Counterfactual Experiments Research Articles

Dynamical Counterfactual Inference under Time-series Model for Waterflooding Oilfield

The performances of numerical simulation and machine learning in production forecasting are severely dependent on precise geological modeling and high-quality history matching. To address these challenges, causal inference is an effective methodology since it can provide a causality for formalizing causality in history, not statistical dependence. In this paper, to dynamically predict oil production from causality existed in waterflooding oilfield, a dynamical counterfactual inference framework is built to predict oil production. The proposed framework can forecast the oil production under non-observation of engineering factors, i.e., counterfactual, and provide the causal effect of engineering factors impacting on oil production. Meanwhile, combining with the practice exploitation in engineering factor impacting on production, a counterfactual experiment is designed to execute counterfactual prediction. Compared with general machine learning and statistical models, our results not only show better performance in oil production flooding but also guide the specific optimization in improving production, which holds more practical application significance.

Occupational reallocation within and across firms: Implications for labor market polarization

This study analyzes how labor market frictions interact with firms’ decisions to reallocate workers across different occupations during labor market polarization. We compare the patterns of occupational reallocation within and across firms in the US and Germany in recent years. We find within-firm reallocation contributes significantly to the decline in employment in routine occupations in Germany, but much less so in the US. We construct a general equilibrium model of firm dynamics and find the model with different firing taxes can replicate the difference in firm-level adjustment patterns across these countries. We conduct two counterfactual experiments for each country, highlighting the different roles played by the within-firm cost of reorganizing the occupational mix and across-firm frictions created by firing taxes. The results suggest the latter plays a more significant role in labor market polarization. Higher firing costs would lead to greater and faster polarization in the US.

The potential burden from urbanisation on heat-related mortality in São Paulo, Brazil

Heatwave events are associated with increased hospital admissions and mortality rates worldwide. Heat exposure is often exacerbated by the Urban Heat Island (UHI) effect, especially in large cities. However, an accurate quantification of the additional burden related to heat from urbanisation is understudied in Latin American cities. Therefore, we used advanced meteorological modelling to simulate 2 m air temperature at 1 km spatial resolution across São Paulo, estimating UHI intensity attributing of all-cause mortality to heat and UHI during the 2014 heatwave. The Local Climate Zone classification system provided input land use parameters. Our model was validated against 31 weather stations and bias-corrected using linear regression. We calculated heat-related all-cause mortality, stratified by age, using the modelled temperature data, estimating 394 heat-related deaths. A counterfactual experiment, replacing urban areas with natural land, quantified the additional UHI burden. We found that the UHI may contribute to 69–70% of heat-related mortality (modelled temperature scenario). This work motivates the use of appropriate urban climate data, including careful model validation and bias correction, when assessing heat exposure and health risk assessment. It also highlights the health impacts from heatwaves in cities such as São Paulo, which are likely to increase due to climate change.

Wealth Elasticity, Agent Heterogeneity and Fiscal Dynamics

The literature on fiscal policy has identified wealth elasticity of labor supply and agent heterogeneity as the linchpin of shaping different transmission mechanisms for fiscal policy. In this paper, I build these two important edifices on an otherwise canonical business-cycle model to conduct quantitative analyses of various fiscal policies. To underscore the relative importance of agent heterogeneity, I compare results from the model with agent heterogeneity with those from a standard model without agent heterogeneity. I also compare results from changes in wealth elasticity to analyze how the model ingredient delivers different macroeconomic and fiscal dynamics. The main finding reveals that in the context of the model formulated in this paper, the capital tax is the least detrimental to business cycles and fiscal states and the consumption tax is the most appropriate policy tool since it yields positive output and tax-revenue multipliers in the short and long runs. All the results from counterfactual experiments suggest that macroeconomic and fiscal outcomes hinge fundamentally on agent heterogeneity and wealth elasticity. An important value that this paper adds to the literature is to offer results from a comprehensive analysis of all possible fiscal policies by considering both essential forces.

Das House Kapital: A Long-Run Theory of House Prices and Housing Wealth

Abstract Over the last 70 years, many advanced countries have experienced growing real house prices and an increasing housing wealth-to-income ratio. To explain these long-run patterns, this paper introduces a novel multi-sector growth model where housing services are produced using non-reproducible land and reproducible structures. Land is also employed in the non-housing sector. First, we identify two fundamental mechanisms driving the long-run increase in the real house price: (i) technological progress in the construction sector lags behind the technological progress of the rest of the economy and (ii) housing production is more land-intensive than non-housing production. Second, we study transitional dynamics for the US, UK, France, and Germany. Our calibrated model explains most of the observed increase in the housing wealth-to-income ratio since 1950. Counterfactual experiments identify initially low stocks of residential structures and non-residential capital as key exogenous drivers for this increase. The associated investment incentives led to a long-lasting construction boom and steadily increasing land scarcity, boosting residential land prices.

Plant-level adjustments to imports and exports at the extensive margin

This paper presents an empirical analysis of plant level responses to the China trade shock based upon a DSGE framework with heterogeneous firms. The empirical analysis shows that soaring imports from China are associated with a higher probability of plant closure. By contrast, firms in export oriented industries are less likely to exit. We rationalize these findings by several counter-factual experiments based upon a DSGE framework. Imports always raise the exit rate but the export-effect is ambiguous. More exports fuel competition among domestic rivals associated with more exits. However, this competition effect disappears when the share of exporters is extremely high. The effects of imports and exports on firm entry are close to zero in both theory and empircs. We also introduce a novel productivity shock channel. Additional export sales better protect firms from other shocks. We show this by introducing negative productivity shocks into the model.

Talent, Geography, and Offshore R&D

Abstract I model and quantify the impact of a new dimension of globalization: offshore R&D. In the model, firms employ researchers across the globe to develop new product blueprints and then engage in offshore production and exporting. Frictions impeding trade and the separation of production from R&D lead to a “market-access” motive for offshore R&D, while cross-country differences in the distributions of firm knowhow and worker ability generate a “talent-acquisition” motive. I discipline the model using empirical facts derived from a new firm-level dataset. Counterfactual experiments show that the two motives can account for a significant portion of the observed offshore R&D. Incorporating offshore R&D amplifies the gains from globalization by a factor of 1.3 and generates new implications for the impacts of traditional forms of global integration, namely trade and multinational production.

Decomposing China’s bilateral export growth: A firm-regional-transactions structural gravity approach

This paper undertakes a decomposition of China’s firm-transaction-level exports with an emphasis on aggregates thereof at the prefectural level. Chinese prefectures are large and, relative to some smaller countries, the larger ones are visible players in the world market. We decompose China’s prefecture-product-level bilateral exports to foreign countries into their main components. These are factor costs, market potential (including trade frictions and foreign expenditure) as well as quality and productivity. We consider the latter two as to be drawn from a bivariate Pareto distribution that is specific to prefectures, products, and countries. This strategy enables retrieving fixed market access costs and permits conducting counterfactual experiments with a focus on the distribution across Chinese prefectures.

Fertility decisions and the norm of intergenerational support to aging parents

AbstractThis paper examines the impact of the social norm of intergenerational support to aging parents on fertility decisions in developing economies. The traditional expectation of receiving support from adult children in old age has historically been a significant factor in the decision to have children, especially in developing countries. The study develops a life cycle model that endogenizes fertility choices and incorporates the expectation of transfers from children based on the filial responsibility norm. We utilize household survey data from Indonesia to estimate earnings profiles and uncertainties over the life cycle, and to estimate the transfers from adult children to parents to indicate the current strength of the norm in 2000s. We conducted counterfactual experiments to explore the impact of the filial responsibility norm on fertility and found that a weakening of the norm could account for a significant proportion of the decline in the total fertility rate.

Competition and regional Phillips curve: Evidence from China.

The product market competition affects the non-neutrality of monetary policy. This paper quantitatively assesses its impact on the slope of the Phillips curve through the channels of nominal and real rigidity. We build a New Keynesian model using the Kimball aggregator and the Calvo staggered pricing scheme. We show that a more competitive market environment has opposite effects on the slope of the Phillips curve by increasing the real rigidity and lowering the nominal rigidity. We then estimated the model using regional data of China. The Bayesian estimation shows that the response of inflation-output trade-off is larger in the region with a high degree of competition. Counterfactual experiments demonstrate that nominal rigidity has a dominant role and accounts for the majority of the difference in the Phillips curve, while the contribution of real rigidity is relatively minor. Our results highlight the key role of nominal rigidity in determining the inflation-output dynamics.

Foreign Competition along the Quality Ladder

AbstractWe document that firms with low prices are more impacted by low-cost competition. To explain this pattern, we propose an empirical model of trade with random-coefficient demand and endogenous product quality. Unlike commonly used demand systems (e.g., constant elasticity of substitution, nested logit), this model generates rich substitution patterns and implies an ‘escape-competition’ effect: in response to low-cost competition, firms have an incentive to upgrade their product quality. The estimation of the model reveals significant heterogeneity in consumer preferences. Counterfactual experiments suggest that the ‘China shock’ was significantly more damaging to firms at the bottom of the price distribution, and that quality upgrading had a limited role at mitigating the shock.

The 1996 Mid-Atlantic Winter Flood: Exploring Climate Risk through a Storyline Approach

Abstract This article explores the application of thermodynamic perturbations to a historical midlatitude, wintertime, rain-on-snow flood event to evaluate how similar events may evolve under different climate forcings. In particular, we generate a hindcast of the 1996 Mid-Atlantic flood using an ensemble of 14-km variable-resolution simulations completed with the U.S. Department of Energy’s global Energy Exascale Earth System Model (E3SM). We show that the event is skillfully reproduced over the Susquehanna River Basin (SRB) by E3SM when benchmarked against in situ observational data and high-resolution reanalyses. In addition, we perform five counterfactual experiments to simulate the flood under preindustrial conditions and four different levels of warming as projected by the Community Earth System Model Large Ensemble. We find a nonlinear response in simulated surface runoff and streamflow as a function of atmospheric warming. This is attributed to changing contributions of liquid water input from a shallower initial snowpack (decreased snowmelt), increased surface temperatures and rainfall rates, and increased soil water storage. Flooding associated with this event peaks from around +1 to +2 K of global average surface warming and decreases with additional warming beyond this. There are noticeable timing shifts in peak runoff and streamflow associated with changes in the flashiness of the event. This work highlights the utility of using storyline approaches for communicating climate risk and demonstrates the potential nonlinearities associated with hydrologic extremes in areas that experience ephemeral snowpack, such as the SRB.

Alzheimer’s Disease mechanistic pathways were discovered through in silico experiments in causal AI based digital twins

AbstractBackgroundBeyond a single‐gene‐based approach, pathway‐based approaches have become important in drug development paradigm. Several pathways are known for the pathogenesis of Alzheimer’s disease (AD), but these may be from animal studies or observational human subject studies. Here, we focus on identification of gene pathways with direct causal relationships to AD through Gemini Digital Twins of AD patients.MethodTwo independent and complementary AD Digital Twins were built using an A.I. platform REFSTM [aitiabio.com], based on Bayesian network models of ADNI and ANMerge consortium data which reverse‐engineered the connectivity of ∼59K and ∼35K multi‐modal variables and a wide range of AD‐related outcomes (cognitive, imaging and biomarker) profiled from 317 subjects (Control:MCI:Dementia = 97:191:29) and 199 subjects (Control:MCI:Dementia = 65:63:71), respectively [Fig1]. The average causal effect of each gene (at the blood gene expression level) on each outcome was estimated through in silico counterfactual experiments. Pathway enrichment analyses were performed on metascape.org.ResultWe identified a total of 573 and 300 AD‐driver‐genes from ADNI‐ and ANMerge‐based Digital Twins, of which 41 overlapped (significance‐of‐overlap = 6e‐15). These 41 robustly validated genes, of which 61% were identified by the network as causal mediators of APOE4 genotype, were enriched in several biological processes known for AD such as ‘regulation of RNA splicing’ and ‘macrophage activation’ [Fig2]. The AD‐driver‐genes were also enriched in other interesting pathways in both Digital Twins, such as ‘immune response’ and ‘oxidative stress’. Furthermore, the ADNI‐specific AD‐driver‐genes, representing a sample population with a higher proportion of MCI patients, were related to additional processes such as ‘mitochondrial respiration’ and ‘ribosomal sub‐complexes pathways’, previously known to be involved specifically in early stage AD.ConclusionWe prioritized several AD‐related pathways for potential drug development identified from causal driver genes discovered and validated in two independent Digital Twins. Some of these pathways are more related to proteopathic biochemical phase of AD, rather than cellular phase, suggesting an area of a potential early intervention. Investigating causal genes from two complementary consortium studies with a different proportion of early vs. late stage AD patients may be a promising approach to identify disease‐stage‐specific and disease‐stage‐common signatures in the pathway‐based drug discovery efforts.

The unequal effects of the health-economy trade-off during the COVID-19 pandemic.

Despite the global impact of the coronavirus disease 2019 pandemic, the question of whether mandated interventions have similar economic and public health effects as spontaneous behavioural change remains unresolved. Addressing this question, and understanding differential effects across socioeconomic groups, requires building quantitative and fine-grained mechanistic models. Here we introduce a data-driven, granular, agent-based model that simulates epidemic and economic outcomes across industries, occupations and income levels. We validate the model by reproducing key outcomes of the first wave of coronavirus disease 2019 in the New York metropolitan area. The key mechanism coupling the epidemic and economic modules is the reduction in consumption due to fear of infection. In counterfactual experiments, we show that a similar trade-off between epidemic and economic outcomes exists both when individuals change their behaviour due to fear of infection and when non-pharmaceutical interventions are imposed. Low-income workers, who perform in-person occupations in customer-facing industries, face the strongest trade-off.

International trade and the risk in bilateral exchange rates

Exchange rate volatility falls after a trade deal, driven by a decline in the systematic component of risk. The average trade deal increases trade by 50 percent over five years, reducing systematic risk by a third of a standard deviation across countries. We examine this connection in an Armington model where the structure of trade networks determines the risk in exchange rates. We estimate our model to current data and find i) that countries at the periphery of the world trade network benefit the most from lower trade barriers and ii) that a counterfactual experiment of a trade war between the US and China shows a global increase in currency risk, with effects concentrated among peripheral countries.

Moral Support and Performance

This study presents unique empirical evidence on the importance of moral support for performance. We take advantage of an unusual change in Argentinean football legislation. In August 2013, as a matter of national security, the Argentinean government forced all teams in the first division to play their games with only home team supporters. Supporters of visiting teams were not allowed to be in stadiums during league games. We estimate the effect of this exogenous variation of supporters on team performance and find that visiting teams are on average about 20% more likely to lose without the presence of their supporters. As a counterfactual experiment, we run the analysis using contemporaneous cup games, where the visiting team supporters were allowed to attend, and find no effect of the ban on those games. Moreover, the ban does not seem to bias the decisions of referees, the lineups, or the market value of the teams, suggesting that the effect on team performance is due to the loss of moral support rather than other factors. Finally, we find that moral support is more relevant when there is equal power between the two teams, suggesting that moral support compensates the power of monetary resources. This paper provides a proof of concept of moral support as an important nonmonetary resource, even in settings with high monetary incentives. This paper was accepted by Yan Chen, behavioral economics and decision analysis. Supplemental Material: The e-companion and data are available at https://doi.org/10.1287/mnsc.2021.02906 .

Intergenerational transfers in China: What are the patterns of the transfers and when do the transfers occur?

AbstractChina's social safety net is still underdeveloped, hence family support in the form of intergenerational transfers often serves as a substitute for the public transfer system. Using data from the China Health and Retirement Longitudinal Study, this paper finds that both upstream inter‐vivos transfers (from children to parents) and downstream inter‐vivos transfers (from parents to children) are prevalent in urban China. Moreover, the relative income status of the parent and children has an impact on inter‐vivos transfers. To investigate what economic factors generate the observed patterns of inter‐vivos transfers, this paper adopts a general equilibrium life‐cycle model in which overlapping generations are altruistically linked and calibrates the model to match data from urban China. Counterfactual experiments of removing one source of economic risk or modifying the social security replacement rate from the baseline model at a time reveal that intergenerational transfers mainly serve as informal insurance against the income risk of the children.

Do gold prices respond more to uncertainty shocks at the zero lower bound?

This paper employs a nonlinear Interacted-VAR model to evaluate the effects of economic uncertainty shocks on gold prices at the zero lower bound (ZLB) regime. We use different measures of economic uncertainty and estimate their interaction effects with monetary policy stances. We find that the responses of gold prices to uncertainty shocks are much stronger and more persistent during the ZLB periods. The regime-dependent effects are shown not to be driven by the business cycle, and are robust to alternative economic uncertainty measures. Our counterfactual experiment further demonstrates that the role of monetary policy stance is quantitatively important for explaining the different responses of gold prices to uncertainty shocks. These results - viz, the added value that gold assets appear to provide to diversified investment portfolios during periods of heightened economic uncertainty - have important implications regarding the circumstances whereby gold's likely effectiveness as a protective shield can be maximized.

Occupations, retirement, and the value of disability insurance

Occupations vary greatly in physical intensity, and there are, correspondingly, many differences in later-life work disability risk, retirement patterns, and applications for Social Security Disability Insurance (SSDI)—a national program that insures shocks to work productivity due to disability, with about nine million current beneficiaries in the U.S. In light of its widespread coverage and large differences in utilization across occupations, this paper aims to measure the value of the SSDI program across a broad population and the extent to which it influences the major choice of occupation. Using data from the Health and Retirement Study and O*NET, I estimate a life-cycle equilibrium model of occupational choice, work, savings, and disability at older ages, and find that incorporating occupations and preference heterogeneity is an integral part of understanding work and SSDI application behavior. While SSDI coverage is nearly universal and the premiums from workers are uniform, estimates suggest that the value of the program varies greatly, from being worth 2.1 to 14.5 percent of earned income—depending on preferences and choice of occupation—though for all groups it is welfare improving. I also find that SSDI plays an important role in the choice of occupation for older workers, providing an insurance value that results in over ten percent more people choosing physically intense, blue-collar occupations at older ages. This overall effect, however, masks the underlying selection of less risk-averse individuals into blue-collar jobs, which, if not accounted for, would lead to overstating the effects of the SSDI program on behavior. A counterfactual experiment aimed at eliminating the distortion created by uniform SSDI tax rates shows that—even without any additional employer accommodation of disabled workers—occupation-specific tax rates lead to fewer aggregate work-disabled years.

Stable outcomes and information in games: An empirical framework

Empirically, many strategic settings are characterized by stable outcomes in which players’ decisions are publicly observed, yet no player takes the opportunity to deviate. To analyze such situations in the presence of incomplete information, we build an empirical framework by introducing a novel solution concept that we call Bayes stable equilibrium and computationally tractable approaches for estimation and inference. Our framework allows the researcher to be agnostic about players’ information and the equilibrium selection rule. In an application, we study the strategic entry decisions of McDonald’s and Burger King in the US. While the Bayes stable equilibrium identified set is always (weakly) tighter than the Bayes correlated equilibrium identified set, our results show that the former can be substantially tighter in practice. In a counterfactual experiment, we examine the impact of increasing access to healthy food on the market structures in Mississippi food deserts.