Economics · Finance · Data analysis

Prathamesh
Singh.

I’m an economics graduate from the University of Houston. I use Python and SQL to investigate business costs, compare financial risks and test forecasts against simple alternatives.

Explore my work Download résumé

Houston, Texas · Open to analyst opportunities

From the portfolioConsumer finance / 2026

When are mortgage prices
actually comparable?

A Houston study checks which borrower groups are represented across five lenders.

41.38%

of those loans share common support.
That limits what a lender comparison can tell us.

Source: HMDA 2025. Matching uses available public borrower and loan characteristics. Price differences remain descriptive, with unobserved differences unresolved.
Explore the mortgage study

Selected work

Selected projects

Independent projects in finance, economic research and operations. Each includes the source data, code and limitations.

Banking / Cost analysis

Why did bank deposit
costs fall?

Across 4,018 screened banks, quarterly deposit interest expense fell by $5.01 billion. The analysis separates changes in balances, interest-bearing mix and a derived rate measure.

A finance team can use the breakdown to frame a pricing review before making assumptions about customer behavior.

Read the case study
Accounting bridgeQ2 2025 to Q2 2026
Deposit expense bridge in billions of dollars: balance growth plus 5.24, interest-bearing mix plus 0.06, proxy-rate effect minus 10.31, total change minus 5.01.
Balance growth
+$5.242B
Interest-bearing mix
+$0.057B
Proxy-rate effect
−$10.310B
Total change
−$5.012B
FDIC data. Screened matched cohort. An accounting attribution does not establish a causal pricing effect.

Energy / Data quality

The newest dataset
wasn’t ready to use.

Twenty-six state summaries in the 2025 EIA reliability release failed reconciliation. The analyst workbench keeps the discrepancies visible and uses the reconciled 2024 release for comparisons.

2023Pass
2024Pass
2025Held back

51 available state summaries checked per year. The predictive model also failed its promotion gate.

Inspect the reliability work

Operations / Forecasting

A better rental forecast.
A planning tradeoff to test.

The selected model reduced historical test error from 18.70% to 13.35%. The next question is how forecast errors affect capacity when shortages and excess capacity have different costs.

Actual daily rentals and selected ridge forecasts across 184 historical test days from July to December 2012.
UCI Bike Sharing. Historical test, not a current demand forecast. Planning costs are hypothetical.
See the forecast and planning analysis

Research note

A recession model that lost to its baseline.

Adding credit spreads improved on a Treasury-curve model, but a simple historical probability still performed better. The candidate was not promoted.

Read the model audit

Explore the full collection, including insurance, real estate and working capital.

All 19 projects

A little about me

Economics at Houston.
Questions across industries.

I earned my B.S. in Economics at the University of Houston, with a minor in Entrepreneurship. My coursework included economic data analysis, behavioral economics, money and banking, and energy economics.

Those interests carry into my projects. I work through financial disclosures, compare lending markets and test operational forecasts. I’m looking for an analyst role where I can develop that work with a team and help explain the findings to the people making decisions.

PythonSQLStataExcel

Python and SQL are demonstrated in the linked projects. Stata and Excel are listed on my résumé.

Contact

Have a role or a question
about the work?

LinkedIn ↗GitHub ↗Résumé ↓