How we calculated Nepal's rental tax revenue leakage
A transparent account of the data sources, methods, and assumptions behind our estimate of Rs. 5.35 Arba in annual uncollected rental TDS.
Domain-level multiplication is used rather than a national average because Nepal's rental market is too geographically concentrated — in both rent levels and household counts — for any single average to produce an accurate national total. Kathmandu Valley Urban alone accounts for 62.6% of estimated national rental tax potential; averaging it with Madhesh Rural as if they were equivalent would understate the total by Rs. 3.49 Arba.
The Three Data Sources
CBS Data: From Raw Provincial Files to One National Dataset
Scraped 7 provincial HTML tables from the CBS portal. Each file contained local-level household ownership data structured per district and municipality.
Cleaned, flattened, and merged all 7 provincial files into a single nationwide local-level dataset with standardized column structure and district classification.
Aggregated rented household counts by the 15 analytical domains (province × urban/rural), ready for merge with NLSS consumption data.
NLSS Data: From Consumption Survey to Rent Estimates
The Nepal Living Standards Survey publishes average annual household consumption by domain, broken down by spending category. We extracted the rent share for each of the 15 analytical domains.
By multiplying each domain's average consumption by its rent share percentage, we derived the estimated annual rent per household — without relying on any self-reported rent figures.
IRD Data: Official Collection vs. Estimated Potential
The IRD's own annual reports show rental tax collection stagnating at roughly 65% of their internal target — and their target itself is a fraction of true potential.
The IRD set a target of Rs. 4.49 Arba but collected only Rs. 2.90 Arba. Our model estimates a true potential of Rs. 8.26 Arba at 10% TDS — meaning the actual leakage is 3× larger than even the IRD acknowledges.
The Rs. 4,722 figure appearing in our summary is the simple unweighted average of 15 domain monthly rents. It is illustrative only — it was never multiplied by total households to produce our estimate. The actual calculation was performed at domain level: each of the 15 province × urban/rural segments was calculated separately using its own household count and its own rent figure derived from NLSS consumption data for that segment. The 15 results were then summed. Using a single national average in the formula would produce a significant underestimate driven by the outsized weight of Kathmandu Valley Urban, which has both the highest rent (Rs. 14,284/month) and the largest household count (3,01,509) of any domain.
Data Engineering Pipeline
To construct our model, we executed a rigorous consolidation process to map high-level consumption behaviors to granular, local-level housing realities.
The Calculation, Step by Step
Household count
Note: We use the official CBS Census 2078 household count to maintain a highly conservative, undisputed baseline, rather than extrapolated estimates.
Annual rent per household
Note: This is a weighted national average. Kathmandu Valley Urban households pay an estimated Rs. 14,284/month; rural Sudurpaschim households pay an estimated Rs. 2,565/month.
Total annual rental income
Tax potential
Note: We also present the 7.5% scenario (Rs. 6.19 Arba) as a conservative estimate given that effective rates may be lower for smaller landlords.
Revenue leakage
What the IRD's own collection data implies
Working backwards from the IRD's FY 2081/082 collection of Rs. 2.90 Arba: if we assume this collection represents 10% TDS on all declared rental income, the implied total declared rental income nationally is Rs. 29 Arba. Divided across the CBS census count of 8,50,562 rented households, this implies an average declared monthly rent of approximately Rs. 2,841 per household.
The Nepal Living Standards Survey places the actual average monthly rent at Rs. 4,722. In Kathmandu Valley, the average is Rs. 14,284.
Rs. 2,841 is below the cost of a single room in any urban ward of Nepal today. This is not primarily the result of deliberate evasion. It is the result of an economy that has no documentation infrastructure — and therefore no way to declare what it genuinely pays.
Methodological Disclosures & Scope Limitations
This research provides a macroeconomic estimation of rental tax revenue leakage derived from aggregating official national datasets. To maintain statistical transparency and academic rigor, this section explicitly outlines the mathematical assumptions, data boundaries, and structural constraints inherent in the model.
The Asymmetrical Baseline & Market Scope
The Conflation of Commercial and Residential Baselines (Mathematical Asymmetry)
A critical structural limitation of this model is the mathematical asymmetry between the calculated tax potential and the reported tax collection.
- The Potential (Residential Framework): The potential side of the equation (derived from CBS and NLSS data) is constrained strictly to residential household consumption capacity. It does not measure the formal corporate or commercial real estate market (e.g., banks, dedicated retail towers, industrial leases).
- The Collection (Blended Reality): The Inland Revenue Department’s (IRD) reported FY 2081/082 rental tax collection of Rs. 2.90 Arba is a blended federal figure. It aggregates Tax Deducted at Source (TDS) swept from both corporate commercial leases and formal residential contracts.
By subtracting a blended (commercial + residential) collection from a purely residential potential, the equation mathematically compares overlapping but distinct economic bases. Because commercial collections represent a significant portion of the IRD's Rs. 2.90 Arba pool, subtracting this blended number from the residential potential artificially compresses the calculated deficit. Therefore, the resulting leakage calculation represents a highly conservative, minimum macroeconomic floor rather than a perfectly isolated sector analysis.
Mixed-Use Conflation and Classification Ambiguity
While the CBS Census 2078 structurally targets residential "households," it suffers from inherent classification ambiguities due to the high density of mixed-use real estate in Nepal’s urban centers. A significant volume of rented properties in municipalities function simultaneously as dwellings and commercial micro-enterprises (e.g., ground-floor retail shops, home-based workshops, or local clinics with attached living quarters).
Because census enumerators record these under a single household classification based on occupancy, our baseline count of 850,562 "rented households" inadvertently captures an unquantifiable volume of commercial activities. If a portion of these households are generating commercial revenues rather than strictly residential consumption, the statutory tax rate applied would shift from residential thresholds to corporate/commercial TDS brackets. This introduces a dual-directional variance: it may slightly overstate the purely residential count, but it simultaneously understates the total revenue potential, as commercial rental values command a massive premium over residential market rates.
Exclusion of Legal Thresholds and Variable TDS Scenarios
The baseline calculation does not explicitly account for low-income tax exemptions, localized municipal rebates, sub-letting friction, high-density communal sharing (e.g., multiple students or laborers informally sharing a single rented room under a non-contractual verbal agreement), or progressive tax brackets for high-volume landlords.
To compensate for this legal and policy variance, the model introduces an interactive sensitivity analysis via a variable Tax Deducted at Source (TDS) threshold parameter (5%, 7.5%, and 10%):
- The 10% Scenario: Represents the standard statutory federal TDS rate typically applied to formal institutional or corporate tenancies under the IRD framework.
- The 7.5% and 5% Scenarios: Act as discount parameters to simulate lower effective tax rates. These lower bounds account for individual-to-individual residential leases, smaller rental agreements that fall under local municipal jurisdictions, or cases where the effective tax burden is mitigated by legal standard deductions.
Consequently, a portion of this calculated "leakage" represents economic activity that practically falls outside the formal regulatory reach of the IRD, meaning the legally enforceable tax base is narrower than the total mathematical deficit presented.
Data Distribution & Proxy Variables
Consumption Proxy and Hedonic Imputation (NLSS-IV Alignment)
The rental share variable utilized in this model is extracted from the Nepal Living Standards Survey IV (2022/23). According to the National Statistics Office's official methodology (Section 11.2.5), the NLSS estimates the "flow of housing services" by utilizing a national hedonic regression model to impute rental values for non-renters, blending observed market rents with self-assessed valuations.
Because this national accounting method measures the total macroeconomic utility of housing consumption—including imputed rent for non-cash arrangements (e.g., rent-free arrangements with relatives, corporate housing, or caretakers)—our mapping of NLSS consumption data to IRD taxable revenue generates a theoretical maximum tax capacity. It inherently captures a broader economic base than the strictly realized, contractual cash transactions that the IRD is legally mandated to tax.
Statistical Skewness (Mean vs. Median)
This model utilizes the arithmetic mean derived from domain-specific consumption aggregates, as raw household-level microdata was unavailable to calculate the exact median. In real estate economics, housing expenditure typically follows a highly right-skewed (log-normal) distribution, where high-value luxury rentals disproportionately pull the arithmetic mean upward. Applying the mean uniformly across all households within a domain likely introduces an upward bias in the total rental potential for low-and-middle-income renters. Readers should treat the calculated potential as a synthetic macroeconomic capacity rather than an exact representation of the median renter's financial reality.
Temporal Asynchrony
The predictive engine relies on datasets collected across different time horizons: the structural household volume is baseline-anchored to the CBS Census 2078 (2021), consumption ratios are derived from the NLSS IV (2022/2023), and collection metrics are pulled from the IRD FY 2081/082 (2024/2025). This temporal gap fails to mathematically account for intermediate urban migration spikes, localized real estate shocks, or real-wage inflationary adjustments over this three-year window. Conversely, the official CBS census count is known to under-count actual renter volume compared to consumption surveys, which acts as a counter-balancing conservative factor.
Institutional Context
Alignment with IRD’s Own Findings (Section 1.4.5)
Despite the structural constraints of utilizing proxy data, the underlying conclusion of the model aligns directly with the government's own institutional assessment. The Inland Revenue Department’s Annual Progress Report FY 2081/082 (Section 1.4.5) explicitly acknowledges the systemic failure in this sector, stating:
"...घर जग्गा बहालमा हुने न्यून घोषणाले राजस्व परिचालनको थप सम्भाव्यतामा नकारात्मक असर परेको छ।"
(...Under-declaration in house rent negatively impacts further revenue potential).
Jurisdictional Fragmentation (Schedule 8)
A primary driver of the data opacity in this sector is regulatory fragmentation. Under Schedule 8 of the Constitution of Nepal, basic house rent tax collection falls under Local Government jurisdiction (Municipalities/Wards), while federal corporate rental TDS is handled by the IRD. The absence of a unified digital registry between federal and municipal bodies prevents exact data reconciliation, necessitating the use of the macroeconomic estimation models presented in this research.
The limits of this estimate
This estimate is a macroeconomic baseline, not an exact accounting audit. It relies on consumption surveys that include 'imputed rent' (the economic value of non-cash housing arrangements) and uses arithmetic means, which may skew higher due to luxury rentals. Furthermore, there is a structural asymmetry in the data: while the calculated tax potential focuses strictly on residential households, the subtracted actual tax collection (Rs. 2.90 Arba) is a blended figure that includes commercial corporate leases. It also assumes a uniform statutory TDS rate and does not explicitly filter out mixed-use commercial properties, informal communal living arrangements, or municipal tax exemptions. However, because it subtracts blended commercial collections from a strictly residential baseline, and relies on the official CBS census—which is known to under-count actual renter volume—the resulting Rs. 5.35 Arba figure serves as a highly defensive, minimum macroeconomic floor for the true systemic revenue deficit.