Business-Day and Calendar-Day Conventions in Curve Discount Factors

# rateslib curves: Why does changing day count convention affect discount factors?


# rateslib curves: Why does changing day count convention affect discount factors?












I'm using the rateslib Python package to build a yield curve from Brazilian DI1 futures using piecewise/spline interpolation. My expectation was that changing the day count convention of the curve itself (e.g. from 'bus252' to 'act360') wouldn't affect the discount factors much, assuming the instruments and rates used to solve the curve are unchanged.

However, when I change the convention of the curve from 'bus252' to 'act360', I get noticeably different discount factors when querying the curve at the same date.

For example,

```
mixed_curve[dt.datetime(2026, 6, 24)]
```

returns:

0.871460 for convention='bus252' 0.869912 for convention='act360'

Here’s the curve setup code (plot for the code also attached):

```
di1_info = {'N25': 14.90, 'Q25': 14.91, 'U25': 14.93, 'V25': 14.94, 'X25': 14.94, 'Z25': 14.94, 'F26': 14.94, 'G26': 14.94, 'H26': 14.92, 
            'J26': 14.90, 'K26': 14.83, 'M26': 14.80, 'N26': 14.73, 'V26': 14.48, 'F27': 14.21, 'J27': 14.00, 'N27': 13.79, 'V27': 13.61,
            'F28': 13.46, 'J28': 13.37, 'N28': 13.34, 'V28': 13.33, 'F29': 13.32, 'J29': 13.33, 'N29': 13.35, 'V29': 13.35, 'F30': 13.39, 
            'J30': 13.41, 'N30': 13.43, 'F31': 13.48, 'F32': 13.55, 'F33': 13.58, 'F34': 13.58, 'F35': 13.61
            }
...

bra = rateslib.Cal(holidays=[dat.strptime(_, "%Y-%m-%d") for _ in di1_pricing_holidays], week_mask=[5, 6])

mixed_curve = rateslib.curves.Curve(
    nodes= discount_estimate_nodes,
    interpolation="log_linear",
    t = t,
    id="mixed_curve",
    calendar = bra,
    convention='Act360'
)

zcs_args = dict(frequency="A", calendar=bra, curves="mixed_curve", currency="brl", convention="bus252")
instruments = []

for element in di1_intervals:
   date1 = dt.datetime(element[0].year, element[0].month, element[0].day)
   date2 = dt.datetime(element[1].year, element[1].month, element[1].day)
   instruments.append(rateslib.ZCS(date1, date2, **zcs_args))

solver = rateslib.Solver(
    curves=[mixed_curve],
    instruments=instruments,
    s=di1_rates,
    algorithm  = "levenberg_marquardt"
)
```

My question: Why does changing the convention parameter of the curve object change the resulting discount factors, even when all instrument inputs and solver settings are unchanged?

## Answer by Dimitri Vulis (score 2)

https://quant.stackexchange.com/a/83702

besides the interpolation, Brazilian BUS/252 behaves differently from other daycounts, that accrue on calendar days, in that: the numerator of the period fraction has at least $\frac{2}{7}\approx.29$ fewer days, because Saturdays and Sundays are not accruing. There are also many Brazil holidays that sometimes cluster. For example, the Carnival is Monday through (Ash) Wednesday; and both Good Friday and Easter Monday are not accruing.

The denominator is $\frac{365-252}{365}\approx.31$ which doesn't exactly offset the difference in days of accrual. The fraction for a short period that also includes Brazil holidays can turn out to be surprisingly different from 30/360 or Actual/365.