I came across this graph:
Now, this graph plots the percentage change with respect to the year 1994 for all subsequent years. I wanted to know whether we can come up with a graph where we plot percentage change with respect to ONLY THE PREVIOUS YEAR, that is, for 1996, we plot change w.r.t 1995, for 1997, we plot change w.r.t 1996, and so on, and also if such a graph has seen previous use.
You can plot anything you want that you can calculate unambiguously. Here the number of offenses known in $1995$ was $95.7\%$ of the number in $1994$ and the number in $1996$ was $91.6\%$ of the number in $1994$. The number in $1996$ was then $\frac {0.914}{0.957} \approx 0.955=95.5\%$ of the number in $1955$ You could plot these as change from the previous year. Whether that is a good idea depends on what you want to show or what idea you want to push. If you did this with this data you would show a decrease of a few percent per year without big changes from one year to the next. It would not give people the idea of the cumulative effect of a few percent per year for five years, which is a $20\%$ or so reduction. People would be much more prone to focus on the fact that the reduction $1997-1998$ was larger than the reduction $1996-1997$. Whether that is good or bad is not a mathematical question.