Michigan

The American Community Survey's 5-year estimate of the number of people living in the area (ACS B01003), read at each geographic level directly rather than summed up from smaller areas. Served as-is with no further calculation; areas with zero recorded population are dropped from the data entirely.

The American Community Survey's 5-year median age of all residents in the area (ACS table B01002), read straight from the ACS for that geography with no aggregation or derived arithmetic. Census sentinel codes for unavailable, not-applicable and suppressed values are set to null, and areas with no population are dropped.

Total population divided by the area's land area in square miles, using the land area recorded on the Census TIGER boundary file (which excludes water).

Median annual real estate tax paid on owner-occupied housing units, taken straight from ACS table B25103 with no further computation. ACS top-codes this median at 10001, meaning $10,000 or more, and that code is served unchanged.

The median annual real-estate tax bill across owner-occupied homes with and without a mortgage (ACS B25103), divided by the median home value (ACS B25077). Left blank when either median is a bound ACS publishes in place of a figure -- a home value of $2,000,000 or more or under $10,000, or a tax of $10,000 or more -- and also when the tax figure is a literal zero.

Median home value (ACS B25077) divided by median household income (ACS B19013). Left blank wherever ACS censored either input, meaning home values published as "$9,999 or less" or "$2,000,000 or more" and incomes published as "$250,000 or more", because a ratio of two capped medians is not a measurement.

Median monthly gross rent, taken directly from ACS B25064 with no further calculation. Values at the top of the distribution arrive as the ACS top-code 3,501 rather than a true median, and that code is served as reported rather than blanked out.

FHFA's published one-year percent change in the House Price Index for this area, taken as-is from FHFA's annual HPI file for the matching geography (tract, ZIP, county or state) rather than recomputed here. Juniper reads the most recent annual data year available in the pipeline.

The highest FHFA house price index reading in 2022-2023 minus the lowest reading in 2019-2020, divided by that 2019-2020 low, times 100.

The highest FHFA house price index reading in 2005-2007 minus the lowest reading in 2009-2012, divided by that 2005-2007 peak, times 100. A negative value means the 2009-2012 low was still above the pre-crash high.

The 1991-2020 average daily high temperature across June, July and August, from the NOAA station nearest the tract or ZIP centroid that reports it, and left empty when no such station lies within about 200 km. County values are population-weighted averages of their tracts, and state values population-weighted averages of their counties.

The 1991-2020 NOAA normal of daily low temperature over December, January and February, read from the nearest station that measures it to the tract or ZIP centroid, and left blank when no such station lies within about 200 km. County values are population-weighted averages of their tracts and state values population-weighted averages of their counties, in both cases counting only the units a station actually covers.

The 1991-2020 annual average daily temperature normal from the NOAA weather station nearest the tract or ZIP centroid that reports temperature, matched separately for each climate metric so different metrics can draw on different stations. An area with no reporting station within 200 km is left blank; county values are population-weighted averages of their tracts, and state values are population-weighted averages of their counties.

NOAA's 1991-2020 normal for the average number of days a year with a daily low below 32 degrees Fahrenheit, taken from the nearest weather station reporting that measurement to the tract or ZIP centroid, and left blank when no such station lies within about 200 km. County values are population-weighted averages of their tracts and state values of their counties, counting only the areas that have a reading.

NOAA's 1991-2020 normal for total yearly precipitation, in inches, read from the nearest station that measures precipitation to the tract or ZIP centre; no such station within about 200 km leaves the value blank rather than zero. County figures are population-weighted averages of their tracts and state figures of their counties, in each case averaging only the areas that carry a value.

The 1991-2020 average total yearly snowfall in inches, from the NOAA weather station nearest the area's center that actually measures snowfall; a station that does not measure snow is skipped rather than read as zero, and an area with no snow-measuring station within about 200 km is left blank. County values are population-weighted averages of their tracts, and state values population-weighted averages of their counties.

FEMA National Risk Index modelled expected annual loss from coastal flooding, in dollars, divided by the same area's FEMA building-value exposure expressed in thousands of dollars; an area with no reported building value is left blank rather than shown as zero. This covers coastal flooding only, not inland or riverine flooding, and county figures are population-weighted averages of their tract values while state figures are population-weighted averages of their county values, counting a place where FEMA models no coastal flood loss as zero rather than as missing.