Corruption and Growth Under Weak Identification

The goal of this paper is to revisit the influential work of Mauro [1995] focusing on the strength of his results under weak identification. He finds a negative impact of corruption on investment and economic growth that appears to be robust to endogeneity when using two-stage least squares (2SLS). Since the inception of Mauro [1995], much literature has focused on 2SLS methods revealing the dangers of estimation and thus inference under weak identification. We reproduce the original results of Mauro [1995] with a high level of confidence and show that the instrument used in the original work is in fact 'weak' as defined by Staiger and Stock [1997]. Thus we update the analysis using a test statistic robust to weak instruments. Our results suggest that under Mauro's original model there is a high probability that the parameters of interest are locally almost unidentified in multivariate specifications. To address this problem, we also investigate other instruments commonly used in the corruption literature and obtain similar results.

Performance Evaluation of the New Connecticut Leading Employment Index Using Lead Profiles and BVAR Models

Dua and Miller (1996) created leading and coincident employment indexes for the state of Connecticut, following Moore's (1981) work at the national level. The performance of the Dua-Miller indexes following the recession of the early 1990s fell short of expectations. This paper performs two tasks. First, it describes the process of revising the Connecticut Coincident and Leading Employment Indexes. Second, it analyzes the statistical properties and performance of the new indexes by comparing the lead profiles of the new and old indexes as well as their out-of-sample forecasting performance, using the Bayesian Vector Autoregressive (BVAR) method. The new indexes show improved performance in dating employment cycle chronologies. The lead profile test demonstrates that superiority in a rigorous, non-parametric statistic fashion. The mixed evidence on the BVAR forecasting experiments illustrates the truth in the Granger and Newbold (1986) caution that leading indexes properly predict cycle turning points and do not necessarily provide accurate forecasts except at turning points, a view that our results support.

Effects of Study Area Size on Geographic Characterizations of Health Events: Prostate Cancer Incidence in Southern New England, USA, 1994-1998.

ABSTRACT : BACKGROUND : We consider how representations of geographic variation in prostate cancer incidence across Southern New England, USA may be affected by selection of study area and/or properties of the statistical analysis. METHOD : A spatial scan statistic was used to monitor geographic variation among 35,167 incident prostate cancer cases diagnosed in Massachusetts, Connecticut and Rhode Island from 1994 to 1998, in relation to the 1990 populations of men 20+ years of age living in that region. Results from the combined-states analysis were compared to those from single-states. Impact of scanning procedures set to examine up to 50% or no more than10% of at-risk populations also was evaluated. RESULTS : With scanning set to 50%, 5 locations in the combined-states analysis were identified with markedly distinct incidence rates. Fewer than expected cases were estimated for nearly all Connecticut, Rhode Island and West Central Massachusetts, whereas census tracts on and around Cape Cod, and areas of Southwestern Connecticut and adjacent to greater Boston were estimated to have yielded more than expected incidence. Results of single-state analyses exhibited several discrepancies from the combined-states analysis. More conservative scanning found many more locations with varying incidence, but discrepancies between the combined- and single-state analysis were fewer. CONCLUSION : It is important to acknowledge the conditional nature of spatial analyses and carefully consider whether a true cluster of events is identified or artifact stemming from selection of study area size and/or scanning properties.

Do Proctored Exams Matter In Online Classes?

Does the format of assessment (proctored or un-proctored exams) affect test scores in online principles of economics classes? This study uses data from two courses of principles of economics taught by the same instructor to gain some insight into this issue. When final exam scores are regressed against human capital factors, the R-squared statistic is 61.6% for the proctored format exams while it is only 12.2% for the un-proctored format. Three other exams in the class that had the proctored final were also un-proctored and also produced lower R-squared values, averaging 30.5%. These two findings suggest that some cheating may have taken place in the un-proctored exams. Although it appears some cheating took place, the results suggest that cheating did not pay for these students since the proctored exam grades were 4.9 points higher than the un-proctored exam grades although this difference was significantly different at only the 10% level. One possible explanation for this is that there was slightly higher human capital in the class that had the proctored exam although this must have occurred by chance since the students did not know if the exams were going to be proctored in advance so there is no issue of selection bias. A Oaxaca decomposition of this difference in grades was conducted to see how much was due to human capital and how much was due to the differences in the rates of return to human capital. This analysis reveals that 17% of the difference was due to the higher human capital with the remaining 83% due to differences in the returns to human capital. It is possible that the un-proctored exam format does not encourage as much studying as the proctored format reducing both the returns to human capital and the exam scores.