The links to the papers are actually found here:
I would also like to point out a higher level issue with science in general. Gun violence is a highly emotional and political issue. Scientists are not immune to that pressure any more than most people. A epidemiologic study of homicide research found " Most studies and reports surveyed failed in at least one way to observe basic standards of the science of epidemiology."
Ok, not that having started lets do a critique of these articles in turn.
Firearm availability and homicide: A review of the literature
First, from the abstract: "No longitudinal cohort study seems to have investigated the association between a gun in the home and homicide."
So from the outset this study admits to being a snapshot rather than a trend. Its use of studies relying on international and even interstate data without the appropriate controls provided by time automatically removes the author's ability to imply anything like a causation route to this finding, as noted both in the abstract and the findings page. More on this in the critique of study 2, but the comparison of Chicago homicide rates to San Francisco homicide rates in 2001 gives you virtually no actual information. Chicago might have higher gun ownership rates and homicide rates, but they might also have had higher homicide rates when they had lower gun ownership rates, without the appropriate data its fallacious to assume a relationship.
A more extensive study done by the National Academies Press http://www.nap.edu/catalog.php?record_id=10881#toc, found no relationship between those two variables when longitudinal studies were included. They also found an inverse relationship between gun ownership and other crimes. Finally, when they restricted gun ownership to legal gun ownership they found an inverse relationship (not a causal, a correlation) between legal gun ownership and homicide.
Firearm Availability and Homicide Rates across 26 High-Income Countries
This study also lacks a time series component sufficiently long enough to account for factors other than gun ownership to ascribe a relationship with.
To again look at the abstract of the study we see shockingly bad methodology:
Methods: Homicide rates for the early 1990s come from 26 of 27 of the highly industrialized or high-income countries with greater than 1 million population as classified by the World Bank. Two common proxies for gun availability are used, the percentage of suicides with a firearm, and theCook index, the average of the percentage of suicides with a firearm and the percentage of homicides with a firearm.
Results: In simple regressions (no control variables) across 26 high-income nations, there is a strong and statistically significant association between gun availability and homicide rates.
First, they use two proxies to identify firearm ownership rates that are correlated, which we should remember from stats is a no no.
Second, they use poor proxies, use of a firearm for suicide is extremely correlated to cultural backgrounds, sex and age to name a few. At a minimum those variables would need to be controlled for in order to come up with a basic proxy, but the study does no such thing. A good example of this included in the study is Switzerland, which has a low suicide rate in general, a lower use of suicide by firearm than the average used in the study, but actually has higher gun ownership rates
Third, they use a "simple regressions (no control variables)" meaning that no other factors were considered, at all. None. They attempt no common cause control, no other correlation factors, no explanatory model, no population dynamic control, nothing. This is below the requirements for the junior taking a statistics class at the University of Washington. http://faculty.washington.edu/cadolph/?page=42 This study would fail topics 10, 11 and 12 of that class, and they authors would have received no credit for problem set 5 and probably failed the final.
Fourth, their time horizon prevents any substantive use of underlying variables (even if they weren't excluded intentionally) behind the data. Most developed countries have had relatively stable gun laws throughout the 90s. A good example would be that they found a lower homicide rate in London than New York, with higher firearm ownership in the latter than the former. The problem with that analysis is two fold. Not only does the magnitude of difference in ownership rates not account for the homicide level differences, but historically London has always had a lower homicide rate than New York, even when New York had strict gun control laws and London had lax laws and higher ownership rates.
Rates of Household Firearm Ownership and Homicide Across US Regions and States, 1988–1997
I would like to point out an interesting sentence in the report: "although other studies found no association with homicide perpetration." So while they find a connection between ownership and homicide rates, they note that studies have found that they firearms aren't committing the crimes in question. That is a glaring common cause finding, that argues that homicide rates have nothing to do with a firearm since those homicides aren't being committed by firearms. The study the authors cite point out that most studies of the nature of the type Rod pointed out use homicides as reported by Police, not by Courts, meaning that they include a high number of cases where the crime is dismissed as self-defense. The study also found that gun ownership tended to reduce the likelihood of violence in a criminal encounter, but increased the proportion of fatalities in situations where there was violence, which also paints a very different picture from the one painted by the Harvard authors.
Finally, this study also has an extremely constricted time horizon, 1988-1997. In the caveats section the authors noted that gun ownership rates did not change in noticeable rates during this time (less than three percent concentrated in two census region) so the regression association found was entirely between regions (implying a regional cultural differences) rather than tied to gun ownership rates directly.
State-level homicide victimization rates in the US in relation to survey measures of household firearm ownership, 2001-2003.
Unlike the study above, this study did control for many factors and tied (at a state, not an individual level like the Kleck-Hogan paper) increased firearms to increased firearm homicide rates.
I will point out that the time series used is even more restrictive than the one before (gun ownership changed by an even smaller rate and was even more concentrated). Since I have discussed the time series limitation on several occasions I will simply say that they apply here as well. By measuring different people at the same time you are searching for differences in the person, not differences in the gun rate.
More importantly, the study notes in its caveats section that if cases in which the homicide was declared self defense are removed the statistical significance disappears. That means that the entire significance of their finding revolves around cases where a gun owner used the gun to defend themselves.
So in the end the conclusion of the studies is really reduced to two findings. (1)There are significant cultural differences that underlie both gun ownership and homicide rates in regions. (2) In areas (within the US) where there are higher gun ownership rates, there is an increased use of those firearms in self-defense scenarios.