On January 2016, General Electric announced it was moving its headquarters from Fairfield, Connecticut to Boston, Massachusetts. The political impact on the state of losing a storied, blue-chip manufacturing and technology conglomerate was significant. But what about the economic impact? In this paper we gauge the economic impact of GE’s departure. Specifically, we examine the performance of Total NonFarm Employment for the Bridgeport-Norwalk-Stamford area. We compare performance before and after January 16, 2016. We do so against the performance of the same series at the national level. Over the period between January 2016 and February 2017, we find a cumulative decline of 66,000 jobs in the Bridgeport-Stamford-Norwalk Area. This is considerably more than the specific (and oft-cited) 200 jobs that GE moved to Boston. However, while the impact on Total Employment measured above is statistically significant and attributable to GE, a simple glance at the time series reveals that the environment in Connecticut had turned for the worse prior to GE’s announcement. Looking back via a formal statistical change-point analysis of the same Total Employment series we find statistical evidence that conditions did turn worse in the region in November 2013. Thus, we surmise that GE’s exit was a company bailing a flailing ship.
THE VALIDITY OF CONSUMER SENTIMENT IN SMALL-AREA ECONOMIC FORECASTING: A NAÏVE BAYES ANALYSIS
A.E. Rodriguez & Carolyne Cebrian
Obtaining an accurate picture of the current state and direction of the regional economy is particularly important to local decision-makers, including shopkeepers, academic institutions, and state and local government agencies. Traditional, survey-based sentiment indices have long-existed and are used for this purpose. But current abilities to source online data to map consumer sentiment has kindled interest in their usefulness in regional economic forecasting. The appeal of tailored sentiment indices and other similar online-sourced measures are their seeming immediacy and their ability to capture information in more relevant geographic and product domains. If decision-makers are to profitably rely with reasonable confidence from the increased availability of sentiment indices they will have to learn to effectively integrate domain knowledge, conventional or tailored online sentiment indices and traditional data. Perhaps more importantly, users will have to be assured of sentiment index validity in enhancing regional economic forecasts. We test sentiment index relevance in this paper reproducing results of a popular local forecast. Specifically, we appraise whether there are measurable improvements from the presence of a sentiment index to the New Haven Register’s Economic Scorecard, a popular regional forecast model. The model is a binary directional prediction model. Succinctly, we find measurable improvements in the model’s predictive accuracy of the Economic Scorecard. We speculate as to the generalizability of our results, especially regarding the use of other online-sourced nowcasting metrics.
THE PERILS OF RELYING ON STATE ECONOMIC PERFORMANCE RANKINGS WITHOUT ADJUSTING FOR HETEROGENEITY
A.E. Rodriguez and Brian Marks
State performance rankings are ubiquitous. But most rankings fail to recognize the heterogeneity inherent in the seemingly “objective” variables utilized to structure the ordering. A more parsimonious representation can be accomplished by adjusting the ordering variable by its most important attributes. To demonstrate the procedure, we utilize a state ranking based on Cumulative GDP Growth. We identify the relative importance and sensitivity of several popular variables used in explaining the variation in cumulative gdp growth performance among the states. Once identified, important variables can enhance the effectiveness of legislators and administrators’ policy-making efforts. State performance rankings are recast after adjusting cumulative gdp growth for the important drivers identified. The period examined is 2004-2014. To identify the importance and sensitivity of predictors we utilize random forests via the R packages relaimpo, Boruta, and random forests. Partial dependence depictions of the critical variables identified enable policy inferences. Specifically, we find that the top marginal personal tax rate and the number of state employees exert and uncommonly high influence in explaining variation in state performance rankings based on cumulative gdp-growth. The method proposed here is of general applicability and can be deployed to extract robust policy prescriptions based on a more accurate treatment of data given the limitations of traditional econometric models.
THE NEW HAVEN REGION AFTER THE GREAT RECESSION: A SHIFT-SHARE ANALYSIS
Steven Gillette, Jurgena Hysoli, Sean Kingsepp, Vanessa Lopez, Miles Mortali, Drew Ortone, Nathan Pitruzzello, Esin Cakan, Brian Marks, and A.E. Rodriguez
Shift-share analysis is a decomposition technique that is commonly used to measure attributes of regional change. In this method, regional change is decomposed into expected and regional (idiosyncratic) parts. We use it here to scrutinize the performance of the greater New Haven region relative to the performance of the national economy. We do so for the years following the great recession: June 2009 through September 2016. The approach provides a sense of the comparative advantage or disadvantage of the various sectors in our region. The results are distressing: the greater new haven region is lagging and underperforming badly. If it is to inform policy, the results here are likely to recommend generalized policies of broad applicability rather than targeted emphasis on a few seeming outperforming areas as appears to be the case with the current administration.
shift-share, comparative advantage, trends, and regional development