Robust likelihood inference for public policy

Issues of public policy are typically decided by non‐specialists who are increasingly informed by statistical methods. In order to be influential, inferential techniques must be widely understood and accepted. This motivates the author to propose likelihood‐based methods that prove relatively insensitive to the choice of underlying distribution because they exploit a peculiarly stable relation between two standard errors and a 95% coverage probability. The author also notes that bootstrap and jackknife estimates of variance can sometimes be strongly biased. In fact, symbolic computations in R suggest that they are reliable only for statistics that are well approximated by averages whose distributions are roughly symmetric. The author thus proposes to transform the classical likelihood ratio into a statistic whose variance can be estimated robustly. He shows that the signed root of the log‐likelihood is well approximated by an average with a roughly symmetric distribution. This leads to Cox‐Tukey intervals for a Student‐like statistic and to simple confidence intervals for most models used in public policy.     

Hybrid and Size‐Corrected Subsampling Methods

This paper considers inference in a broad class of nonregular models. The models considered are nonregular in the sense that standard test statistics have asymptotic distributions that are discontinuous in some parameters. It is shown in Andrews and Guggenberger (2009a) that standard fixed critical value, subsampling, and m out of n bootstrap methods often have incorrect asymptotic size in such models. This paper introduces general methods of constructing tests and confidence intervals that have correct asymptotic size. In particular, we consider a hybrid subsampling/fixed‐critical‐value method and size‐correction methods. The paper discusses two examples in detail. They are (i) confidence intervals in an autoregressive model with a root that may be close to unity and conditional heteroskedasticity of unknown form and (ii) tests and confidence intervals based on a post‐conservative model selection estimator.     

Representative Bureaucracy, Organizational Strategy, and Public Service Performance: An Empirical Analysis of English Local Government

The theory of representative bureaucracy suggests that organizationsperform better if their workforces reflect the characteristicsof their constituent populations. The management literatureimplies that the impact of representative bureaucracy is contingenton organizational strategy. Our empirical evidence on Englishlocal government is inconsistent with the basic theory of representativebureaucracy but supports a moderating effect of organizationalstrategy. Representative bureaucracy is negatively associatedwith citizens' perceptions of local authority performance. However,organizations pursuing a prospector strategy are able to mitigatethis negative relationship.     

Controlling motherhood: observations on the culture of the La Leche League*

Les arguments de cet article sont avant tout méthodologiques: que les caractéristiques distinctives d'un groupe peuvent être identifiees grâce à une analyse de ses activités de controle social; et qu'un examen des comportements de contrôle d'un groupe devrait commencer par l'étude de sa culture avec une attention particulière pour sa problématique centrale. Des éléments à l'intérieur de la culture d'un groupe contiennent des sources importantes de contrôle du comportement et contribuent à la nature de l'ordre social à l'intérieur de ce groupe. C'est dans la culture que des sanctions normales et des limitations de ces sanctions sont soit déduites, soit explicitement articulées. Le mode d'analyse est illustre par l'étude de la culture de la Ligue La Lèche, une organisation d'auto-assistance pour les femmes qui souhaitent alaiter leurs bébés au sein. L'article conclut que le succès de la ligue à encourager ses membres à s'adapter à un style de vie substantiellement plus coûteux peut être attribuer à quelques caractéristiques importantes du contenu de son problème central et son articultation avec d'autres aspects de la culture. Les données analysées proviennent des publications de la Ligue La Lèche, d'observations de participants aux réunions de la Ligue et d'interviews avec des membres de la Ligue. Le site de recherche est Ottawa, en Ontario. The arguments of this paper are primarily methodological: that distinguishing characteristics of a group can be identified through an analysis of its social control activities; and that an examination of a group's control behaviours should begin with a study of its culture, with special attention to its central problematic. Elements within the culture of a group contain important sources of behavioural control and contribute to the nature of the social order within that group. It is within the culture that normal sanctions, and limitations upon sanctions are either inferred or are explicitly articulated. The mode of analysis is illustrated through the study of the culture of the La Leche League, a self-help organization for women who want to breastfeed their babies. The paper concludes that the League's success in encouraging its members to adapt a lifestyle with substantial costs can be attributed to some important characteristics of the content of its central problem and its articulation with other aspects of the culture. The data analyzed come from La Leche League publications; participant observation at League meetings; and interviews with League members. The research site is Ottawa, Ontario.     

Accommodating Uncertainty in Comparative Risk

Comparative risk projects can provide broad policy guidance but they rarely have adequate scientific foundations to support precise risk rankings. Many extant projects report rankings anyway, with limited attention to uncertainty. Stochastic uncertainty, structural uncertainty, and ignorance are types of incertitude that afflict risk comparisons. The recently completed New Jersey Comparative Risk Project was innovative in trying to acknowledge and accommodate some historically ignored uncertainties in a substantive manner. This article examines the methods used and lessons learned from the New Jersey project. Monte Carlo techniques were used to characterize stochastic uncertainty, and sensitivity analysis helped to manage structural uncertainty. A deliberative process and a sorting technique helped manage ignorance. Key findings are that stochastic rankings can be calculated but they reveal such an alarming degree of imprecision that the rankings are no longer useful, whereas sorting techniques are helpful in spite of uncertainty. A deliberative process is helpful to counter analytical overreaching.     

Traces of Norwegian influence in the development of social services in Minnesota, USA, 1890–1920

This article explores areas where Norwegian Americans contributed to the burgeoning social welfare system being established in Minnesota from 1890 to 1920. The array of services developed was due in part to the belief among native-born Americans that immigrant groups should establish their own helping networks and care for themselves. The Lutheran Church played a central part in this development by initiating and sponsoring new undertakings. Issues discussed during this period are found to be still central among both immigrant groups and 'natives' today. Concerns regarding mother tongue, religion, integration in schools and use of existing welfare systems continue to stir up 'natives' and 'settlers' alike.     

Stability and change in levels and structure of subjective well-being: USA 1972 and 1988

It has long been recognized that more needs to be known about the dynamic behavior of measures of subjective well-being. As the Social Indicators Movement matures, it becomes possible to explore issues of stability and change in measures of perceived life quality. To contribute to this research need, this paper reports results of administering similar measures of subjective well-being to several representative samples of the adult American population assessed 16 years apart — in 1972 and in 1988. Results show that in the United States, which has experienced reasonably continuous economic prosperity and no major social disruptions during these 16 years, thestructure of the subjective well-being measures (i.e. how they relate to one another and how concern-level measures predict global-level measures) has remained remarkably constant. This is in accord with expectation and enhances confidence in the validity and usefulness of the measures. Changes were found in thelevels of some of the measures: For the total population, feelings about life-as-a-whole, oneself, and one's own health became notably more positive; evaluations of own income and the national government also gained; but assessments of one's own family life, social relationships, community, and job remained fairly stable. Some of these gross changes were general throughout the population, but others occurred only for certain age, racial, or socioeconomic groups. Younger cohorts and people with higher educations and incomes showed more gains in subjective well-being than other groups.     

Population issues and social indicators of well-being

Relating demographers' measures of various population characteristics (size, growth/decline, density, age/sex structures, migration, et cetera) to measures of well-being recently developed within the social indicators movement promises to provide new knowledge about the linkage of population and well-being that can enhance decision making about important population issues. A conceptual schema is presented that suggests specific relationships to examine at various levels of aggregation, that helps to classify research already done in this area, and that helps to identify "holes" in the knowledge base. Some special methodological features of research in this area suggest considerable time and care will be required to produce dependable new knowledge. These include: (a) the inherent multilevel nature of the relationships (involving properties of individuals and collectivities); (b) the slow rate at which population characteristics change; (c) the absence of much good well-being data from the past; and (d) the limited nature of the collectivities for which population data are available.     

The robustness of chain block designs and coat-of-mail designs

Chain block designs are relatively vulnerable to loss of information when missing values or outliers may occur. An alternative class of designs, coat-of-mail designs, are proposed and the relative robustness of the two types of design are compared.