Kristian Gleditsch's Polity Data Archive for older versions of POLITY dataset. See also Will H. Moore's Polity Data Page.
Variable POLITY2 is widely used by researchers on the cause and effect of democracy. For criticism on this aggregate variable, see Gleditsch and Ward (1997) "Double Take: A Reexamination of Democracy and Autocracy in Modern Polities," Journal of Conflict Resolution, 41(3): 361-383.
Vreeland (2008) proposes a modified POLITY2 variable (called X-Polity) that purges the presence of civil conflict, by arguing that the Polity coding scheme juxtapose civil conflict with democracy.
Variable EXCONST is used as the measure of "limited government", hence secure property rights, by Acemoglu and Johnson (2005). Acemoglu, Johnson, and Robinson (2001) show that this variable in year 1900 or the year of independence is correlated with Political Risk Service's measure of secure property rights and European settlers' mortality (see Table 3 of Acemoglu, Daron, Simon Johnson, and James A. Robinson (2001) "The Colonial Origins of Comparative Development: An Empirical Investigation," American Economic Review, 91(5), pp.1369-1401). Besley and Persson (2011) use this variable to measure the degree to which the group of citizens in power cannot exclude the opposition groups in the allocation of government spending. See Glaeser et al. (2004) for criticism against the use of this variable as a measure of institutions.
Variable PARCOMP is found to be the main driving force for democracy to have higher manufacturing wages (see Rodrik 1999, section IV).
Variable DURABLE is used by Smith (2004) to examine whether oil wealth enhances or undermines regime durability.
If you use variable REGTRANS, be careful. The description in the codebook may be different from how the variable is actually coded in the data file. See page 53 of Jong-A-Pin and de Haan (2008, Econ Journal Watch).