Why do expenditures and revenues not always add up to the values in the budget summaries?

Each difference I've identified is documented. There are several possible reasons:
(1) Sometimes the value in the department budgets just don't match what is in the budget summary.
(2) In a few cases, dedicated funding directly offsets an expenditure (WMATA). In other cases the dedicated funding shows up as a revenue stream and an expenditure (HUD housing subsidies). I try to be consistent by choosing the later representation.
(3) I made a mistake.
(4) There's a mistake in the budget documents.
(5) The value for an adopted budget year differs from one budget to the next.
(6) Something I don't understand about PAYG.

Why do department budgets not match the summary at the front of the document?

I have no idea. This happens several times specifically for the Department of Environmental Services in the more recently adopted budgets. When there was a discrepancy between the budget summary and the department budgets, I use the individual department budget and not the amount in the budget resolution. If you know why this is, please let me know and I'll update.

Dedicated funding shows up differently in the budget?

Sometimes, yes. For example, Housing and Urban Development (HUD) provides funding for affordable housing subsidies that is booked as revenue in the Section 8 Fund. This money shows up as revenues. The housing subsidy itself also shows up as an expenditure. This seems like a good way to report the budget item: it highlights both the fact that the federal government is providing money, and that money is being spent on something (housing, in this case). The downside to this approach is that the link between the revenue and spending is not always clear. For example, if the county reduces spending on housing subsidies, HUD will stop paying the county the funding, which is something you wouldn't know just by looking at the budget numbers.

However, this isn't always the way that these types of "pass through" subsidies are reported. For example, in adopted budget summary at the start of the document, WMATA funding is specified excluding (or net of) the state and regional contributions. The budget shows the contribution for Arlington's share of WMATA expenses, but the reported number leaves out the state aid, the regional gas tax, and the I-66 toll money. This stands in contrast to the Section 8 funding, for which all sources of revenues are reported and all expenditures are reported. For consistency, I put WMATA funding into the budget as both a revenue and expenditure.

Do adopted values change for the budget for some departments?

So, for example, the 2019 budget has an adopted budget for 2019. In the 2020 budget, they will report the 2019 adopted budget again alongside the 2020 adopted budget. The value for the 2019 adopted budget in the 2019 budget will not precisely match what is reported in the 2020 budget for a few departments. For spending, I typically use whatever the earliest adopted value is: so for the 2019 adopted budget, I would report it from the 2019 budget. For revenues, I use the value in the Arlington Open Data resource, which I think reports the latest adopted version. The differences are usually pretty small.

Why is there a difference in the actual PAYG numbers from the budget summary?

In the budget summaries, PAYG capital projects are listed including a transfer that is not included in the total expenditures in the individual budget page. I don't know which is right, the difference is not explained in the PAYG documentation. In the absence of further guidance, I go with what is on the individual PAYG budget and not what is in the summary.

Why do you have a tree view and a table view?

They show different things. The tree view gives a visual sense of the relative magnitude of budget items. However, the tree view is not very good for making year-to-year comparisons or finding a budget item if you don't know where to look. Therefore, I include a table version with a search and sort capabilities. The table makes it easier to see the spending categories that are driving overall changes to the budget, and it is easier to use the table to search and to find individual budget items of interest.

How do you compute the individual amount spent on each program?

I take the amount of taxes you paid, and I multiply that amount by the share of the category of spending. So, if you're looking at police, and police at 20 percent of the budget, I assume that 20 percent of your taxes go toward the police. This is just an approximation.

Why is the change for some of these categories infinite?

That's because this expenditure or revenue did not exist in the previous year, therefore it is infinitely larger than it was last year. This sometimes happens because programs are new, but other times the programs get split off of an existing budget item.

There are a lot of infinite and -100 percent changes, why is that?

The county, when reporting its budget, often breaks out certain department expenditures into more granular category. So just because something looks "new", does not mean it is new. It might just been part of another budget item in the past. This goes in reverse too; sometimes budget lines get consolidated. They're not terminated, just end up wrapped up in another line item.

There are a lot more categories of revenues than are reported in the revenue summary part of the budget, why is that?

I pull the data for the revenues summary part of Arlington's Open Data portal. It's a bit more detailed than the budget summary provided in PDF form on the budget page.

I see negative numbers in the revenues. How are those represented in the tree?

They don't show up in the tree visualization. They will show up accurately in the table.

If I look at the actual revenues and spending, it looks like we're running huge surpluses. Why is that?

Adopted revenues and adopted spending, ignoring any errors, should be the same. In reality though, they don't have to add up. Revenues and spending don't match spending. But still, the county uses its left over money at the end of the year. But that doesn't really account for such a big difference. Two other things tend to account for most of the "surplus." First, Arlington lists money left-over in the General Fund as a source of revenue. This number is huge, it's 9 figures. In 2016, it was $155 million. This isn't really isn't a "revenue" since there is no comparable expenditure. Once I figure out exactly what part of this counts and what does not, I may revise the revenues. The other issue is that some longer-term projects end up being several years long, and the spending timing does not exactly match what is in the adopted budget. So for this reason, revenues and expenditures may be mismatched across years giving the false impression of deficits or surpluses where there are none.

When I look at a category, I see a particular amount spent. When I click on that category to get more detail, the numbers add up to a greater amount. Why is that?

That is almost certainly that there are negative numbers in the budget that cannot be visualized. For example, in adopted 2019, Real Estate Taxes bring in $730,267,530. Click on that value, and the value is $739,142,530. What isn't shown are the refunds and offsets which bring that number down to the correct net $730,267,530 (which is $9.6 million). You can find those values in the table view for the budget.

Why do the FTEs not add up to what is in the summary?

A few years ago, APS broke up the Capital Projects category into Capital Projects and Bonds. The Bonds category has expenditures related to issuing, refinancing, and repaying debt (the bonds). These expenditures, when they got split, got removed from the operating budget. You can see the "Bonds" category in the APS budget documents, but it isn't included in any of the totals. Same for the employees; the FTEs got removed. With the expenditures, that money is repaid eventually, so it really is still in the operating budget. The expense is just deferred to a future budget. The FTEs, however, don't show up in future budgets. They are employees doing work right now. So while APS may appear to treat expenditures and FTEs the same in the budget document, they're really not. I reconcile that inconsistency by putting the people who work on debt issues back into the tally of county staff as they were before APS made this reporting change.

What is the Capital Improvement Budget?

Unlike the regular budget, this budget is for things that are long-lived infrastructure: schools, parks, roads, water, sewer, etc. They tend to be expensive, long-term projects.

But we don't pay for this all at once?

The funding for these things comes from a variety of sources: some are from dedicated external funding (from the state or federal government), some of the money is borrowed and paid back later with future tax and fee revenues, some of the money has been saved up in one of the county's funds, some is paid for out of current tax revenues, and rest of the money comes from other sources.

I don't see the Long Bridge Aquatics Center in the most recent CIP, where is it?

Once a CIP is approved, the project does not necessarily come up again. Even if the project isn't completed by the time the next two-year CIP is released, it's still approved. And the CIP is a ten-year plan. This is why I include the last four CIPs in the database: so users can look back at all of the projects that may not have been completed but are still on the table. The Long Bridge Aquatics Center is one such item. It was approved in the 2014 CIP, and hasn't shown up since as a line item (although it is mentioned in the text).

So you said the CIP was a ten year plan, but so why did you not include the 2010 CIP?

The 2010 CIP was the last five-year CIP. They transitioned to ten-year plans in the 2012 CIP. (Note: I refer to the CIP by the year it was approved.)

How important is the CIP?

I personally think that the CIP is just as important, if not more important than the regular budget. First, the latest 10-year CIP is about $3 billion. That's about $300 million per year. But add in the projects from previous CIPs and future CIPs, and we're talking about a very substantial investment in infrastructure each year. The number is not all that much smaller than the rest of government spending. And unlike the operating budget, which offers some year-to-year spending flexibility, many of the projects in the CIP commit Arlington to maintenance and operations costs far into the future. But the CIP is also important for what's not in the CIP: opportunity cost and option value.

What is opportunity cost and why does it matter to the CIP?

Opportunity cost is the loss of a resources next best use. For example, if the county owns vacant land and builds a school on it, there is an opportunity cost of using that land. The land could have been used for houses, maintenance facilities, parks, open space, etc. In this example, the way we measure the opportunity cost is through the cost (market value) of the land. But since the county already owns the land, that cost does not show up in the CIP. The land's market value reflects its value as housing, parkland, space for business, etc. The county is losing a resource, which may have quite a substantial market value, which is not showing up in the budget. By contrast, the general government and school operating budgets don't tend to fund things that come with high, unlisted opportunity costs. For example, hiring a police officer generally doesn't deprive the county of any resource beyond the money spent to hire him or her.

What is option value, and why does it matter to the CIP?

Option value is the value of environmental resources. For example, people get benefits from open spaces and trees in their neighborhoods. Cut down the trees or pave over the open spaces, and the community loses those benefits. The loss of these benefits does not show up in the CIP.

Can you explain why the summary of assessments is a little slow?

If you're downloading all zip codes, your browser has to download a 10MB file and then parse nearly 70,000 records to find the information you want. That's a bit of work.

What are the errors in the data?

Some of the zip codes are typos. For example, there are a several properties in the far north that are listed under the 22202 zip code, which is in south Arlington (Crystal City, Aurora Highlands, Arlington Ridge, etc.). Some of the buildings appear to be mislabeled. There is at least one building on Jefferson Davis highway which is labeled a smaller office building, but appears to be greater than seven floors.

I know this particular property was assessed differently, why is that?

Arlington could have made a mistake in the data, I could have made a mistake reporting it, or the property may have been reassessed later in the year. I report the last assessment in the year.

How accurate is the assessment data?

I'm not entirely sure. I find some of it a bit confusing. The underlying data provided by Arlington, except for a few typos, appears to be pretty high quality. For example, the 2017 Comprehensive Annual Financial Report lists $82.4 billion worth of assessed property (about $8 billion of which is tax-exempt), and the total of all of my data is $82.2 billion. I think that's pretty close, given that there are 70,000+ properties in the dataset!

What assets does the county own?

The county owns a lot of assets--including things such as roads, land, physical capital (cars, buses, equipment, buildings), and other stuff--but for the purposes of this website, we report their financial assets which are used to meet both the day-to-day needs of county operations as well as act as a reserve of longer-term funding for infrastructure investment. The county invests the money it has in a variety of securities and bank accounts, many of which are the same types of investments available to individual investors.

Does the county have a checking account?

Yes, actually it does.

What types of investments does the county own?

The county owns typical bank certificates of deposit (CDs). The county owns agencies, which are a kind of federal debt. Arlington owns money market accounts, savings accounts, corporate bonds, and investments in the Virginia Investment Pool and SNAP.

What is the Virginia Investment Pool and Virginia SNAP?

As a service to local governments, many of which are smaller and do not have professional money managers, Virginia established these pools of investments to help local governments earn money on their investments and remain compliant with federal regulations. You can read more about them here: VIP and SNAP.

Arlington is investing money? Isn't that risky?

All investing is risky, but some investments carry less risk than others. The two main sources of risk in these types of assets are interest (and inflation) risk and default risk. If Arlington owns a bond and the interest rate goes up, that bond becomes less valuable. If Arlington needed to sell that bond before it matured (was repaid), it would take a loss. If Arlington owns a bond and the bond-issuer did not repay it, it would take loss. Nonetheless, both of the risks in Arlington's portfolio appear to be quite low. Most of the bonds are very short-term bonds, which limits the effect that changes in interest rates (and inflation) have on them. Furthermore, Arlington invests in high-quality and federal bonds, limiting the risk that the borrower will not repay the funds. The downside is that the interest earned on these is not high. Overall, Arlington's asset portfolio seems consistent with what I would consider prudent financial stewardship, although I can't say I know how the state investment funds operate.

Is it possible to earn more money on these investments?

It may be possible. If some of these assets are earmarked for specific funds, and those funds are not going to spend the money, it might be possible to invest the money in longer-term investments. Longer-term investments typically offer higher-interest rates, and if the county knows it does not need to spend the money in the next few years, Arlington may be able to structure its investments to take advantage of the higher returns. There is probably not a ton of money to be had doing this, though.

Why are Arlington's bonds sorted by both issue and maturity?

The county owes money, and there are two ways to organize it. First is by issue: this is sorted by the year in which the money was borrowed. The advantage of this view is that it is easy to see trends in county borrowing. The second is by maturity: this is sorted by the year in which the county needs to pay back the money (excepting interest, which is paid regularly, typically twice a year). The advantage to sorting this way is that it is easy to see how much money the county will owe each year as the debts are repaid.

What are the school, government, and utility funds?

Each bond issue generates funds that are typically split among three different funds: general government, schools, and utilities. The money goes toward one or more of these funds. School funds, for example, might be used to build schools. Utilities funds might be used to construct things like water treatment plants.

What are variable rate coupons?

Most of the bonds do not have this feature, Arlington pays the same amount of money out at regular intervals (twice a year) until the debt is settled. The county makes regular coupon--or interest--payments on each of its bonds. A few of the bonds have variable coupons: The rate changes each year. Interest payments listed on the page are based on the *first* of the coupons. Variable rate coupons typically go up over time, meaning the interest increases over time.

How is interest computed?

Annual interest here is an estimate, and should not be taken as the actual amount paid in each year. The amount can vary because the coupon may be variable (see above). Alternately, the interest payment occurs several times a year, and if the bond matures or is repaid before the end of that year, there may be fewer than a full year's worth of interest payments. Lastly, the bonds might be callable, which means that they'd be refinanced before they're due and the interest would change.

What is a callable bond?

Most municipal bonds are callable. That means the issuing agency--Arlington county--can call the bond after a certain amount of time or on certain dates. This means that the county can pay off the bond early. In most cases, this happens when interest rates go down. In this case, Arlington can issue new bonds at lower interest rates, use the proceeds to pay off the higher interest debt, and pocket the difference.

What is the bond premium amortization schedule?

A bond premium amortization measures the difference between the market price for the bonds and the face value of the bond (how much money is owed when the bond needs to be paid).

Why are pensions not listed here?

That's a future project. In the 2017 Comprehensive Annual Financial Report, the auditor states that the county is about $90 million short of its pension obligations. If true, the pension would be quite well funded.

If true?

It's not true. The county is not almost certainly undervaluing future pension liabilities. This isn't an Arlington-specific issue, almost everyone does this. In order to figure out how far underfunded a pension is, you need to figure out how expensive the pension benefits are using something called a "discount rate," which is 6.75 percent in this case. Although consistent with general practice, this discount rate is almost certainly too high. I refer interested readers to a couple of summaries of the issues around valuing pension liabilities here and here. Using 6.75 percent as the discount rate, the auditor estimates a $90 million pension liability. Using a more conservative 5.75 percent discount rate, the money owed by Arlington jumps to $373 million. I'm guessing that the 5.75 percent discount rate is still entirely too high, but that's just a guess.

Is this an immediate problem?

It's better to address this sooner rather than later, but it's probably not going to become a critical for quite a long time. It could be decades before this becomes an issue. In this county, it will probably manifest as more of a chronic, ill-timed problem. In the event that the investments have a run of bad years, the value of the pension fund will fall behind badly as investments underperform their expectations, while payouts and obligations remain stable. At that point, the county will need to make additional contributions to shore up the fund. The problem will feel more serious because the county will need to make those additional contributions during a time when investments and the economy aren't doing well, and taxpayers will feel the pinch a lot more.

Why is there a separate category for school budgets?

The school budget is complicated, and APS does a good job of providing a substantial amount of detail. So it seemed worth it to give schools their own graph separate from general county spending for visitors only interested in the school budget.

Why is there a breakdown of each department's / school's spending after 2017 but not before?

In the 2018 budget, APS recategorized its spending from 7 categories to 6 categories. The old 7 categories do not go into the new 6 categories the same way for each department. So instead of figuring it out on a department-by-department basis, I just report the totals in the earlier years.

There's a lot of variation in some of the smaller categories from the 2017 to 2018 budgets. Did funding change that much?

It's possible I made a mistake copying the data. I try to validate the data as I input it, but it's possible it slipped through. In one or two cases, the numbers just don't add up in the original document. Usually the differences are very small though (less than $1,000). Some departments get moved around, in particular items related to assessments seem to get moved in and out of the Superintendent's Office budget on a regular basis.

Why are the capacity projections the same year-to-year?

These only get updated every few years.

But the enrollment projections change on a yearly basis?


Why is there no H-B Woodlawn in the middle schools section?

The middle and high school stats are combined and reported under high schools.

Why are there no capacity projections for 2015?

The last capacity projection I see was released in fall 2016.

The demographics in the suspensions are different than the ones on the civil rights page. Is that correct?

The data on the "Student Race" statistical repository page for APS is compiled in October. I think that since suspensions are yearly, that the demographic data is drawn from a different time period as students move in and out of the schools during the school year.

Why not include data on average class size?

Average class size is important, however, the reporting is challenging beyond the elementary schools. By the time the students reach high schools, students are all on different schedules and take different classes. A uniform, school-wide measure of average class size is probably not very instructive. I do not think that I could add much value beyond what APS already offers for school class size reporting.

What are "unlocated" incidents?

Sometimes, the incident has a latitude and longitude that place it outside of any civic association. It could be an error, but in my experience, most of them are located in the Pentagon parking lots, Arlington Cemetery, or some park not attached to a civic association. About 90 percent of the incidents can be located within a civic association's boundaries.

You have incidents in Cherrydale that aren't in Cherrydale. Why is that?

Cherrydale has a donut-shaped civic association. It would have taken a bit of work to exclude things inside the donut for all incidents, and they still seemed relevant to civic association, so I left them in the report.

Where does the data for the zone come from?

Arlington graciously provides mapping information from the zones. It is available on their open data website for anyone to use.

Is this a complete map of all zoning in Arlington?

It is as complete as Arlington provides, with one exception. Object number 610 has been omitted, there are too many points for me to render it correctly using the approach I'm using.

Why on earth is Don Tito on the homepage?

Why wouldn't Don Tito be on the main page? Don Tito carries outsized cultural significance in this area, and it seemed appropriate to recognize it. To illustrate: A friend overhead the following conversation on the Metro: Person One: "Bro, I found a group house a block from Don Tito!" Person Two: "Dude, that's great! How much does it cost?" Person One: "Bro! It's a block from Don Tito!" In all fairness, I've only been there once. Someone bought me a hat from the place. The (brunch) food was pretty decent. I would happily eat it again. And I learned that apparently you can dispense rosé from a tap. Who knew?

What's your background in web design and hosting?

Absolutely none. I bought a style sheet from pixelarity.com and started messing around with the design. I took a few hours of an on-line class in JavaScript from Code Academy. I use the Google a lot. If you like the site and want to help improve it, please contact me.

What other ambitions do you have?

I like to build a "build your own" school boundary app, to give people a feel for the kinds of tradeoffs that APS staff need to make and make it consistently available If you have experience with this kind of stuff and would like to help, please let me know. I would like to automate more of the data gathering processes and do more with the assessment and real estate sale data, which can be used to predict changes to assessments and tax revenues on a more granular level.

What visualization libraries are you using?

The maps use the Google Maps API. The treemaps (spending, funds, etc.) use the d3 (v3) libraries. The code is derived from bost.ocks.org and modified for my purposes. Mike Bostock, former NYTimes visualization expert and developer of d3, made this code available. Given how widely people copy it on his website, I'm assuming that the only item he meant to copyright was the style sheet. Please correct me if I'm wrong and I'm happy to license the work or move to something else if necessary. The assets and liabilities are drawn with d3 (v5). Special again, thanks to Mike Bostock, who made his d3 code available at bl.ocks.org. The tables are made with DataTables.

Where are the pictures from?

A few pictures on the home page are from Arlington County. All of the pictures that are not mine were made available under the Creative Commons 2.0 license. The rest of the pictures are mine, and also available to anyone under the same license.