Readme for Historical WFH variables: Summary: This series describes the share of full days worked from home by employed individuals ages 20-64 who reported earnings greater than $20,000 2019 dollars. For 1965 and 1975 we use data from the American Historical Time Use survey. Individuals report their activities for a 24 hour period and we mark them as fully working from home if they are paid to WFH for at least 6 hours in the day. For 1980, 1990, and 2000-2019 we use data from the American Community Survey. Individuals report their usual transportation to work last week. If they report WFH, we mark them as fully working from home. From May 2020 to present, we use data from the Survey of Working Arrangements and Attitudes. Individuals report the days they worked a full day at home. To calculate real earnings we use PCEPI 2019 = 100. AHTUS overall: Years: 1965, 1975, 1985, 1993, 1995, 1998, 2003-2012, 2018 Universe: 1965 is 19 +, 1975 is 18 +, more recent is 15+. Earlier samples are 2,000-4,000. More recent is around 12,000. Weight with: "Recwght" which is their recommended sample weight. This weight accounts for population / sample distribution by age group, sex, distribution of days of the week, and good quality diary entries. Sampling scheme: For the 1965 sample (Multinational Comparative Time Budget Research Project), sampled individuals 19-65 from Jackson, Michigan and 44 other US cities that had between 30,000 and 280,000 residents. They randomly selected one household member to complete a time diary of one day from midnight to midnight. All days of the week were given an equal probability of selection for diary days. For the 1975 sample (Time Use in Economic and Social Accounts, funded by the NSF), sampled individuals 18 years or older across the US in the 74 standard MSAs. These individuals recorded a time diary of 24 hours for 4 waves across a 12 month period. (ie an individual could have a 24 hour diary for 10-10-1975; 02-06-1976; 05-02-1976; 09-04-1976). When applicable they would also record their spouse's activities. 1,324 respondents completed all 4 diarys. Attempted to sample so each individual completed a time diary for Saturday, Sunday, and two weekdays. Note that I treat each waves as unique responses. Other variables available: Age, sex, highest education level, number of children in household, age of youngest child, employment status Earnings concept: In the 1975 sample there is an "Employment income from last month". In both the 1965 and 1975 samples there is the approximate quartile of the total household income. AHTUS WFH variable for our series: Years: 1965, 1975 Universe: Restrict to 20-64 year olds who are employed who earn greater than $20,000 in 2019 dollars. I further restrict to diary entries from Monday - Friday. This limits the 1965 sample to 889 respondent and the 1975 sample to 410 respondents. Question: They list their activity for a specific time period as "Paid work at home." I further restrict to individuals who work from home 6 or more hours in a given day. Weight with: "Recwght" which is their recommended sample weight. This weight accounts for population / sample distribution by age group, sex, distribution of days of the week, and good quality diary entries. Construction of WFH variable: If an individual is 20-64 and works from home at least 6 paid hours, they are labeled as 100% for WFH. Else, they are given a 0%. I then take the mean over the sample. ACS overall: Years: 1980, 1990, 2000-2019 Universe: Individuals age 0 - 97. 1980, 1990 samples have 11,000,000 respondents. 2000 sample has 370,000 respondents. 2001 - 2019 samples have 1,000,000 - 3,300,000 respondents. Weight with: "Perwt" which is their weight to indicate how many persons in the US are represented by a given person. It is not clear what variables this weights by, ie age, gender, race, etc. Sampling scheme: 1980 5% State sample, 1 in 20 national random sample of the population. 1990 5% State sample, 1 in 20 national random sample of the population. 2000 ACS weighted sample, 1 in 750 national random sample of the population. Nearly identical to 2000 census long form. 2001 ACS weighted sample, 1 in 232 national random sample. 2002 ACS weighted sample, 1 in 261 national random sample. 2003 ACS weighted sample, 1 in 236 national random sample. 2004 ACS weighted sample, 1 in 239 national random sample. 2005 - 2019 ACS weighted sample, 1 in 100 national random sample. Other variables available: Age, sex, race, hispanic origin, educational attainment, employment status, age of youngest child, occupation, industry Earnings concept: For all samples reports the respondent's total pre-tax wage and salary income for the previous year. ACS WFH variable for our series: Years: 1980, 1990, 2000-2019 Universe: Restrict to 20-64 year olds who are employed who earn greater than $20,000 in 2019 dollars. 1980 sample has 2,800,000 respondents. 1990 sample has 3,500,000 respondents. 2000 sample has 120,000 respondents. 2001 - 2004 samples have 340,000 respondents. 2005 - 2014 samples have 900,000 respondents. 2015 - 2019 samples have 1,000,000 respondents. Question: From 1990 - 2019 was “How did this person usually get to work last week?” and one of the options was “Worked at home”. In 1980 worked at home was not an option, but interviewers were asked to mark worked at home if the respondent worked on a farm where they lived or in an office or shop in the respondent’s home. Weight with: "Perwt" which is their weight to indicate how many persons in the US are represented by a given person. It is not clear what variables this weights by, ie age, gender, race, etc. Construction of WFH variable: If an individual is 20-64 and "usually" worked from home last week, they are labeled as 100% for WFH. Else, they are given a 0%. I then take the mean over the sample. SWAA WFH variable for our series: May 2020 - present. Percent of full days worked from home.