The unit of analysis is the individ

The unit of analysis is the individual in a household. First, data of the entire residential household types in Singapore were obtained from the Department of Statistics, Singapore in 2000. According to the Department of Statistics (2000), Housing Development Board (HDB) dwellings (1-, 2-, 3-, 4-, 5-room or larger) accounted for 79.14 percent (763,036 units), private dwellings (bungalow, semi-detached and terrace houses, condos and apartments) accounted for 14.18 percent (136,719 units), while other housing types accounted for 6.68 percent (64,383 units).
Quotas were then established accordingly to determine the number of the housing types that should be sampled. The households were divided into private housing and public housing units. There are some housing units that do not fall into the above categories but these are negligible in number (6.68 percent) and scattered around the island. It would not be economical to include them in the sampling frame. Since the quotas were established based on public and private housing units only, the valid percentages of breakdown for public and private housing units were 84.8 percent and 15.2 percent, respectively.
Next, a street directory (Singapore Street Directory, 20th Edition, published by Ministry of Law, Singapore, January 2000) was used to determine the locations of the housing type.
There are three types of public dwellings in the Street Directory, namely HDB New Towns, HDB Estates and HDB Precincts. Systematic sampling method was used for the selection. From the alphabetical list of the 20 HDB New Towns, four were randomly selected based on an interval of five (Bedok, Clementi, Queenstown and Tampines). From the alphabetical list of 62 HDB Estates, three were randomly chosen based on an interval of 17 (Fengshan, Nanyang and Taman Jurong). From the alphabetical list of 199 HDB Precincts, five were randomly selected based on an interval of 40 (Bedok Reservoir Garden, Compassvale Place, Kaki Bukit Ville, Rivervale Court and Teck Whye Heights 2).
All blocks in these selected public housing areas were tabulated based on the maps in the latest Street Directory. Stratified sampling method was employed for the selection of five blocks from each of these 12 public housing areas (new towns, estates and precincts). At the individual unit level, a list of all units in each selected block was tabulated. Homes were later selected based on an interval of 10 with the first home selected on the random last digit of the license plate of a passing vehicle. By doing so, the researchers could avoid over-representation of certain room types in the sampling process. For example, apartments with certain unit numbers at all levels in a block are of 5-room type, where the rest are of 4- or 3-room type. A total of 2,120 mail drops (84.8 percent of the total 2,500) were sent to these randomly selected units. On average, 36 mail drops were sent to each block.
A list of private housing (condos, apartments and private houses) was also available in the Street Directory. From the list of 1,972 private housing estates (in alphabetical order), 19 were randomly selected based on an interval of 100. Admittedly, there are problems with this as the number of units in a condo varies. There is therefore some sampling error. However, as with houses, the numbers would be few. At the individual unit level, homes were selected based on an interval of 10 with the first home selected on the random last digit of the license plate of a passing vehicle. There were 380 mail drops sent to these estates. On average, 20 mail drops were sent to each private housing estate. Recipients were invited to do the survey based on the person with the birth date closest to the day they received the mail drop and who is also an Internet user.