Here's a compendium of studies
showing, again - unsurprisingly -, that whether you look across cities, U.S. states, or countries, more availability of guns leads to more gun violence and more gun homicides.
Statistical data make the pro-gun anecdotes/rhetoric/straw-men look like the jokes they are.
No, your study is junk because they didn't have the courage to take any actual academic risk. They say they controlled for povertty, but poverty is not a good indicator of crime, culture is. Turkey is extremely poor by EU standards but they rate like Germans on being proper and honest. Many rich kids will rob or kill you for jollies.
When you take state-by-state firearm rates versus murder rates, you get almost a shotgun pattern (I can IM anyone my spreadsheet so far). The first pass shows that Montana, Wyoming, the Dakotas, Vermont, and other extremely
high gun-ownership states have a murder rate lower than the UK, but so do some other wildly homogenous white states with low gun ownership rates, like Rhode Island or Massachusetts.
If you try to remove the influence of blacks on the data, the outliers shift to Alaska and the Southwest, and the core data is still a random shotun blast. There is no way to remove cultural influence without breaking things down further into Hispanics and Indians, and then Southern Baptist blacks versus rappers, Italians versus English, Japanese versus Somoans, Hondurans versus Mexicans.
The thing is, even if you take it to those levels, you can do it all day long in a futile attempt to tease gun vs. homicide rates out of noise, and you can create all kinds of patterns, and not one of them is valid on the next run. The gun data is actually noise.
ETA: The Harvard study would've browsed through countless junk plots until they started seeing what they wanted using a set of defensible metrics. They also could've shown you anything else you wanted to see. A previous Harvard study reached the exact opposite conclusions and was cited by the US Supreme Court.
When you're trying to tease tenth of a percent trends out of data (1.5 vs 1.6) and your sample includes hundrds of 1's and an occassional 300, whose behavior depends on what happened on MTV the previous night, it is da** hard to extract meaningful information.