Google Community Mobility – Brazil Daily Report (CSV)

Rows: 1,095
Columns: 15

Daily mobility percent change from baseline for Brazilian states and municipalities across six venue categories during 2021. Key dataset for social science and epidemiological time-series modeling in Latin America.

AI analysis

Google Community Mobility – Brazil Daily Report (2021): Analytical Overview

1. Dataset Representation and Research Value

This dataset captures Google's Community Mobility Reports for Brazil throughout the full calendar year 2021, sourced directly from Google's public static hosting infrastructure at https://www.gstatic.com/covid19/mobility/2021BRRegionMobilityReport.csv — a publisher with high credibility and consistent methodology across its global mobility series. The data encodes daily percent-change-from-baseline figures across six venue categories (retail, grocery/pharmacy, parks, transit stations, workplaces, and residential) for Brazilian states and municipalities. With 1,095 rows spanning 365 days across three geographic granularities (national, state, and sub-regional), this is a structurally compact but analytically rich time-series resource. Its primary value lies in social science and epidemiological modeling: mobility shifts in 2021 directly reflect the interplay of COVID-19 wave dynamics, vaccination rollout, and policy interventions in Latin America's largest nation, making it especially potent for causal inference and interrupted time-series studies.

2. Data Quality Observations

The dataset contains 4,039 null cells across 15 columns, but the distribution of missingness is highly concentrated and largely structural rather than indicative of data corruption. Three columns — subregion2, metroarea, and censusfipscode — are 100% null (1,095 of 1,095 rows each), which strongly suggests these granularity levels were either not reported for Brazil in 2021 or were excluded during the API export. These columns should be dropped immediately before modeling to avoid dimensionality noise. An additional 365 nulls appear in subregion1 and iso31662code, consistent with the national-level rows where state identifiers do not apply — this is expected and interpretable, not a flaw. The transitstationspercentchangefrombaseline column carries 24 nulls, the only mobility metric with missing values, which warrants imputation or exclusion depending on analysis context. Duplicate row counts are pending Phase D recomputation, but the clean placeid distinct count of 3 (likely national + ~2 state aggregates) confirms the dataset's deliberately narrow geographic scope.

3. Key Column Distributions and Anomalies

The six mobility metrics reveal sharply divergent behavioral patterns worth highlighting individually. Grocery and pharmacy mobility shows the highest mean (+37.6%, median +40), with a left-skewed distribution (skew = -0.769) and a striking maximum of +129% — indicating significant demand surges, likely tied to panic-buying or lockdown-adjacent stockpiling episodes. Transit stations is the statistical wildcard: with a mean of +34.3 but a median of only +6, a standard deviation of 67.1, a skewness of +1.22, and a maximum of +228%, this column contains the most extreme values and 33 flagged outliers — far more than any other metric. This bimodal-like behavior likely reflects the contrast between COVID restriction periods and post-restriction rebounds. Retail and recreation paints the starkest contraction picture (mean = -11.8, min = -81), with 26 outliers clustering at the negative tail, while residential mobility is the tightest and most stable metric (σ = 3.35, range 0–26, only 24 distinct values), serving as a near-inverse proxy for out-of-home activity. Workplaces show a negative skew (-1.06) with 28 outliers, suggesting several sharp work-from-home event clusters punctuating an otherwise moderate positive mean (+3.8).

4. Recommended Join Key Columns

For cross-dataset correlation, the most reliable join keys are date (Date type, 0 nulls, 365 distinct values — a clean daily temporal anchor) and iso31662code (0 nulls at the state level, with 2 distinct codes representing Brazil's state-level granularity). The placeid column (3 distinct values, 0 nulls) can serve as a geographic entity identifier for joining against other Google-ecosystem datasets or geospatial lookups. countryregioncode ("BR", 0 nulls, 1 distinct) is useful only for multi-country joins. For any sub-regional enrichment, subregion1 is the only viable administrative-boundary key, though its 365 nulls (the national rows) require a filtering step before use. Analysts should index on date + iso31662code as a composite key for state-level longitudinal joins.

5. Recommended Paired Datasets for Correlation Discovery

Several dataset categories would pair powerfully with this mobility data. Brazil COVID-19 case and death time-series (e.g., from Brasil.IO or the Ministry of Health) joined on date would enable Granger causality testing between mobility drops and case-rate lagged responses — a canonical epidemiological use case. Brazil vaccination rollout data (doses administered by state and date) could reveal whether mobility rebounded post-vaccination threshold crossings, testable as a structural break in the retail and transit series. Brazil weather or climate data (temperature, precipitation by state) joined on date + iso31662code would help disentangle pandemic-driven mobility suppression from seasonal effects, particularly for the parks metric (skew = +0.339, high variance). Consumer sentiment or retail sales indices from IBGE or Banco Central do Brasil would validate whether the grocery/pharmacy surge translated to economic throughput. Finally, social media activity or internet traffic datasets for Brazil in 2021 could serve as behavioral triangulation, testing whether residential mobility increases correlate with digital consumption spikes — a rich opportunity for multi-modal social science inference.

Columns

date (date)
country_region_code (string)
country_region (string)
sub_region_1 (string)
sub_region_2 (string)
metro_area (string)
iso_3166_2_code (string)
census_fips_code (string)
place_id (string)
retail_and_recreation_percent_change_from_baseline (integer)
grocery_and_pharmacy_percent_change_from_baseline (integer)
parks_percent_change_from_baseline (integer)
transit_stations_percent_change_from_baseline (integer)
workplaces_percent_change_from_baseline (integer)
residential_percent_change_from_baseline (integer)