0.213
t-stat(IC)	13.036	17.082	16.832
p-value(IC)	0.000	0.000	0.000
IC Skew	-0.118	-0.264	-0.271
IC Kurtosis	0.976	0.272	0.949

Turnover Analysis
	1D	5D	21D
Quantile 1 Mean Turnover	0.008	0.037	0.133
Quantile 2 Mean Turnover	0.004	0.019	0.070
Quantile 3 Mean Turnover	0.003	0.013	0.049

	1D	21D	5D
Mean Factor Rank Autocorrelation	0.999	0.978	0.995

Out[2]:

		1D	21D	5D	factor	factor_quantile
date	asset
1998-02-02	Equity(FIBBG000H83MP4 [AAI1])	0.000000	0.000000	0.000000	-1.967196	1
	Equity(FIBBG000B9XRY4 [AAPL])	-0.033861	0.262698	0.010377	1.908786	2
	Equity(FIBBG000B9Y973 [AATI1])	0.016695	0.123716	0.076894	5.829192	3
	Equity(FIBBG000MDCQC2 [COR])	0.031084	0.002296	0.064112	5.316102	3
	Equity(FIBBG000B9YDD7 [ABF])	0.012302	0.068541	0.054482	3.759253	3
...	...	...	...	...	...	...
2022-12-30	Equity(FIBBG019QV0CG8 [SNAL])	-0.032895	0.052632	-0.006579	2.226664	2
	Equity(FIBBG011RWR2Q4 [ACRV])	-0.040000	0.350000	-0.033333	12.978663	3
	Equity(FIBBG000BCVMH9 [CP])	-0.010874	0.046280	0.028113	1.660714	2
	Equity(FIBBG000K5M1S8 [ENB])	-0.004329	0.043290	0.031831	0.774874	2
	Equity(FIBBG000C32XT3 [IMO])	0.004534	0.128607	-0.004328	3.731341	3

18650675 rows × 5 columns

Tear sheet commentary¶

Factor Distribution table¶

min/max: For each factor quantile, we can check the minimum and maximum values to make sure they are as we expected. Here, everything in quantile 1 is below 0, everything in quantile 2 is between 0 and 3, and everything in quantile 3 is above 3.
count %: Because we defined the bin edges instead of using equal-sized quantiles, the quantiles are different sizes. Less than 10% of stocks are in the distress zone, ~40% are in the grey zone, and ~50% are in the safe zone.

Returns Analysis¶

Mean Relative Return By Factor Quantile: companies in distress perform worse than companies not in distress. There is little difference between the safe zone and the grey zone.

Defining bin edges in Pipeline¶