Overview

Urban noise is one of the most common — and most unequally distributed — environmental stressors in large cities. This project analyzes over 2,498 census tracts across Los Angeles County to answer a deceptively simple question: where do noise complaints cluster, and why do they occur where they do?

Using 311 service request data from 2022–2023, demographic variables from the American Community Survey, land-use indicators, and the CDC Social Vulnerability Index, the analysis applies a full spatial econometrics workflow — from exploratory mapping to LISA cluster detection to three regression models — to identify the socioeconomic and built-environment drivers of noise complaint density across LA neighborhoods.

Are noise complaints significantly clustered across Los Angeles, and how do socioeconomic and land-use characteristics shape these clusters? The answer turns out to be yes — with extremely strong spatial dependence that ordinary regression cannot capture.

2,498Census tracts
analyzed
0.74Spatial lag ρ
(SLM model)
0.76Spatial error λ
(SEM — best fit)
3Regression models
compared

Noise Rate Distribution

Noise complaint rate histogram

Distribution of noise complaint rates per 1,000 residents across LA census tracts — right-skewed with a long tail of high-complaint areas.

Boxplot of noise complaint rate

Boxplot of tract-level noise complaint rates — median, spread, and outliers across the 2,498-tract study area.

Live StoryMap

The full analysis — maps, scatterplots, model outputs, and spatial cluster visualizations — is presented as an interactive ArcGIS StoryMap.

StoryMap preview
ArcGIS StoryMap — Public
Understanding the Geography of Noise
Los Angeles County · 311 Data · Spatial Econometrics · December 2025
Open StoryMap →

Data Sources

Analytical Workflow

Global Moran's I — Spatial Autocorrelation

The Global Moran's I test revealed extremely strong positive spatial autocorrelation in noise complaint rates across LA census tracts — complaints are not randomly distributed, they cluster significantly in space.

Moran's I scatterplot

Moran's I scatterplot — steep positive slope confirms that tracts with high complaint rates are surrounded by other high-rate tracts.

Median income vs noise rate

Median household income vs. noise complaint rate — lower-income tracts systematically bear higher noise complaint burdens.

Interactive Map — Noise Rate by Tract

Explore the normalized noise complaint rate across all 2,498 LA County census tracts. The map reveals the sharp spatial gradient between the dense urban core and suburban periphery.

● ArcGIS — Noise Rate by Census Tract

LISA — Local Spatial Cluster Detection

Local Moran's I (LISA) identified statistically significant hotspot and coldspot neighborhoods. Results confirm that noise inequality in LA is not random — it maps onto well-known patterns of socioeconomic stratification.

High–High Clusters

Central LA — Koreatown, Westlake, Pico-Union, Downtown. Dense multifamily housing, renter-heavy populations, high complaint density surrounded by similarly high tracts.

Low–Low Clusters

Affluent or suburban tracts — the Valley, Palos Verdes, hillside neighborhoods. Quiet areas surrounded by similarly quiet neighbors.

High–Low Outliers

Transitional areas at the edges of hotspot clusters — tracts with elevated complaints surrounded by lower-rate neighbors.

Low–High Outliers

Quiet tracts embedded within noisier surroundings — spatial anomalies highlighting localized noise management or underreporting.

Interactive Map — Noise Rate vs. Median Income

This bivariate map overlays noise complaint density with median household income, making the environmental justice dimension of noise distribution directly visible.

● ArcGIS — Noise Rate × Median Income

Predictor Relationships

Bivariate scatterplots showing the relationship between noise complaint rate and each socioeconomic predictor variable used in the regression models.

Noise rate vs % renters

Noise rate vs. % renters — renter-heavy tracts show consistently higher complaint density, significant in OLS and SLM.

Noise rate vs poverty rate

Noise rate vs. poverty rate — positive relationship confirms that lower-income tracts bear a disproportionate noise burden.

Noise rate vs % young adults

Noise rate vs. young adults (20–34) — negative relationship across all three models; younger neighborhoods report fewer complaints.

Regression Models — OLS → SLM → SEM

OLS regression established significant predictors but residual analysis showed systematic spatial clustering — violating OLS assumptions and requiring spatial econometric models.

OLS
Ordinary Least Squares
Baseline — spatial residuals persist

Identifies significant predictors but violates the independent residuals assumption. Residuals show systematic spatial patterns.

SLM
Spatial Lag Model
ρ = 0.7404 (p < 2.2e-16)

Adds a spatially lagged dependent variable. Strong ρ confirms that a neighborhood's noise rate is strongly predicted by its neighbors'.

SEM
Spatial Error Model
λ = 0.7620 (p < 2.2e-16)

Models spatial dependence in the error term. Best AIC fit among all three models. ✓ Best fit

Variable Significance Across Models

VariableDescriptionOLSSLMSEMDirection
pct_renters% households renting✓ Sig✓ Sig+
pct_poverty% below poverty line✓ Sig+
pct_young% adults age 20–34✓ Sig✓ Sig✓ Sig
med_incomeMedian household income✓ Sig
pct_multiunit% multi-family housing✓ Sig✓ Sig

Model Diagnostics & Residual Analysis

OLS residuals vs fitted values

OLS residuals vs. fitted — non-random pattern confirms spatial dependency, motivating the move to spatial econometric models.

AIC comparison across models

AIC comparison — SEM achieves the lowest AIC, confirming that spatial error structure explains noise patterns better than pure spillover (SLM) or no spatial modeling (OLS).

SEM coefficient plot

SEM coefficient plot — direction and significance of each predictor after accounting for spatial error dependence. Income effect becomes clearly negative.

Residual distribution comparison

Residual distribution comparison across OLS, SLM, and SEM — SEM residuals most closely approximate a normal distribution, validating model fit.

Key Findings

The High–High LISA clusters — concentrated in Koreatown, Westlake, Pico-Union, and Downtown — are predominantly lower-income, renter-heavy, and minority neighborhoods. This reflects a well-documented pattern of environmental burden falling disproportionately on communities with the least political and economic power to address it.

Policy Implications

Targeted Noise Mitigation

Direct noise enforcement resources to High–High LISA cluster neighborhoods — Koreatown, Westlake, and Pico-Union — where complaint density and spatial spillover are both highest.

Housing Insulation Standards

Improve acoustic insulation requirements in renter-heavy, multifamily housing tracts. Noise burden in dense residential areas is partly a structural problem addressable through building codes.

Environmental Justice Framing

Treat noise complaint density as an environmental justice indicator. 311 analytics can serve as an early warning system for neighborhood stress before more severe decline patterns emerge.

311 as Urban Sensing

Use spatially modeled 311 data — not raw counts — as input to urban monitoring systems. Normalizing by population and accounting for spatial dependence transforms complaint data into a meaningful policy signal.

Future Work

Three directions would strengthen this analysis. First, integrating land-use zoning and night economy data (bars, venues, commercial density) would sharpen the built-environment predictors. Second, multi-year time-series spatial models would reveal whether noise inequality is growing or shrinking. Third, comparing 311 patterns against actual decibel readings from street sensors would clarify how much of the spatial signal reflects real acoustic exposure versus differential reporting behavior across communities.