Diffusion of Bilateral Investment Treaties 829 purposes of sending a proinvestment signal to international markets.47 Both UNCTAD and the World Bank's ICSID track the date and signatories of BITs. While the two sources basically agree,UNCTAD's list is more recent and more comprehensive.48 As the equation above indicates,our independent variables take on one of four analytic forms:(1)independent factors associated with the "home" country;(2)independent factors associated with the "host"country;(3)factors associated with the relationship between host and home countries;and(4)spatial lags of the dependent variable. Spatial lags of the dependent variable:Competition and cultural emula- tion.To assess the source and strength of the various influences of policy dif- fusion we construct a series of spatial lags,modeled largely after those in Sim- mons and Elkins.49 Spatial lag models treat spatial dependence in the same way time-series models treat serial correlation.3o Instead of(or,in our case,in addition to)lagging the value of the dependent variable one unit in time,one"lags"it one (or more)units in space to capture the behavior of neighbors.Thus,the general formulation of the spatial lag above,Wy,in which W is an N by N by T spatial weights matrix that maps the distances between units for each year and y*is an N by T matrix of values that represent some function of the dependent variable for all countries other than i.In our case,y*represents the sum total of BITs in force for each host country in a given year.s!We then compute the weighted average of y by dividing the sum of its product with W,by a row sum of W.Thus,the spa- tial lag for country i using the weight w is as follows, wgy+wk+…+wny后 wy+wk十·十wm This computation of Wy*allows an intuitive interpretation of the spatial lag:it is the average number of BITs in force among other host countries,weighted by some distance to country i.Since the spatial lag is endogenous,we lag it one year 47."As the great majority of BITs are ratified,it is reasonable to assume that,in the perception of investors,signing a BIT is the crucial action:Once a BIT is signed,or expected to be signed,the market has absorbed it or begins to absorb it";UNCTAD 1998,106. 48.Our comparison of the two data sets found that,for the years they overlapped (1959-97). UNCTAD included over two hundred treaties not included in the ICSID database. 49.Simmons and Elkins 2004. 50.Anselin 1988. 51.This formulation assumes that it is the accumulation of treaties among peers,not the"event"of their recent signing,that provokes a response.We also experimented with an event trigger by calcu- lating the number of BIIs signed by others in the previous year.In fact,the results were largely comparable.purposes of sending a proinvestment signal to international markets+ 47 Both UNCTAD and the World Bank’s ICSID track the date and signatories of BITs+ While the two sources basically agree, UNCTAD’s list is more recent and more comprehensive+ 48 As the equation above indicates, our independent variables take on one of four analytic forms: ~1! independent factors associated with the “home” country; ~2! independent factors associated with the “host” country; ~3! factors associated with the relationship between host and home countries; and ~4! spatial lags of the dependent variable+ Spatial lags of the dependent variable: Competition and cultural emula￾tion. To assess the source and strength of the various influences of policy dif￾fusion we construct a series of spatial lags, modeled largely after those in Sim￾mons and Elkins+ 49 Spatial lag models treat spatial dependence in the same way time-series models treat serial correlation+ 50 Instead of ~or, in our case, in addition to! lagging the value of the dependent variable one unit in time, one “lags” it one ~or more! units in space to capture the behavior of neighbors+ Thus, the general formulation of the spatial lag above, Wy *, in which W is an N by N by T spatial weights matrix that maps the distances between units for each year and y * is an N by T matrix of values that represent some function of the dependent variable for all countries other than i+ In our case, y * represents the sum total of BITs in force for each host country in a given year+ 51 We then compute the weighted average of y * by dividing the sum of its product with W, by a row sum of W+ Thus, the spa￾tial lag for country i using the weight w is as follows,
wij yj * wik yk * {{{ win yn * wij wik {{{ win This computation of Wy * allows an intuitive interpretation of the spatial lag: it is the average number of BITs in force among other host countries, weighted by some distance to country i+ Since the spatial lag is endogenous, we lag it one year 47+ “As the great majority of BITs are ratified, it is reasonable to assume that, in the perception of investors, signing a BIT is the crucial action: Once a BIT is signed, or expected to be signed, the market has absorbed it or begins to absorb it”; UNCTAD 1998, 106+ 48+ Our comparison of the two data sets found that, for the years they overlapped ~1959–97!, UNCTAD included over two hundred treaties not included in the ICSID database+ 49+ Simmons and Elkins 2004+ 50+ Anselin 1988+ 51+ This formulation assumes that it is the accumulation of treaties among peers, not the “event” of their recent signing, that provokes a response+ We also experimented with an event trigger by calcu￾lating the number of BITs signed by others in the previous year+ In fact, the results were largely comparable+ Diffusion of Bilateral Investment Treaties 829