I have an implied vol discrete grid, obtained from market data. To obtain prices from these implied vols, a dividend model with discrete and proportional dividends is used.
How can I verify if there are calendar arbitrages in this implied vol grid?
This paper (Gatheral, Jacquier. Arbitrage free SVI volatility surfaces) says, in Lemma 2.1 on page 3, that if dividends are proportional only, then verifying that the total variance is increasing is a necessary and sufficient condition of no-arbitrage.
Though, in my case dividends are not only proportional, there are also cash dividends.
On page 4 of the same paper, definition 2.2 (without proof) says that the increasing total variance is a sufficient condition of non-arbitrage, without specifying any assumption on dividends.
I want to know if I interpreted correctly this definition 2.2, i.e. if there are both cash and proportional dividends, then
- If I demonstrate that total variance is increasing, then I am sure that there is no arbitrage in the vol grid.
- If there are some points where total variance is not increasing, I am not sure that if there is or if there is not arbitrage on that point, i.e. the increasing total variance condition is sufficient for non-arbitrage, but not necessary.
- Mathematically: if dividends are both discrete with cash and proportional, then $\frac{\partial}{\partial T} \left[ \sigma^{2}(k, T)T \right] >0\Rightarrow \text{No-arbitrage}$
Are these 3 assertions correct?