Auckland's transportation authority, Auckland Transport, provides much of its internal data to the public.  Here is a link to data it provides on Auckland's bicycle infrastructure. The data are licensed under the Creative Commons, Attribution 4.0 (International). Notice that the data package includes a shapefile that one can load into GIS software, such as QGIS. 

The attribute table for the shapefile includes the year in which each item of infrastructure was built.  This allowed me to strip-out all the infrastructure interventions that were built in 2018 and after.  I had to do this because the most recent published New Zealand Census data on journey-to-work are for the 2018 census, with the census date being March 6 of that year.  (Incidentally, it was a rain-free day, so weather would not have impeded commuter cycling.) As a first cut, I deleted bicycle infrastructure that provides no physical barrier between the cyclist and the motor vehicle traffic.  I did so because I wished to isolate the infrastructure types that, I assume, would appeal most to potential commuter cyclists.  (As you will see in the webpage introducing a spatially autocorrelated regression approach, I eventually add back these initially culled infrastructure items to the dataset.) After this initial cut, the dataset included infrastructure items classified as "Off-Road Cycleway," abbreviated ORC;  "Off-Road Shared Path,"  abbreviated ORSP; and "On-Road Protected Cycle Lane," abbreviated ORPCL.  It excluded items classified as "Local Area Traffic Management",  abbreviated LATM; "Shared Zone,"  abbreviated SZ; "On-Road Unbuffered Cycle Lane," abbreviated ORUCL; and "On-Road Buffered Cycle Lane," abbreviated ORBCL, none of which provide a meaningful physical barrier between the bicyclist and the motor vehicle traffic.  I also excluded the category, "Off-Road Trail, (ORT)" because off-road trails are designed for recreational cycling, not commuter cycling.  The names of these infrastructure interventions generally tell you what they are. The exceptions are Shared Zone, which is essentially a pedestrian mall in an urban core area that very slow-moving cars and bicycles are allowed to use, and Local Area Traffic Management, which is a residential street with traffic-calming features, such as speed bumps, chicanes, and mini-roundabouts, to slow the motor vehicle traffic.

Statistics New Zealand, the governmental agency that administers the country's census, publishes census data on how people journeyed to work on census day. You can download the raw data used in my analysis here. Be sure to download the dataset for statistical area 1 (SA1), which is the smallest geographic unit for which New Zealand census data are aggregated and which I used as the unit of analysis in my data modeling. From this point on, I'll refer to SA1s as "area units" to avoid technical jargon. Area units typically have a residential population between 100-200 people, although densely developed urban cores with apartment buildings usually have more.

If you go to Table 5b, you will find the raw data for how people commuted to work on census day, 2018.  To preserve confidentiality, all counts are randomly rounded to base 3. In cases where rounding to base 3 would not have adequately preserved confidentiality, Statistics New Zealand reports "C" in the field.  I stripped-out records that had this entry in any relevant field. One shortcoming with the journey-to-work data is that the census respondent could choose only one mode for how they commuted to work when an unknown number of them were multimodal commuters, such as a commuter who  bicycled to a ferry terminal on the way to work.  This respondent could have ticked either "ferry" or "bicycle" but not both on the census form. Thus, the journey-to-work data are not entirely accurate, but they are the best available.

One of the control variables I used in the regression analyses is the fraction of commuters from each area unit who worked at a job that likely would not require a change in clothing after arriving at work after bicycling in. I assume that holding such a job would make a commuter more likely to cycle.  Census Table 3a contains these data. I elaborate on how I used these data in the Transformed Data page of this website.

Another control variable I included was the fraction of the commuters from the area unit who held a university degree. I am not making the implausible assumption that university graduates know more about the financial and health benefits of commuter cycling.  Everyone knows this.  I used holding a university degree as a proxy variable for having an above-average psychological disposition to engage in deferred gratification.  After all, acquiring a university degree requires enduring much pain and suppression of immediate desires in the present in order to enjoy benefits in the future, just like enduring the physical strain and discomfort of cycling to work instead of driving allows one to enjoy future physical and financial benefits. I assume that having an above-average propensity to engage in deferred gratification makes one more likely to bicycle to work. You can find the raw data on the educational attainment of the residents of each area unit in Table 2 of the census data.

The census data are geocoded.  Each record in each table includes the unique identifying numerical code for each area unit.  This allows mapping of the data.

Statistics New Zealand also publishes shapefiles for use in mapping the census data.  You can download the shapefiles used in my analysis here. Just type "Statistical Area 1 (Generalized)" in the search bar, and you should go right to the appropriate page.

The shapefiles use a coordinate reference system that allows accurate mapping of each area unit. In addition, the attribute table for each shapefile includes the unique numeric identifying codes for the area units. The Transformed Data webpage explains how I used these shapefiles in my GIS analysis.