Bayesian Structural Time Series Model Research Articles

Quantifying cumulative effectiveness of green stormwater infrastructure in improving water quality

Stormwater runoff is one of the main sources of pollution in streams and receiving water bodies of major cities. Green Stormwater Infrastructure (GSI) is a set of distributed stormwater best management practices that absorb excess water, filter out sediment and pollutants, and help recharge groundwater.Despite the increasing popularity of GSI as means of stormwater management, our knowledge of their cumulative performance is limited. In this research, we apply an empirical approach to study the effectiveness of GSI in improving the water quality of four major receiving water bodies of Seattle, Washington, at the watershed scale. We use a Bayesian structural time series model and synthetic control method to build counterfactual scenarios of water quality in absence of GSI implementation and estimate the causal impact of GSI on water quality. We use monthly time series data of water quality parameters (water temperature, dissolved oxygen, surface PAR, chlorophyll a, Secchi depth, pH, light transmission, and fecal coliform) in Seattle's urban watersheds from 2004 to 2017. We also use a set of nine control variables to estimate the counterfactual water quality parameters. Our findings show that GSI improve some water quality parameters such as chlorophyll a, light transmission and Secchi depth, but increase water temperature and decrease dissolved oxygen in some water bodies.

A Time-Series Analysis of the Association Between Occupational Health Policies and Opioid Prescription Patterns in United States Active Duty Military Service Members From 2006 to 2018.

The Department of Defense (DoD) implemented numerous occupational health policies to guide opioid prescribing to active duty military service members (ADSM). This retrospective time series analysis evaluated the impact of DoD policies on opioid prescribing trends in ADSM from 2006 to 2018. Bayesian structural time-series models with a Markov chain Monte Carlo algorithm for posterior inference and a semi-local linear trend were constructed to estimate the impact of polices. Results indicate annual opioid proportions significantly decreased after the introduction of occupational health policies introduced in 2011 to 2012. Collectively, occupational policies were associated with a significant reduction (6.6%) in annual opioid rates to ADSM following 2012. This observed effect was associated with approximately 121,000 less opioid prescriptions dispensed in 2018 alone. Occupational health policy interventions were associated with reductions in opioid prescribing within the DoD.

Quantifying Cumulative Effectiveness of Green Stormwater Infrastructure in Improving Water Quality

Stormwater runoff is one of the main sources of pollution in streams and receiving water bodies of major cities. Green Stormwater Infrastructure (GSI) is a set of distributed stormwater best management practices that mimic natural systems, absorbing excess water, filtering out sediment and pollutants, and helping to recharge groundwater. Despite the increasing popularity of GSI as means of stormwater management, our knowledge of their performance is limited. Even though studies have shown effectiveness of individual installations in many cases, it is not clear whether the combination of these plants will deliver the same benefits. This research studies the effectiveness of GSI in improving the water quality of four major receiving water bodies of Seattle, Washington at the watershed scale. Using the monthly time series data of water quality indices from 2004 to 2017, as well as a set of nine control variables, we predict the counterfactual water quality levels in absence of GSI in each watershed and estimate the cumulative effect of GSI using a Bayesian structural time series model. Our findings show that GSI improves some physical water quality indices such as chlorophyll a, Secchi depth and light transmission, but increases the temperature in some waters.

Crude oil price analysis and forecasting: A perspective of “new triangle”

Abstract In this paper, the new structural characteristics and core influencing factors of the crude oil prices are summarized based on previous representative research results. Firstly, a newly dynamic Bayesian structural time series model (DBSTS) is developed to investigate the oil prices. In particular, Google trend is introduced as an indicator to reflect the impact of search data on the oil price. Secondly, the spike and slab method is employed to select core influence factors. Finally, the Bayesian model average (BMA) is utilized to predict the oil price. Experimental results confirm that the supply and demand of global crude oil and the financial market are still the main factors affecting the oil price. Furthermore, Google trend can reflect the changes in the crude oil price to a certain extent. Moreover, the impact of shale oil production on the oil price is gradually increasing, yet remains relatively small. In addition, the DBSTS model can identify turning points in historical data (such as the 2008 financial crisis). Finally, the findings suggest the DBSTS model has good predictive capabilities in short-term prediction, making it suitable for analyzing the crude oil prices.

Florida's Opioid Crackdown and Mortality From Drug Overdose, Motor Vehicle Crashes, and Suicide: A Bayesian Interrupted Time-Series Analysis.

In 2011, Florida established a prescription drug monitoring program and adopted new regulations for independent pain-management clinics. We examined the association of those reforms with drug overdose deaths and other injury fatalities. Florida's postreform monthly mortality rates-for drug-involved deaths, motor vehicle crashes, and suicide by means other than poisoning-were compared with a counterfactual estimate of what those rates would have been absent reform. The counterfactual was estimated using a Bayesian structural time-series model based on mortality trends in similar states. By December 2013, drug overdose deaths were down 17% (95% credible interval: -21, -12), motor vehicle crash deaths were down 9% (95% credible interval: -14, -4), and suicide deaths were unchanged compared with what would be expected in the absence of reform. Florida's opioid prescribing reform substantially reduced drug overdose deaths. Reforms may also have reduced motor vehicle crash deaths but were not associated with a change in suicides. More research is needed to understand these patterns. Bayesian structural time-series modeling is a promising new approach to interrupted time-series studies.

Noncooperative dynamics in election interference.

Foreign power interference in domestic elections is an existential threat to societies. Manifested through myriad methods from war to words, such interference is a timely example of strategic interaction between economic and political agents. We model this interaction between rational game players as a continuous-time differential game, constructing an analytical model of this competition with a variety of payoff structures. All-or-nothing attitudes by only one player regarding the outcome of the game lead to an arms race in which both countries spend increasing amounts on interference and counterinterference operations. We then confront our model with data pertaining to the Russian interference in the 2016 United States presidential election contest. We introduce and estimate a Bayesian structural time-series model of election polls and social media posts by Russian Twitter troll accounts. Our analytical model, while purposefully abstract and simple, adequately captures many temporal characteristics of the election and social media activity. We close with a discussion of our model's shortcomings and suggestions for future research.

Estimating Causal Effects in the Presence of Partial Interference Using Multivariate Bayesian Structural Time Series Models

Researchers regularly use synthetic control methods for estimating causal effects when a sub-set of units receive a single persistent treatment, and the rest are unaffected by the change. In many applications, however, units not assigned to treatment are nevertheless impacted by the intervention because of cross-unit interactions. This paper extends the synthetic control methods to accommodate partial interference, allowing interactions within predefined groups, but not between them. Focusing on a class of causal estimands that capture the effect both on the treated and control units, we develop a multivariate Bayesian structural time series model for generating synthetic controls that would have occurred in the absence of an intervention enabling us to estimate our novel effects. In a simulation study, we explore our Bayesian procedure’s empirical properties and show that it achieves good frequentists coverage even when the model is misspecified. Our work is motivated by an analysis of a marketing campaign’s effectiveness by an Italian supermarket chain that permanently reduced the price of hundreds of store-brand products. We use our new methodology to make causal statements about the impact on sales of the affected store-brands and their direct competitors. Our proposed approach is implemented in the CausalMBSTS R package.

A Bayesian Structural Time Series Approach for Predicting Red Sea Temperatures

Sea surface temperature (SST) is a leading factor impacting coral reefs and causing bleaching events in the Red Sea. A long-term prediction of temperature patterns with an estimate of uncertainty is thus essential for environment management of the Red Sea ecosystem. In this work, we build a data-driven Bayesian structural time series model and show its effectiveness in predicting future SST seasons with a high accuracy, and identifying the main predictive factors of future SST variability among a large number of factors, including regional SST and large-scale climate indices. The modeling scheme proposed here applies an efficient hierarchical clustering to identify interconnected subregions that display distinct SST variability over the Red Sea, and a Markov Chain Monte Carlo algorithm to simultaneously select the main predictors while the time series model is being trained. In particular, numerical results indicate that monthly SST can be reliably predicted for five months ahead.

A national intervention to reduce imaging for low back pain by general practitioners: a retrospective economic program evaluation using Medicare Benefits Schedule data

BackgroundThe overuse of diagnostic imaging for low back pain (LBP) in Australia results in unnecessary cost to the health system and, for patients, avoidable exposure to radiation. The 2013 NPS MedicineWise LBP program aimed to reduce unnecessary diagnostic imaging for non-specific acute LBP in the Australian primary care setting. The LBP program delivered referral pattern feedback, a decision support tool and patient information to 19,997 (60%) of registered Australian general practitioners (GPs). This study describes the findings from evaluation of the effectiveness of the 2013 LBP program at reducing X-ray and computed tomography (CT) scans of the lower back, and the financial costs and benefits of the program to the government funder.MethodsThe effectiveness of the 2013 LBP program was evaluated using population-based time-series analysis of administrative claims data of Medicare Benefits Schedule (MBS) funded X-ray and CT scan services of the lower back. The CT scan referral trend of non-GP health professionals was used as an observational control group in a Bayesian structural time-series model. A retrospective cost–benefit analysis and cost-effectiveness analysis was conducted using program costs from organisational records and reimbursement data from the MBS.ResultsThe 2013 NPS MedicineWise LBP program was associated with a statistically significant 10.85% relative reduction in the volume of CT scans of the lumbosacral region, equating to a cost reduction to the MBS of AUD$11,600,898. The best available estimate of program costs was AUD$141,154. Every dollar of funding spent on the 2013 LBP program saved AUD$82 of funding to the MBS for CT scan reimbursements. Therefore, from the perspective of the Australian Government Department of Health, the 2013 LBP program was cost saving. The program cost AUD$2.82 per CT scan averted in comparison to the scenario of no program. No association between the 2013 NPS MedicineWise LBP program and the volume of X-ray items on the MBS was observed.ConclusionsThe 2013 NPS MedicineWise LBP program reduced CT scan referral by GPs, in line with the program’s messages and clinical guidelines. Reducing this low-value care produced savings to the health system that exceeded the costs of program implementation.

Bayesian Structural Time Series and Regression Modeling for Sustainable Technology Management

Many companies take the sustainability of their technologies very seriously, because companies with sustainable technologies are better able to survive in the market. Thus, sustainable technology analysis is important issue in management of technology (MOT). In this paper, we study the management of sustainable technology (MOST). This focuses on the sustainable technology in various MOT fields. In the MOST, sustainable technology analysis is dependent on time periods. We propose a method of sustainable technology analysis using a Bayesian structural time series (BSTS) model based on time series data. In addition, we use the Bayesian regression to find the relational structure between technologies. To show the performance of our method and how the method can be applied to practical works, we carry out a case study using the patent data related to artificial intelligence technologies.

Using machine learning to quantify the impacts of genetically modified crops on US midwest corn yields

Global food security is becoming increasingly stressed by growing populations and climate change. To compensate for these stresses, crop yields must increase throughout the upcoming century. One of the more prominently featured solutions entails genetically modified crops, but their impacts on yields are contested. Here, we leverage machine learning techniques to examine the effects genetically modified crops have had on US corn yields. In particular, a principal components analysis conducted on US Midwest county yields reveals that the commercialization of genetically modified corn accentuated preexisting spatial disparities in production and explains approximately 6–12% of the region's inter-county variation in yields from 1980 to 2015. Additionally, counterfactual yield trajectories predicted by Bayesian structural time series models using non-genetically modified crops as synthetic controls suggest that the adoption of this biotechnology amounted to an approximate 13% increase in overall US corn yields from 1996 to 2015.

Assessing the effect of advertising expenditures upon sales: A Bayesian structural time series model

AbstractWe propose a robust implementation of the Nerlove‐Arrow model using a Bayesian structural time series model to explain the relationship between advertising expenditures of a countrywide fast‐food franchise network with its weekly sales. Due to the flexibility and modularity of the model, it is well suited to generalization to other markets or situations. Its Bayesian nature facilitates incorporating a priori information reflecting the manager's views, which can be updated with relevant data. This aspect of the model will be used to support the decision of the manager on the budget scheduling of the advertising firm across time and channels.

Impact of a public transit strike on public bicycle share use: An interrupted time series natural experiment study

IntroductionPromoting active transportation is an important public health objective. Limited research has examined the potential of interventions that highly constrain transportation and their potential impact on cycling. From November 1-7th, 2016, Philadelphia's transit workers went on strike, stopping all transit services in the city. MethodsWe used the strike event as a natural experiment to examine the impact of public transit strikes on use of Philadelphia's bicycle share program. We estimated the impact of the strike using two separate approaches, interrupted time series and Bayesian structural time series models. We estimated the impact of the intervention overall and stratified by membership type (members and non-members). Models controlled for the weather in Philadelphia (daily temperature and precipitation), and the rate of bicycle share use per 100,000 people in Washington, Boston, and Chicago. ResultsWe estimate the strike caused an increase of between 86 and 92 trips per 100,000 population (57% increase in use) on average in Philadelphia during the strike period. After the strike ridership quickly returned to baseline, decreasing by 80 trips per 100,000 population after the strike. Similarly, members and non-member ridership increased by 41 and 49 trips per 100,000 population on average during the strike period and quickly returned to baseline, respectively. ConclusionsOur results suggest that interventions that highly constrain transit can increase active transportation but the behavior may not be sustained after transit becomes available again.

Can Google search data help predict macroeconomic series?

We make use of Google search data in an attempt to predict unemployment, CPI and consumer confidence for the US, UK, Canada, Germany and Japan. Google search queries have previously proven valuable in predicting macroeconomic variables in an in-sample context. However, to the best of our knowledge, the more challenging question of whether such data have out-of-sample predictive value has not yet been answered satisfactorily. We focus on out-of-sample nowcasting, and extend the Bayesian structural time series model using the Hamiltonian sampler for variable selection. We find that the search data retain their value in an out-of-sample predictive context for unemployment, but not for CPI or consumer confidence. It is possible that online search behaviours are a relatively reliable gauge of an individual’s personal situation (employment status), but less reliable when it comes to variables that are unknown to the individual (CPI) or too general to be linked to specific search terms (consumer confidence).

Forecasting the 2021 local burden of population alcohol-related harms using Bayesian structural time-series.

Background and aimsHarmful alcohol use places a significant burden on health services. Sophisticated nowcasting and forecasting methods could support service planning, but their use in public health has been limited. We aimed to use a novel analysis framework, combined with routine public health data, to improve now‐ and forecasting of alcohol‐related harms.DesignWe used Bayesian structural time–series models to forecast alcohol‐related hospital admissions for 2020/21 (from 2015 to 2016).SettingEngland.ParticipantsWe developed separate models for each English lower‐tier local authority.MeasurementsOur primary outcome was alcohol‐related hospital admissions. Model covariates were population size and age‐structure.FindingsNowcasting validation indicated adequate accuracy, with 5‐year nowcasts underestimating admissions by 2.2% nationally and 3.3% locally, on average. Forecasts indicated a 3.3% increase in national admissions in 2020/21, corresponding to a 0.2% reduction in the crude rate of new admissions, due to population size changes. Locally, the largest increases were forecast in urban, industrial and coastal areas and the largest decreases in university towns and ethnically diverse areas.ConclusionsIn 2020/21, alcohol‐related hospital admissions are expected to increase in urban and coastal areas and decrease in areas associated with inward migration of younger people, including university towns and areas with greater ethnic diversity. Bayesian structural time–series models enable investigation of the future impacts of alcohol‐related harms in population subgroups and could improve service planning and the evaluation of natural experiments on the impact of interventions to reduce the societal impacts of alcohol.

Go Wide or Go Deep? Assortment Strategy and Order Fulfillment in Online Retail

Problem Definition: Expansions in product assortment undertaken by online retailers often engender operational challenges. In undertaking such expansions, retailers exercise a choice between expanding assortment depth or width. We empirically examine the relative impacts of assortment depth and width on a retailer’s order delivery performance and the subsequent effects on sales revenue. We further investigate the role of logistic flexibility in this relationship. Academic/Practical Relevance: Order delivery timeliness is a critical measure of operational success in online retail. The relative impact of expansions in product assortment depth and width on order fulfillment is theoretically unclear thereby making it difficult for managers to choose between these two options from an operational standpoint. We contribute to theory and practice by empirically clarifying the role of assortment strategy on order fulfillment performance. Methodology: Employing a unique panel dataset comprising more than 200 million orders, we study the effects of assortment strategy on delivery timeliness using an instrumental variable approach. We then utilize a two-stage model to estimate the eventual impact of order delivery performance on sales. Furthermore, we employ a matched difference-in-differences approach and a novel Bayesian structural time-series model that exploit the opening of new warehouses to confirm the relationship between delivery timeliness and improved sales. Results: We find that assortment width has a greater negative impact on order delivery timeliness relative to depth. Specifically, adding five subcategories increases average delivery times by 0.21 days. We identify that this is mediated by a proliferation of multi-item orders that induces logistical inefficiencies. Further, these delivery delays result in a 6.0% drop in future sales across all offered subcategories. Managerial Implications: Our study aids managerial-decision making by providing empirical evidence pertaining to the operational consequences of expansions in product assortment and offering insights on potential mitigation strategies.

Impact Evaluation of a Parolee-Based Focused Deterrence Program on Community-Level Violence

We estimate the impact of a parolee-based focused deterrence (“pulling levers”) intervention on community-level firearm and non-firearm violence in Rockford, Illinois, via a retrospective, quasi-experimental design. Focusing on incidents of firearm violence in Rockford over a period of 60 months (38 months pre-intervention, 22 months post-intervention), program impact is assessed using Bayesian Structural Time Series (BSTS) models, constructing a synthetic control-based counterfactual time series from National Incident-Based Reporting System (NIBRS) data from 59 non-treated cities of similar size. Relative to the synthetic control counterfactual, the intervention was associated with significant reductions in both firearm and non-firearm violence, particularly robberies, ranging from 6% to 30%. Consistent with research at other sites, these findings support the notion that focused deterrence strategies centered on high-risk parolees may result in reductions in firearm violence at the community level. The BSTS approach is a useful application for producing counterfactuals in retrospective quasi-experimental impact evaluations.

Effects of political instability in consolidated destinations: The case of Catalonia (Spain)

Catalonia is one of the most successful tourist regions in Spain. This article analyses the immediate impact that the instability associated to the recent political situation in Catalonia has had on the arrivals and spending of international tourists in the region using the classical Box-Jenkins method (ARIMA) and the more recent Bayesian Structural Time-Series Models. The results obtained indicate that during the final quarter of 2017, political events led to a reduction in the arrivals and spending of tourists in the region, although whether this impact reached statistical significance depends on the estimation method used.

Brexit Announcement: Immediate Impact on British Tourism in Spain

The United Kingdom constitutes the principal tourist source market for Spain. This research note analyzes the immediate impact of the Brexit on British tourism in Spain using the Bayesian structural time series models framework. The results obtained show that between July 2016 and September 2017, Brexit has not produced any initial negative effect on the arrival of British tourists or on their spending in Spain.

Using a Bayesian Structural Time–Series Model to Infer the Causal Impact on Cigarette Sales of Partial and Total Bans on Public Smoking

Abstract The Bayesian structural time series model, used in conjunction with a state–space model, is a novel means of exploring the causal impact of a policy intervention. It extends the widely used difference–in–differences approach to the time series setting and enables several control series to be used to construct the counterfactual. This paper highlights the benefits of using this methodology to estimate the effectiveness of an absolute ban on smoking in public places, compared with a partial ban. In January 2006, the Spanish government enacted a tobacco control law which banned smoking in bars and restaurants, with exceptions depending on the floor space of the premises. In January 2011, further legislation in this area was adopted, removing these exceptions. The data source used for our study was the monthly legal sales of cigarettes in Spain from January 2000 to December 2014. The potential control series were the monthly tourist arrivals from the United Kingdom, the total number of visitors from France, the unemployment rate and the average price of cigarettes. Analysis of the state–space model leads us to conclude that the partial ban was not effective in reducing the tobacco sold in Spain, but that the total ban contributed significantly to reducing cigarette consumption.