Discretionary Accruals
Calculate the discretionary accruals of the modified Jones model, the part of a company’s accruals that its change in revenues and its fixed assets do not explain and that is therefore attributed to management’s discretion. Jones (1991) modelled the accruals a company needs as a function of its change in revenues (working capital moves with sales) and its property, plant and equipment (depreciation). Dechow, Sloan and Sweeney (1995) subtracted the change in receivables from the change in revenues, because revenue booked on credit can itself be managed, and found this modified model the most powerful at detecting earnings management. Large positive discretionary accruals point at earnings that were managed up.
The formula is as follows:
\[\text{Total Accruals} = \frac{\text{Net Income} - \text{Cash Flow from Operations}}{\text{Total Assets}(t-1)}\] \[\text{Estimation:}\;\; \text{Total Accruals} = \frac{a_{1} \cdot 1}{A(t-1)} + \frac{a_{2} \cdot \Delta \;\text{Revenue}}{A(t-1)} + \frac{a_{3} \cdot \text{PPE}}{A(t-1)} + e\] \[\text{Non-Discretionary Accruals} = \frac{a_{1} \cdot 1}{A(t-1)} + \frac{a_{2} \cdot (\Delta \;\text{Revenue} - \Delta \;\text{Receivables})}{A(t-1)} + \frac{a_{3} \cdot \text{PPE}}{A(t-1)}\] \[\text{Discretionary Accruals} = \text{Total Accruals} - \text{Non-Discretionary Accruals}\]Here A(t-1) is the total assets at the start of the year, ΔRevenue and ΔReceivables the changes over the year and PPE the property, plant and equipment at the end of it. The coefficients a1, a2 and a3 are estimated with ordinary least squares without an intercept, as in Jones (1991); with intercept=True a constant a0 is added to the estimation and the Non-Discretionary Accruals (Kothari, Leone and Wasley, 2005).
For more information, see the following papers:
- Jones, J.J. (1991). “Earnings Management During Import Relief Investigations.” Journal of Accounting Research, 29(2), 193-228.
- Dechow, P.M., Sloan, R.G. and Sweeney, A.P. (1995). “Detecting Earnings Management.” The Accounting Review, 70(2), 193-225.
- Kothari, S.P., Leone, A.J. and Wasley, C.E. (2005). “Performance Matched Discretionary Accrual Measures.” Journal of Accounting and Economics, 39(1), 163-197.
Also known as: modified Jones model, abnormal accruals, Jones model, earnings management.
Discretionary Accruals in Python or via the MCP
Install the Finance Toolkit with:
pip install financetoolkit -U
Then call get_discretionary_accruals as shown below.
from financetoolkit import Toolkit
toolkit = Toolkit(
["AAPL", "MSFT"],
api_key="FINANCIAL_MODELING_PREP_KEY",
start_date="2005-01-01",
end_date="2025-12-31",
)
toolkit.models.get_discretionary_accruals().loc[:, "2021":]
Or, after setting up the Finance Toolkit MCP, ask:
What were the total, non-discretionary and discretionary accruals of Apple and Microsoft from 2021 to 2025, according to the modified Jones model estimated from 2005?
Which returns:
| Item | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| (‘AAPL’, ‘Total Accruals’) | -0.0289 | -0.0637 | -0.0384 | -0.0695 | 0.0014 |
| (‘AAPL’, ‘Non-Discretionary Accruals’) | -0.1208 | -0.084 | -0.0696 | -0.0707 | -0.0794 |
| (‘AAPL’, ‘Discretionary Accruals’) | 0.0919 | 0.0203 | 0.0312 | 0.0012 | 0.0808 |
| (‘MSFT’, ‘Total Accruals’) | -0.0513 | -0.0488 | -0.0417 | -0.0738 | -0.067 |
| (‘MSFT’, ‘Non-Discretionary Accruals’) | -0.1076 | -0.0973 | -0.1047 | -0.0961 | -0.1061 |
| (‘MSFT’, ‘Discretionary Accruals’) | 0.0563 | 0.0485 | 0.063 | 0.0223 | 0.039 |
Parameters
get_discretionary_accruals accepts the following parameters.
| Parameter | Type | Description | Example |
|---|---|---|---|
method | str | How the coefficients are estimated. “time_series” estimates them for every year on the company’s own earlier years, as Jones (1991) and Dechow, Sloan and Sweeney (1995) did, with the change in revenues in the estimation and the change in receivables subtracted only for the year itself. “cross_section” estimates them per year across the tickers of the Toolkit, with the change in revenues minus the change in receivables in the estimation as well, which is meaningful with many companies of one industry. Defaults to “time_series”. | "time_series" |
min_observations | int | The fewest observations a regression is estimated on: earlier years of the company with “time_series”, tickers with “cross_section”. At least 3, the number of coefficients. Defaults to 10, the minimum estimation period of Dechow, Sloan and Sweeney (1995). | 10, the minimum estimation period of Dechow, Sloan and Sweeney (1995) |
intercept | bool | Whether to add a constant to the regression, as Kothari, Leone and Wasley (2005) do, in both methods. It removes the average level of the accruals from the other coefficients at the cost of a degree of freedom (one observation more is needed). Defaults to False, the formulation of Jones (1991) and Dechow, Sloan and Sweeney (1995). | False, the formulation of Jones (1991) and Dechow, Sloan and Sweeney (1995) |
trailing | int | None | Sum the flows over this many periods (e.g. 4 quarters, a trailing twelve months) and compare with the values trailing periods back, so that with quarterly data every change is year over year. Balance sheet items are taken at the period end. The window counts calendar periods, so a period without a report makes it NaN. Defaults to None, which is 4 for quarterly data (a year) and 1 otherwise; pass 1 for quarterly data to compare quarter over quarter. | 4 |
rounding | int | None | The number of decimals to round the results to. Defaults to None, which uses the Toolkit’s rounding. | 2 |
show_columns | str | list[str] | None | The columns to keep, a name or a list of names, matched on every column level, or the rows when no column has those names (such as the line items of a financial statement). Names that do not exist are logged, and when none exist the result is returned in full. Defaults to None, which keeps every column. | None, which keeps every column |