ind1<-rep(c("E","W"), times=20)
ind2<-sample(100:150, 40)
y<-c(1:40)
index<-data.frame(cbind(ind1, ind2, y))

x1<-sample(c("E","W","N"), 40, replace=TRUE)
x2<-sample(100:150, 40)
x3<-rep(0, times=40)
data<-data.frame(cbind(x1,x2,x3))

index1<-split(index, index$ind1)
data1<-split(data, data$x1)
data1$E$x3<-match(data1$E$x2, index1$E$ind2)
data1$W$x3<-match(data1$W$x2, index1$W$ind2)

merge(data, index, by.x=c("ind1", "ind2"), by.y=c("x1", "x2"), all.x=TRUE, all.y=FALSE)

new_data <- data

new_data$x3 <- ifelse(
    is.na(match(paste(data$x1, data$x2), paste(index$ind1, index$ind2))),
    0,
    index$y)

require(dplyr)
left_join(data, index, by = c("x1" = "ind1", "x2" = "ind2"))

#Write the row number of index in x3 which matches
data$x3 <- match(interaction(data[c("x1", "x2")]), interaction(index[c("ind1","ind2")]))

#In case you want to return 0 instead of NA for nomatch
data$x3 <- match(interaction(data[c("x1", "x2")]), interaction(index[c("ind1","ind2")]), nomatch=0)

#Instead of >interaction< you could also use paste as already suggested by Dinre
data$x3 <- match(paste(data$x1, data$x2), paste(index$ind1, index$ind2))