How to Mine Pittsburgh Post-Gazette Death Notices for Community Data
Most people treat death notices as a formality. They are not. When you sit down and actually read through them, the Post-Gazette's archive becomes one of the most underused local history databases in Western Pennsylvania. I have spent more years than I want to admit pulling obituaries and running them against census data, city directories, and old church records. The pattern recognition is where the real work happens. The paper publishes death notices daily. Each one typically lists the deceased, age, date of death, surviving family members, sometimes former residence, employer, and funeral home. That single notice is a data point. Stack enough of them and you get a picture of neighborhoods changing, families migrating, and institutions closing. I started doing this systematically after working on a housing study that needed demographic baseline data for neighborhoods that had been redrawn multiple times since 1980. Census tracts do not always match what people actually live in anymore. Getting the notices is straightforward. The Post-Gazette moved most of its archive online through its digital subscription portal. You can search by name, date range, and keyword. There is also the Pittsburgh Post-Gazette death notices section on their website, which allows free browsing by date. For bulk work, you will want to move past manual copying. I use a simple spreadsheet template with columns for name, date, age, stated address, names of survivors, and any mentioned institutions like churches or employers. Filling these out takes about three to five minutes per notice once you are calibrated. Doing it by hand for a full year in a specific neighborhood comes out to roughly four hundred notices, which is about forty hours of work. That is the grind part.
Where people mess up is treating every notice as equally reliable. They are not. A notice from 1995 might list an address on Murray Avenue that was actually the deceased's last known residence, not their current one. By 2005, that same street might have been rezoned or the building demolished. I ran into this exact problem when tracking a family line through the Lawrenceville area. The death notices kept listing a house on East 7th Street that did not appear on any modern map. It turned out the building had been condemned in 2003 and the address was reused for a new construction project two years later. The workaround was to cross-reference the address with the Allegheny County assessment records and a few old Sanborn fire insurance maps. Once I confirmed the physical location had changed, I stopped treating the notice address as a current residence and started treating it as a last known location marker. That shift in interpretation made the whole dataset usable. Family names in the notices are another goldmine that people rarely exploit properly. When a notice mentions surviving siblings or children, those names are often not present anywhere else in the record for that deceased person. I built a simple linking script that pulls sibling and child names from each notice and connects them to other notices in the dataset. Running this on about six hundred notices from the Strip District over a fifteen-year period revealed three extended families that had effectively vanished from the census because they were spread across multiple zip codes. The notices showed they were still there, just living in different buildings on the same block. This kind of clustering does not show up in standard demographic reports. Employer mentions are less common now than they used to be. Back in the 1980s and early 1990s, many notices listed the company the person worked for. U.S. Steel, Westinghouse, Alcoa, various hospital systems. Those references are useful for understanding which industries were still employing local residents at the time of death. If you see a cluster of steelworker notices spiking in a certain neighborhood during 1986 and then dropping off sharply by 1992, that is a data point that aligns with the blast furnace closures. It is not speculation. It is right there in the text.
Church and organization mentions work the same way. A sudden increase in notices referencing a specific parish or fraternal group in a particular area can indicate that community was still active and organized. A decline in those references often signals that the institutional framework around that neighborhood had eroded. I use this alongside membership records when they are available, but the notices alone are enough to flag shifts. There are real limitations to this approach. The coverage is not uniform. Notices from wealthier neighborhoods tend to be longer and more detailed. Notices from working-class areas or places with higher turnover can be sparse. Some families choose not to publish notices at all, or they publish them elsewhere. The digital archive goes back to the mid-1990s reliably, but earlier material requires visiting the library microfilm collection or paying for third-party database access. The Post-Gazette's own archive subscription costs money and the search interface is not great for bulk extraction. You can export results, but the formatting requires cleanup before it is usable for analysis. If you are doing this for academic or professional purposes, I would recommend pairing the death notice data with the Pennsylvania Vital Records system for official death certificates. The certificates provide different fields like exact cause of death, medical certifier, and burial location. They do not have the family network information that the notices do. Using both together gives you the structural data from the certificates and the relational data from the notices. Trying to rely on one source alone will leave gaps.
Get the Full Details
The most practical workflow I have found is to export the notices you need by date range, paste them into a plain text editor to strip the HTML formatting, run a basic parsing script to pull out the key fields, and then load everything into a relational database or at least a well-structured spreadsheet. From there you can do location analysis, name correlation, and trend mapping. The initial setup takes a few hours if you know what you are doing. It is worth it if you need granular community-level data that does not exist anywhere else. For people who just want to look up a specific person, the Post-Gazette website has a search function that is adequate. Go to the death notices section, enter the name, and filter by date. Results are usually immediate. Paid subscription accounts get full-text access and earlier archives. Free accounts are limited to recent notices and some older ones that have been indexed publicly. The value of this work depends entirely on how carefully you handle the source material. The notices are not raw truth. They are curated statements provided by families and funeral homes, often edited for length and tone. But within that editing process lies a lot of signal if you know where to look. Address changes, family relocations, employer names, church affiliations, and the simple accumulation of who is dying where and when tells you more about a community than most published reports do.
