Candidates for improvements to the Toronto public financial records

The budgetpedia.ca project has been working on presenting Toronto budget data since July of 2015, and I have personally been researching aspects of the budget since about 2008. budgetpedia.ca released its first version in November 2016, with as much data as we could find and process given the resources we had available. This includes summary budget data from pdf documents, somewhat more detailed budget data from Toronto’s open data portal, and some actual data (revenues, expenses, and cost elements) from the Toronto audited financial statements. There are other sources that we have not yet had time to add, including City variance reports, and FIR (financial information return) data submitted to the province, as well as capital budgets.

Over that time period, we have learned about some problems regarding Toronto’s budget data. This blog is a summary of some those, which we think could, and should, be corrected by the City as a matter of basic standards.

The purpose of these recommendations is to encourage the City to adhere to the highest standards of reporting, so that figures that are published by the City can be trusted by all parties. This in turn supports constructive discussions about issues related to the data.

We hope that these guidelines can help make City of Toronto budget data more accessible for all.

There are 4 specific, and 5 general, recommendations.

Specific recommendations

1. Accurate budget reporting

Remove distortions from operating budgets.

There are two main distortions in the budget figures presented by the City.

One is based on double counting of inter-divisional charges. If for example the IT division charges (say) Parks, Forestry, and Recreation (PFR) for computer services, that amount is posted in inter-divisional recoveries for IT, and inter-divisional charges for PFR. When expenses and revenues are combined (called ‘net’) these cancel out, but on the expense side the IT costs appear in both IT and PFR divisions, causing double counting. For the 2017 budget this practice, taken together over all programs, amounts to a doubling of ~$301.5M.

The other distortion is caused by the fact that capital staff (staff dedicated to capital projects such as building new facilities) are charged to the operating budget, and are only removed from the operating budget during the year as they are charged to capital projects. We estimate this to be about $360M. Though this is just plain wrong (capital and operating costs should not be mixed), at the very least there should be an offset in revenue, which we have not yet found. We are also not sure if this cost is built into capital budgets as well, which would be another double-counting.

In any case, these two distortions result in an over-stating of the 2017 gross operating budget by about $661.5M, or about 5.3% — certainly a material item. So the budget of $12.4B (or $12.32B — depending what source you use, that’s another story), should show at about $11.8B instead.

The solution:

• Capital staff costs should not be included in the operating budget
• There should be two versions of the budget, one including inter-divisional charges, and one excluding them. Those two budgets should obviously reconcile to the penny.

2. Matching staffing figures

Report disaggregated operating and capital staffing figures.

We think it’s important to present staffing figures with financial figures. (This means presenting operating staffing budgets with operating budgets.) For one thing staffing is about 1/2 the cost of running the City. For another, we think it’s important to make the point that it’s real people — friends and neighbours — that make up the City staff. Citizens helping citizens. Not a faceless, soulless bureaucracy as it’s so often portrayed, even by itself.

Yet it turns out that the budget operating staff figures are even harder to find and reconcile than the basic revenue and expense budgets. For some years up to 2015, budgeted staffing figures were reported correctly in a disaggregated way as operating, capital, and consolidated — good. But for 2017 only consolidated — aggregated — figures have been published, making it impossible to get accurate operational staff figures. There was one exception in a 2017 Briefing Note for staff recommended staffing budgets, but we had to make a number of estimations based on past patterns to arrive at final operating figure breakouts. This problem makes it difficult to show accurate historical trends, and reduces the value of the datasets for further analysis.

The solution:

• the City should resume publishing staffing budgets broken out by operating and capital staff.

This solution should be implemented in conjunction with terminating the practice of including capital staffing costs in operating budgets. Budgets must make sense to be usable!

3. Consistent formats

Use consistent formats, and retrofit past formats where possible.

As mentioned above, budgeted staffing formatting and content has changed radically recently, for the worse, and in such a way as to make it impossible to perform some comparisons with financial budgets.

Changes in formats to operating budget data published through Toronto’s open data portal are also problematic.

It’s good that the City has been publishing operating budget data through the open data portal since 2011. However there have been radical changes to the formats year over year. In general the City has been ‘drifting’ toward releasing Activity level and Cost Element level data, which is good, but the formats and content keep changing.

• 2015 and 2017 saw Activity level data. 2016 (approved version) did not show Activity level data.
• Up to 2016 Cost Element data was accompanied by Cost Element codes. 2017 data was not. None of the other data was accompanied by codes. Codes help with consistency in year-over-year comparisons.
• Some line items with the same Cost Element codes have been differentiated by description only (like Subway Extension property tax revenue). This creates a special case for processing.
• The first years of data were not accompanied by Property Tax revenue budgets; the latter years have.

The solution:

• Work with the user community to identify a comprehensive, generic format for operating budget data, stick to it, and revise historical datasets to conform to current standards.

4. Basic reconciliation

Generate reports and open data from the same sources.

We have come to expect reports from different sources to disagree, to the point that if such reports agree, we are suspicious. The opposite should be true — reports from different sources should always agree, and lack of agreement should cast suspicion.

For example, the gross operating budget from summary reports and from the Toronto open data portal disagree by ~$80M. The first 2017 variance reports base gross operating budgets disagree with the summary reports by about $8M.

The solution:

• generate reports (on an automated basis) from the same source.

General Recommendations

1. Programs should not be reported in Corporate Accounts

Move Parking Tags from Corporate Accounts to Agencies.

We think a clear definition of a program (at the highest level being mostly divisions or agencies) is activities involving staff. Yet Parking Tag Enforcement and Operations (PTEO) is buried in Corporate Accounts under non-program revenues and expenses. We are told the reasoning has to do with the fact that various jurisdictions are involved with “tickets”. Yet the City publishes PTEO Analyst Notes, within which this activity is identified as a program. Moreover, and most importantly, it is seen as a program. In addition, it is managed by the Toronto Police Service, as a unit. This makes it a clear analogy to Wheel Trans in relation to the TTC.

Moreover in staffing reports, the almost 400 PTEO staff positions are reported in association with Corporate Accounts. This strikes us as being similar to getting a bank statement claiming that, aside from money, there are several people in our bank account. It makes no sense.

So Parking Tag Enforcement and Operations should be listed under Agencies in the City’s standard format. This would be a much clearer disclosure of the $65.6M expenditure and $102.4M revenue involved.

2. Publish all summary datasets that appear in Analyst Notes

Match open datasets to datasets published in Analyst Reports.

Open summary datasets should systematically match dataset fragments included in Analyst Notes. These include:

• gross revenues (to Revenue Elements)
• gross expenses (to Cost Elements)
• staff, including temp/permanent and union status
• key performance indicators
• fees breakdown
• others, as appropriate, such as quantitative service data

Again, this should be planned in consultation with the community.

3. Open data repositories should contain all basic budget data and reports

Create convenient repositories standard reports, including summary and detail budgets, staffing reports, and variance reports.

Of the report sources we used for the first version of budgetpedia.ca, only audited financial statements were readily available (back to 2000).

Gross revenue, expenses, and staffing budget reports were extremely hard to find, and even after considerable search effort, some years had to be filled in from ‘prior’ columns of subsequent years (‘prior’ columns never seem to agree with actual prior year reports).

The solution:

• publish each report type sequences going back to (as close as possible) amalgamation, in the toronto.ca/open repository

4. Reliance on automation

Rely on automation to produce published tables.

In gathering, processing, and importing the budget sources of data, our impression is that the source data tables are still largely generated by hand, which is inherently error-prone. It would be better to rely on automated systems.

5. Motivation by open data

Embrace open data to improve data quality.

Our final point is that publishing data (or anything else) tends to focus the mind of the authors and publishers to improve the quality of the published materials. Therefore the more the city can rely on automation to produce the budget data, up to and including granular geo-coded cost centre data, the better the quality of the data will be for everyone, inside and outside government.

If such open data were all in machine-readable format (as it should be), this would also mercifully spare the outside community from having to ‘scrape’ data from pdf reports, an aggravating, error-prone, and wasteful exercise.

Henrik Bechmann is the project lead of budgetpedia.ca. The opinions expressed here are his own.