If I want to fit the Nelson-Siegel-Svensson (NSS) model to a set of spot, forward, or discount rates, my intuition says that the data should of course be in percentage form.
For example, I should use $r = 0.05$, instead of $r = 5$, as they are often quoted.
Is this correct?
Additionally, why does Svensson divide the spot rate by 100 when computing the discount factors, if the spot rate formula is already in percentage form ?
$$ d(m;b) = \exp\left( - \frac{i(m;b)}{100} m \right) $$
where:
- $i(m;b)$ is the spot rate.
- $m = T - t$ is the time to maturity.
- $b = [\beta_0, ... , \beta_3, \tau_1, \tau_2]$ is the NSS parameter vector.