add_fpsim() allows you to add a variety of dyadic foreign policy similarity measures to your (dyad-year, leader-dyad-year) data frame

## Usage

add_fpsim(data, keep)

## Arguments

data

a data frame with appropriate peacesciencer attributes

keep

an optional parameter, specified as a character vector, about what dyadic foreign policy similarity measure(s) the user wants returned from this function. If keep is not specified, the function returns all 14 dyadic foreign policy similarity measures calculated by Haege (2011). Otherwise, the function subsets the underlying data to just what the user wants and merges in that.

## Value

add_fpsim() takes a (dyad-year, leader-dyad-year) data frame and adds information about the dyadic foreign policy similarity, based on several measures calculated and offered by Frank Haege.

## Details

For the dyad-year (and leader-dyad-year) data, there must be some kind of information loss in order to reduce the disk space data like these command. In this case, all calculations are rounded to three decimal spots. I do not think this to be terribly problematic, though I admit I do not like it. If this is a problem for your research question (though I can't imagine it would be), you may want to consider not using this function for dyad-year or leader-dyad-year data.

Haege's (2011) article reads at first glance as agnostic about which of these particular measures you should consider a "preferred" or "default" measure of dyadic foreign policy similarity. Indeed, the 2011 publication in Political Analysis mostly drives the point home that S has important limitations and the multiple variants Haege calculates are not substitutable. This means a user interested in measuring dyadic foreign policy similarity might have to cycle through all of them to assess their varying effects whereas a user interested in this as just a control variable for the model can (probably) get by with picking just one and not belaboring the measure any further.

### Suggested Defaults

An evaluation of the data, the article, and an email exchange with the author leads to the following points the user should consider. What follows is a rationale for why users should think of kappa as a default measure for dyadic foreign policy similarity, though why the "valued" equivalent for the alliance data is an inadvisable default. The example at the end of the document offers the operational "nudge" for what the user should want from this function.

• The choice of measure will in part depend on the temporal domain. If the user has just a post-WWII sample, the UN voting measures offer better coverage. We're all partial to the alliance data, though, because of its 19th century coverage.

• Haege implores the use of chance-corrected measures, like Cohen's (1960) kappa or Scott's (1955) pi. Of the two, Haege suggests kappa over pi. The rationale is the user would need to build in a very strong assumption that the baseline propensity of forming a tie in the dyad is the same for both members of the dyad to make Scott's (1955) pi as appropriate an estimate as Cohen's (1960) kappa even as both have the important chance correction.

• The choice of squared versus absolute distances is arbitrary. Users probably do not think about the differences, or know about the differences. S was usually calculated with absolute differences in software packages, though this was never usually belabored to the user. Comparability with S might be an argument in favor of absolute distance as a default, but keep in mind that squared distances are much more commonly used in most other types of distance and association metrics.

• The choice of binary or valued is also a design choice for the user to consider on the full merits, though the practice of valuing alliance ties on a quantitative scale builds in strong assumptions about the scale of alliance strength as presented in something like the Correlates of War or ATOP typology. S has traditionally done this by default, which is another reason its application in a lot of quantitative peace science research is suspect.

## References

### The Main Source of the Data

For any use of these data whatsoever (except for Tau-b), please cite Haege (2011). Data are version 2.0.

• Haege, Frank M. 2011. "Choice or Circumstance? Adjusting Measures of Foreign Policy Similarity for Chance Agreement." Political Analysis 19(3): 287-305.

Tau-b is calculated by me and not Haege, and no additional citation (beyond citing the package) is necessary.

### Citations for the Particular Similarity Measure You Choose

Additional citations depend on what particular measure of similarity you're using, whether Kendall's (1938) Tau-b, Signorino and Ritter's (1999) S, Cohen's (1960) kappa and Scott's (1955) pi. Haege (2011) is part of a chorus arguing against the use of S, though S measures are included in these data if you elect to ignore the chorus and use this measure. Likewise, Tau-b is in here, though it is not a good measure of dyadic foreign policy similarity for reasons that Signorino and Ritter (1999) mention. Haege (2011) argues for a chance-corrected measure of dyadic foreign policy similarity, either Cohen's (1960) kappa or Scott's (1955) pi.

• Cohen, Jacob. 1960. "A Coefficient of Agreement for Nominal Scales." Educational and Psychological Measurement 20(1): 37-46.

• Kendall, M.G. 1938. "A New Measure of Rank Correlation." Biometrika 30(1/2): 81--93.

• Scott, William A. 1955. "Reliability of Content Analysis: The Case of Nominal Scale Coding." Public Opinion Quarterly 19(3): 321--5.

• Signorino, Curtis S. and Jeffrey M. Ritter. "Tau-b or Not Tau-B: Measuring the Similarity of Foreign Policy Positions." 43(1): 115--44.

### Citations for the Underlying Data Informing the Similarity Measure

Haege (2011) also suggests you cite the underlying data informing the similarity measure, whether it is UN voting or alliances. In his case, he recommended a Voeten citation from 2013 and the alliance data proper. In the case of the alliances, I know Gibler's (2009) book is recommended even if the alliance data have since been updated (and reflected in this measure). In the UN voting data, my understanding is the 2017 paper in Journal of Conflict Resolution is also the preferred citation.

• Bailey, Michael A., Anton Strezhnev, and Erik Voeten. 2017. "Estimating the Dynamic State Preferences from United Nations Voting Data." Journal of Conflict Resolution 61(2): 430--456.

• Gibler, Douglas M. 2009. International Military Alliances, 1648-2008. Washington DC: CQ Press.

Steven V. Miller

## Examples

if (FALSE) {
# just call library(tidyverse) at the top of the your script.
library(magrittr)
# The function below works, but depends on
# running download_extdata() beforehand.
# kappaba: kappa for binary alliance data if you have pre-WWII data.
# kappavv: kappa for UN voting data if you just post-WWII data.