Economics Applications: Research-Specific Prompts

Abstract:
We examine the impact of remote work on employee productivity using a randomized
controlled trial conducted at a Fortune 500 company. 823 employees were randomly
assigned to either work-from-home (treatment) or office-based (control) conditions
for 6 months. Using detailed productivity metrics and time-use data, we find that
remote workers were 13% more productive (p<0.01). Administrative data from 2019-2020.
Data available via company partnership. Analysis code provided in supplementary materials.

RESEARCH QUESTION:
Does remote work affect employee productivity?

RESEARCH DESIGN:
- Type: Experimental
- Specific method: Randomized controlled trial (RCT)

DATA:
- Sample: Fortune 500 company employees
- Sample size: 823
- Time period: 2019-2020 (6-month treatment)
- Geographic scope: United States (implied)
- Data source: Administrative records, productivity metrics

IDENTIFICATION STRATEGY:
- Treatment: Work-from-home assignment
- Comparison: Office-based work (control group)
- Key assumption: Random assignment ensures treatment/control groups are comparable

OUTCOME MEASURES:
- Primary: Productivity metrics (13% increase)
- Secondary: Time-use data

REPRODUCIBILITY ASSESSMENT:
- Data availability: Restricted (company partnership required)
- Code availability: Mentioned (supplementary materials)
- Preregistration: Unknown

Papers to analyze:

[Paste all 10 research design extracts from Step 1]

1. Remote Work RCT (Chen et al., 2023): 13% productivity increase, Fortune 500, N=823
2. Hybrid Work Observational (Rodriguez, 2022): Mixed results by occupation, Survey N=2,400
3. WFH in Academia (Johnson et al., 2024): 8% decrease, UK universities, N=156 faculty
[... etc for all 10 papers]

ESTABLISHED FINDINGS:
What do we now know? (consensus findings across multiple papers)
- Finding 1: Remote work effects vary significantly by occupation type (6 of 10 papers)
- Finding 2: Productivity measurement methodology strongly influences results (5 of 10 papers)
- Finding 3: Short-term effects (<6 months) may differ from long-term (4 of 10 papers)

CONTRADICTIONS:
Where do papers disagree?
- Contradiction 1: Chen (2023) found 13% productivity increase, but Johnson (2024) found 8% decrease
- Possible reasons: Different sectors (corporate vs. academic), measurement approaches (objective metrics vs. self-report), sample characteristics (tech workers vs. faculty)

METHODOLOGICAL GAPS:
What methods haven't been tried?
- Gap 1: Only 2 of 10 studies use experimental designs (RCTs); rest are observational
- Gap 2: No studies exceed 12 months duration; long-term effects unknown
- Gap 3: No studies examine heterogeneous effects by worker characteristics (age, experience, family status)

CONTEXTUAL GAPS:
What populations/regions/time periods are understudied?
- Gap 1: Only 1 study from developing countries; 8 are US/UK-based
- Gap 2: No studies of remote work in manufacturing or service sectors (all knowledge work)
- Gap 3: All studies are pre-2025; post-pandemic norms may differ

THEORETICAL GAPS:
What mechanisms or explanations haven't been explored?
- Gap 1: Little research on why productivity effects vary (commute time savings? Reduced distractions? Different?)
- Gap 2: No examination of spillover effects (family, health, local economy)
- Gap 3: Team dynamics and collaboration quality understudied (mostly individual-level analysis)

HIGHEST-VALUE RESEARCH QUESTIONS:
Based on these gaps, what 3 studies would most advance the field?

1. Long-term RCT (2+ years) with heterogeneous effect analysis by worker demographics
   - Why it matters: Current evidence limited to short-term; policy needs long-run estimates
   - Suggested method: Multi-year RCT with rich demographic data and mechanism surveys

2. Developing country context study using administrative productivity data
   - Why it matters: Vast majority of workers globally are outside US/UK samples
   - Suggested method: Partner with large employer in India/Brazil/Nigeria for quasi-experimental design

3. Mechanism study examining causal pathways (commute savings vs. distraction reduction vs. scheduling flexibility)
   - Why it matters: Understanding why effects occur enables better policy design
   - Suggested method: Factorial RCT varying specific remote work components, mediation analysis

Task: Format as APA 7th

Sources to format:

Chen J and Rodriguez M, "Remote Work and Productivity: Evidence from a Randomized Experiment" Journal of Labor Economics 2023 vol 41 issue 2 pages 234-267 DOI: 10.1086/723456

Johnson et al 2024 Work from Home in Academia: productivity effects British Journal of Industrial Relations v62(3) pp445-478

rodriguez 2022 "hybrid work arrangements" survey evidence Management Science vol 40 pages 1234-1256
[... paste remaining 7 papers in rough format]

FORMATTED CITATIONS:

Chen, J., & Rodriguez, M. (2023). Remote work and productivity: Evidence from a randomized experiment. Journal of Labor Economics, 41(2), 234-267. https://doi.org/10.1086/723456

Johnson, T., Williams, A., & Lee, S. (2024). Work from home in academia: Productivity effects. British Journal of Industrial Relations, 62(3), 445-478.

Rodriguez, M. (2022). Hybrid work arrangements: Survey evidence. Management Science, 40, 1234-1256.

[... remaining citations properly formatted]

FLAGGED ISSUES FOR MANUAL REVIEW:

Paper 3 (Rodriguez, 2022):
- POTENTIAL ISSUE: Volume number 40 seems high for Management Science (typically vol. 60s-70s in 2022)
- RECOMMENDATION: Verify volume/issue number in original source

Paper 5 (Author name unclear):
- MISSING INFORMATION: First author initial not provided in rough citation
- RECOMMENDATION: Check original paper for complete author name

Paper 8 (Journal name abbreviated):
- VALIDATION CHECK: "AER" expanded to "American Economic Review" - verify this is correct journal

Dataset description:
Panel dataset of worker productivity from 2013-2022. Variables include worker_id,
firm_id, year, age, education, wage, hours_worked, productivity_score, remote_work_dummy.
Data from 5,000 workers across 50 firms. Source: Company HR records merged with
productivity tracking software.

Sample of data:
worker_id,firm_id,year,age,education,wage,hours_worked,productivity_score,remote_work
1001,5,2015,34,Bachelor,65000,40,87.3,0
1001,5,2016,35,Bachelor,68000,40,89.1,0
1001,5,2017,36,Bachelor,72000,42,91.2,1
1002,12,2015,28,Master,78000,45,76.5,0
1002,12,2016,29,Master,NA,40,82.1,0
1002,12,2017,30,Master,85000,40,NA,1
1003,12,2015,-5,PhD,95000,35,94.2,0
1004,5,2015,45,high school,52000,168,45.3,0
1005,8,2015,31,Bachelor,67000,40,88.7,0
1005,8,2015,31,Bachelor,67000,40,88.7,0

COMPLETENESS CHECK:
- Missing values: wage (2.3%), productivity_score (1.8%), remote_work (0.1%)
- Systematic missingness: Wage missingness concentrated in 2015-2016 (possible system migration issue?)
- Imputation recommendation:
  * For wage: Forward-fill within worker_id (use prior year's wage + inflation adjustment)
  * For productivity_score: DO NOT impute - may introduce bias. Flag observations and run robustness check excluding them
  * For remote_work: Assume 0 (office work) for missing values if pre-2020, manual review for post-2020

VALIDITY CHECK:
- Out-of-range values:
  * Age = -5 for worker_id 1003 (impossible - likely data entry error)
  * hours_worked = 168 for worker_id 1004 (24 hours/day × 7 days = impossible; likely weekly instead of daily)
  * productivity_score should be 0-100 scale - verify max/min values
- Logical inconsistencies:
  * Worker 1005 appears twice in 2015 (duplicate row - likely merge error)
  * Education should be categorical but appears as text with inconsistent capitalization ("Bachelor" vs "high school")
- Timestamp issues: None detected in sample

CONSISTENCY CHECK:
- Coding inconsistencies:
  * Education: "Bachelor" vs "high school" vs "Master" vs "PhD" - inconsistent capitalization
  * RECOMMENDATION: Standardize to title case or create numeric codes (1=High School, 2=Bachelor, 3=Master, 4=PhD)
- Unit inconsistencies:
  * hours_worked appears to mix daily/weekly reporting (40 is reasonable daily, 168 is weekly)
  * RECOMMENDATION: Check firm-level patterns; may need firm-specific conversion
- Format inconsistencies:
  * remote_work coded as 0/1 but column header says "remote_work" not "remote_work_dummy" - verify interpretation

STATISTICAL FLAGS:
- Outliers:
  * hours_worked = 168 (99.9th percentile)
  * Age = -5 (impossible outlier - ERROR)
  * wage range needs verification (min/max check against industry standards)
- Unexpected distributions:
  * Duplicate row for worker 1005 suggests possible broader duplication issue
  * Check for duplicate worker_id × year combinations across full dataset
- Correlation oddities:
  * Cannot assess from 10-row sample - recommend correlation matrix on full data

RECOMMENDED CLEANING STEPS:

Priority 1 (must fix before analysis):
1. Remove duplicate row: DROP observation where worker_id=1005 AND year=2015 appears twice
2. Fix impossible age: Investigate worker_id=1003 age=-5 (check source data, likely birth year miscoded)
3. Verify hours_worked units: Check with data provider if some firms report weekly vs daily hours
4. Standardize education coding: Create categorical variable with consistent levels

Priority 2 (should fix):
1. Handle wage missingness: Implement forward-fill within worker + inflation adjustment
2. Flag productivity_score missingness: Create indicator variable for missing productivity (use in robustness)
3. Standardize text fields: Convert education to title case before creating categories
4. Run duplicate check: Verify no other duplicate worker_id × year combinations exist

Priority 3 (nice to have):
1. Create derived variables: tenure = year - first_year_observed
2. Add validation flags: Create indicator variables for imputed/corrected values
3. Generate summary statistics by firm: Check for firm-level data quality patterns

DATA QUALITY SCORE: 68/100

Justification:
- Major issues: Impossible values (-5 age), duplicates, unit inconsistencies reduce score significantly
- Missing data: Moderate (~2%) but non-random patterns concerning
- Fixable problems: Most issues are correctable with Priority 1-2 steps
- After cleaning: Expected score would rise to 85-90/100

Interpret the regression results below:

Regression output:

Dependent Variable: log_hourly_wage
Independent Variables:
  years_education: 0.089 (SE: 0.012, p<0.001)
  experience: 0.034 (SE: 0.008, p<0.01)
  experience_squared: -0.0006 (SE: 0.0002, p<0.05)
  female: -0.152 (SE: 0.031, p<0.001)
  remote_work: 0.071 (SE: 0.028, p<0.05)

R-squared: 0.342
N: 2,456
Controls: industry (15 categories), region (5 categories), year fixed effects
Standard errors: Clustered by firm

PLAIN-LANGUAGE SUMMARY:
This analysis examines how education, experience, gender, and remote work affect wages, controlling for industry, location, and time trends. Using data from 2,456 workers, we find that education and experience significantly increase wages, women earn less than men even after controlling for qualifications, and remote work is associated with higher wages. The model explains about one-third of wage variation.

KEY FINDINGS:

Variable: Years of Education
- Effect: Each additional year of schooling increases wages by approximately 8.9%
- Significance: p<0.001 (highly statistically significant)
- Practical meaning: A bachelor's degree (4 years) over high school predicts 35.6% higher wages, holding other factors constant

Variable: Experience
- Effect: Each year of experience increases wages by 3.4%, but this effect diminishes over time (experience squared is negative)
- Significance: p<0.01 for experience, p<0.05 for experience squared
- Practical meaning: Early career wage growth is steeper than mid/late career. Peak wage growth occurs around 28 years of experience (calculated from turning point of quadratic), after which returns to experience decline.

Variable: Female
- Effect: Women earn approximately 15.2% less than men, all else equal
- Significance: p<0.001 (highly statistically significant)
- Practical meaning: This represents a substantial unexplained gender wage gap even after controlling for education, experience, industry, and location. This gap could reflect discrimination, unmeasured differences in job roles, negotiation patterns, or other factors not captured in the data.

Variable: Remote Work
- Effect: Workers with remote work arrangements earn approximately 7.1% more
- Significance: p<0.05 (statistically significant)
- Practical meaning: Remote work access is associated with higher wages. This could reflect higher-skilled workers having remote options, compensating differentials, or productivity effects. Causality unclear from this observational data.

MODEL QUALITY:
- R-squared: 0.342 (34.2% of wage variation explained)
- This is moderate explanatory power. About 66% of wage variation remains unexplained, likely due to unobserved factors like specific job roles, individual productivity, firm-specific wage policies, and measurement error.
- Sample size: N=2,456 is adequate for detecting effects of this magnitude with reasonable precision
- Key controls: Industry (15 categories), region (5), and year fixed effects account for sectoral differences, geographic cost-of-living, and time trends

LIMITATIONS:
What this model CANNOT tell us:
1. Causality for remote work: Association observed, but we cannot conclude remote work causes higher wages (could be reverse: high-wage workers get remote access)
2. Mechanisms for gender gap: Model shows gap exists but cannot identify whether it's discrimination, occupational sorting, hours flexibility, or other unmeasured factors
3. Long-term effects: Cross-sectional data cannot reveal how wages evolve within individuals over careers

Causality: Can we claim causal effects? Why/why not?
- Education and experience: Likely causal, but could have omitted variable bias (ability, family background)
- Gender: Descriptive gap, not necessarily discrimination (need audit studies or within-job comparisons)
- Remote work: Definitely NOT causal - selection bias concerns (who gets remote access?)
- For causal estimates, would need random assignment (experiment) or credible quasi-experimental design (IV, RDD, DiD)

POLICY IMPLICATIONS:
If this model is credible, what actions might it suggest? (2-3 specific recommendations)

1. Education Investment: With 9% wage returns per year of schooling, policies subsidizing education (tuition assistance, student loan programs) have strong economic justification. Community college and vocational training programs may be particularly cost-effective.

2. Gender Wage Gap Investigation: The persistent 15% gap after controlling for qualifications warrants deeper investigation. Policy responses could include pay transparency requirements, salary negotiation training for women, or stronger enforcement of equal pay laws. However, more research needed to identify mechanisms.

3. Remote Work Access Equity: If remote work genuinely increases productivity (and thus wages), unequal access could exacerbate wage inequality. Policymakers might consider: (a) ensuring remote work opportunities are not concentrated among already high-wage workers, (b) investing in broadband infrastructure to enable remote work in underserved areas, (c) studying whether formalizing remote work rights reduces wage disparities.