######################## Goal ######################## The following request is for HES Net Worth data from the years ended June 2015 and 2018. By and large, these survey years are very similar and, as such, the data should be comparable across this time series. Estimates on the number of people who own property (aside from their principal residence) by demographics including: - capital value (full value of property) - equity on property (total value of property minus mortgages) - owned capital (the percentage of the full value owned by individual in question) (median across demographic variables) - owned equity on property (the percentage of the equity owned by individual in question) (median across demographic variables) broken down by number of properties owned and: • By full or partly owned (suggest fully owned all properties, owned more than 75% of all properties, owned more than 50% of all properties, owned up to 50% of all properties) • By property type (Principal residence, Holiday homes excluding timeshares, Timeshares, Residential investment (rental) real estate, Other non-investment residential real estate, Non-residential real estate (including commercial), Land only) Note that this covers all types of property, including principal residence, farms, agricultural land, commercial residential property etc. • Personally owned, in a trust, in a business • By region (noted this is of the respondent, not where the property necessarily is) • By age • By total response ethnicity Note all tables includes all people with NZ-located property, regardless of how much property they have. Tables are output with a variable stating the number of properties owned, which can be used to exclude those with only a primary residence. These tables exclude property held overseas and mortgages on properties overseas. A person is counted as being an owner of a property if they had some portion of the asset listed under their id, regardless of whether that was held in a trust, business, or personally. In instances where there are multiple mortgages on one property, these mortgages get aggregated for the purpose of calculating net equity. Note mortgages on overseas properties won't be counted in this data. ######################## Challenges ######################## The data is structured in a way that one individual's share in a property is one row of the dataset. Then their share of the mortgage is another row of the dataset. To calculate some of the above figures, I need to link this mortgage value to the property value. There is a propertyid which is very useful for linking these together, however it is not always populated. There are also three different datasets that could include property information - the business, wealth and trust datasets. A person could have a property listed in any of these datasets, with the mortgage in another of the datasets. Alternatively, a person could have the asset and/or mortgage value split across the three datasets. In addition to this, property ownership can be split between multiple owners within a household, or even, in complex cases, split across multiple households. ######################## General Approach ######################## Rather than only including individuals who have more than 1 property, I've included everyone and assigned them into groups for how many properties they have. Note that for the purposes of these notes, personal properties or mortgages refer to properties or mortgages that are not owned via a trust or a business. #### Part 1 #### 1. Take all personal property data and try to link to the personal mortages data (first by personid, propertytype and propertyid, then by just personid and propertytype - the second attempt is due to many propertyids not being populated) 2. Take the remaining mortgages and try to link to property assets in the trust and business datasets 3. Should end up with a. one file that has all personal properties, some linked to wealth mortgages b. one file with properties in trusts linked to privately held mortgages c. one file with properties in business linked to privately held mortgages d. one file of private mortgages that could not be linked to any property #### Part 2 #### 1. Take all trust property data and try to link to the trust mortages data (first by personid, propertytype and trustassetid, then by just personid and propertytype - the second attempt is due to many trustassetids not being populated) 2. Take the remaining mortgages and try to link to property assets in the personal and business datasets 3. Should end up with a. one file that has all properties held in trust, some linked to mortgages in trust b. one file with properties in personal wealth data linked to mortgages owed by trust c. one file with properties in business linked to mortgages owed by trust d. one file of trust mortgages that could not be linked to any property #### Part 3 #### 1. Take all business property data and try to link to the business mortages data (first by personid, propertytype and trustassetid, then by just personid and propertytype - the second attempt is due to many trustassetids not being populated) 2. Take the remaining mortgages and try to link to property assets in the personal and trust datasets 3. Should end up with a. one file that has all properties held in businesses, some linked to mortgages owed by businesses b. one file with properties in personal wealth data linked to mortgages owed by businesses c. one file with properties held in trust linked to mortgages owed by businesses d. one file of business mortgages that could not be linked to any property #### Part 4 #### 1. Create final wealth, business and trust datasets that include mortgages across wealth datasets 2. Join the wealth, business and trust sets together such that there is one row per person per property 3. I created a person dataset and joined the wealth data to this, and created all variables that were needed (part and full ownership, ownership type (personal, business, trust), ethnicity etc.) 3. created aggregated and rounded output tables off these joined datasets ######################## Limitations ######################## Where are a property is split across households, I have no way of tracking this (given it is a sample survey - only one household will be included). There are a number of mortgages that did not seem to link to any properties - these mortgages were excluded from this data. Less unexpectedly, there are a lot of properties without mortgages - I have no way of validating whether these actually should have a mortgage value attached to them. Due to the propertyid not always being filled in, and not working across datasets (e.g. trust and business datasets have their own ids that are separate, I've made some linkages by person based on property type. This probably risks erroneously linking mortages and properties together (though I have performed checks to ensure that each mortgage is only getting linked to one property and vice versa). The benefit is that a lot of properties and mortgages have now been linked - on the balance, I think this vastly improves the estimates overall. When linking without propertyids (i.e. just by year, personid and propertytype), there are often multiple mortgages that link to one property, and/or multiple properties that link to one mortgage. To prevent double counting, I made sure that each mortgage would only be counted once by assigning it to the highest value property if there were multiplematches. Similarly, I counted each property once by only linking it to the highest value mortgage where there were multiple matches. We're also limited to the accuracy of the data - for example, there might be instances where one partner owns a house on paper but in the survey they recorded the value as being split between them and their partner (or vice versa). There are a couple instances of multiple property values being recorded as one property, too. If there was someone aged under 15 who owned property, this would not be recorded in the survey, so would be excluded. The data here are not official statistics, and we advice caution when using them. ######################## Interpretation ######################## It should be noted that these tables count people, not properties. This is important because many properties will be partially owned by multiple people. Mortgage The median value of the individuals portion of the mortgage Equity The median value of the individuals portion of the asset NetEquity The median value of the individuals portion of the equity (Equity less Mortgage) CapitalValue The median value of the full value of the asset – including the portion owned by other individuals Note that these figures are median figures across all individuals that own property assets within the given demographics. Table 4 - Note that the figures will not add to the total as one individual could hold property personally and through a trust/mortgage.