Working Papers

Growth in Publications and Citations Causes Bias in Measures of Scholarly Citation Patterns  

13 December 2019

Lanu Kim, Christopher Adolph, Jevin West, and Katherine Stovel

Scholars have debated whether internet based technologies have led to greater concentration or dispersal of scientific citations. To address the matter empirically, we replicate measures of inequality commonly applied to scholarly citation distributions using Web of Science data covering four broad scientific fields (Health; Humanities; Mathematics and Computational Sciences; and Social Sciences) during the period 1996–2016. We then simulate values of these measures using just the empirical yearly marginal distributions of papers and citations under the assumption that all papers have an equal chance to be cited. This exercise reveals that most raw inequality trends are substantially driven by changes in the relative number of papers and citations. When we standardize observed inequality measures to account for the changing number of publications and citations, we find little evidence that citations are becoming more dispersed across papers; rather, the dominant trend is for citations to become more concentrated on a smaller segment of the prior literature. This result has important implications for the future of scientific innovation.

Reports and Unpublished Papers

Small Sample Properties of Partially-Observed Rank Data Estimators  

This Version: 6 October 2011

Christopher Adolph

Several estimators from the social science toolkit might be used to model the relationship between imprecisely-observed ranks in a hierarchy, and covariates explaining those ranks. But application of standard methods – such as linear regression, ordered probit, or censored regression – is complicated by the interdependence of rank observations. Monte Carlo evidence shows that estimators which either delete partially observed ranks and/or inappropriately assume ranks are iid perform poorly, yielding inefficient and sometimes biased estimates, and wildly inaccurate confidence intervals. In contrast, a Bayesian partial rank model – designed to impute missing ranks within known bounds, and account for interdependence across ranks – performs well, even when most or all ranks are observed imprecisely.

Report on the 2004 Washington Gubernatorial Election  

Submitted May 12, 2005 to the Superior Court of Chelan County, Washington

Christopher Adolph

Expert witness report on ecological inference and the statistical analysis of elections in Borders v. King County, which contested the 2004 Washington gubernatorial election. The final (unappealed) ruling in favor of Governor Christine Gregoire can be found here. Wikipedia's entry on the 2004 Washington election can be found here, though I cannot vouch for the validity of wiki content.

Evidence on Time Series Cross-Section Estimators and Specifications from Monte Carlo Experiments  

March 19, 2005

Christopher Adolph, David M. Butler, and Sven E. Wilson

Political scientists often and increasingly analyze time-series cross-sectional (tscs) data. These data come with significant problems, such as accounting for unobserved variation across sample units and appropriately specifying dynamics. Furthermore, even though fixed-effects (or least squares dummy variable (lsdv) models) can address unit heterogeneity, least squares (ls) estimation of modelds with fixed-effects and lagged dependent variables are known to be biased. Alternative estimators, mostly from economics, and generally designed for short panels, have been proposed to address this bias, but it is generally not well known how these estimators perform in comparison to simple methods like ls and lsdv on tscs data. The preliminary results we illustrate here suggest that lsdv is generally as good or better than instrumental variables (iv) approaches in terms of bias and efficiency. We examine estimator performance under conditions where the importance of the unit effects and the correlation of the unit effects with the independent variables are allowed to vary and find that lsdv performs well. Unfortunately none of the estimators, particularly ls, perform well when the dynamics of the model are mis-specified. The lesson is that new estimators do not, in general, solve the problem of mis-specifying the model’s dynamics.

Homeowner Association Foreclosures and Property Values in Harris County, 1985–2001  

October 12, 2002

Christopher Adolph

In recent years, homeowner associations (hoas) in Harris County, Texas have filed thousands of lawsuits threatening foreclosure against residents who owed dues, late fees, or fines. An event count analysis of hoa foreclosures by neighborhood from 1985—2001 shows the bulk of these filings occur in neighborhoods with low median home values. Overall, homeowners in the bottom quartile of home value face more than ten times the risk of hoa foreclosure proceedings as those in the top quartile. Legal changes in 1987 and 1995 also seem to have encouraged hoas to bring more foreclosures to court: across the spectrum of home values, the annual pace of filing after 1995 is roughly double the previous decade’s rate. Although hoa foreclosures are ostensibly motivated by efforts to improve property values, neither foreclosure activity nor hoas appear linked with above average home price growth.

Replication:  Data are drawn from HOAdata, which should be consulted for more recent data.

Playing Favorites: How Parties Distribute School Finance by Income and Race  

June 21, 2001

Christopher Adolph

A partisan legislative logic influences the distribution of school funding in the American states. Regression analysis and simulation using data on state aid to American school districts from 1992-1997 reveals three aspects of partisan education policy. First, partisan state governments allocate resources towards their core constituents – as grouped by income and race – while matching each other in funding the median voter. Second, students in low income or densely black districts benefit substantially from Democratic control of the state government, but receive a smaller share of the state education budget under Republican regimes. Finally, the party-mediated effect of race on educational resource distribution is even greater than the party-mediated income bias.