The easiest valid way is the interquartile range of the percentage change which is then normalized for other securities, a process known as shrinkage. Technically this isn't correct, but it is close for the use most people would put it to. It is valid provided proper inference is performed.
Properly, you would solve for the distribution of the volatility parameter given the data as there is no "sufficient statistic," and perform inference on the distribution resulting from the information and any prior knowledge. The distribution is often called the "Posterior distribution." However as this is a bit of a difficult concept, the interquartile range will do for most applications.
What way is there to measure the volatility of a stock or a share, except for percentage change?
