Airflow scales roughly according to the square root of the pressure difference across an opening (from Bernoulli’s principle), so you can’t treat it linearly.
For infiltration calculations, the 4 Pa assumption comes from the Sherman and Grimsrud method, which takes a series of airflow measurements from building depressurization tests at high (i.e 75) pressure and fits the data to predict airflow at 4 Pa (typical building envelope pressure).
The formula for infiltration they use is:
\dot{q}_{inf} = C \Delta{P}^n
where:
P = Pressure [Pa]
\dot{q}_{inf} = airflow per exterior facade area [\frac{m^3}{m^2 s}]
C and n are coefficients for the regression, and are also unknown. You’ll have to make an assumption for one of them. A value of 0.65 for n is typically used. With that you can solve for the C value used in the 75 Pa test and use that value to calculate the 4 Pa airflow, like this:
\begin{align} C_{4_{PA}} &= \frac{\dot{q}_{inf}}{\Delta{P}^n} \\ &= \frac{0.0203}{75^{0.65}} \\ &= 0.0012266 \end{align}
Using that I get:
\begin{align} \dot{q}_{inf} &= 0.0012266 \cdot 4^{0.65} \\ &= 0.00302 \text{ } \frac{m^3}{m^2 s} \end{align}
The one last thing I’ll add is that it’s a little nicer to take the log of all these values first, and do your derivations that way, because the log scales all the values to similar precision and for that reason preserves accuracy a little better. In the non-log version you end up mixing high and low values and that causes some precision loss. But that’s a detail and the above method will work fine.
EDIT
I double checked your initial conversion because 0.00302 seemed too leaky. I’m getting 0.4 cfm/ft2 = 0.002032 m3/m2-s from google, if we substitute that the airflow rate is 0.000302 which seems more reasonable.