It's based on data from historical criminologists such as Ted Robert Gurr [1], Manuel Eisner [2] [3], and others. Gurr primarily used data from other researchers who examined local homicide and coroner and court records that had been kept for centuries in the archives of various towns and cities in England throughout the last millennium and produced a graph (note the logarithmic y-axis in this graph and all three other graphs in this post) in his 1981 paper. Needless to say, we've learned even more since 34 years ago, and Eisner goes into more detail about more recent findings. Here, for example, is a much more comprehensive set of data points from Eisner's 2003 paper, which includes continuous homicide data from the town of Kent from the 16th Century onward, a dataset compiled by the historian J.S. Cockburn [4].
Eisner's papers also go into depth about other Western European countries' past homicide rates, in case England was simply an outlier. Here's a graph from Eisner's 2003 paper showing homicide rates from Scandinavia, Germany, Switzerland, Netherlands, and Italy.
All of these data are also consistent with the data we have on prehistoric homicide and homicide among contemporary hunter-gatherers, where the homicide rates easily reached an order of magnitude higher than even Medieval Europe. Here's a graph from Pinker's 2011 book showing the homicide rate of Western Europe (as an average of the sources mentioned) along with average nonstate societies' homicide rates, which are based on data from countless anthropologists and other researchers and compiled by Lawrence Keeley [5] and Azar Gatt [6] in their respective books.
Hope this helps.
