Urban planning community

# Thread: Demographic estimates and forecasting

1. ## Demographic estimates and forecasting

Now that the 2000 Census data is almost completely obsolete, I'm interested in how everyone goes about estimating current populations and other demographic information (if you do at all). I know the merits and fallacies of many of the various methods...what I am interested in is what other planners actually do.

First, what do you do for current demographic estimates? Do you develop your own estimates based on extrapolation (if so, what's your favorite method), go to all of the effort of performing a cohort-component analysis (births, deaths, in- and outmigration), get estimates from someone else (COG, etc.), or do you buy your data from ESRI or similar?

Second, do you estimate total population, households, and jobs and utilize 2000 Census racial composition, age, income, avg HH size, etc. or do you some how estimate these as well?

Third, what about forecasting? How far out do you go and how to you calculate? Do you extrapolate curves or extrapolate cohort-component info.

Finally, how do you incorporate real-world knowledge into your estimates? Say you know a 5000 unit subdivision was recently approved for your town of 20,000. Do you see this as a sign of things to come and adjust your curve to meet this data point (by changing your formula and therefore the slop of your curve), or do you view it as a one-time bump and raise the curve to meet it and keep your slope?

Sorry if this post was too detailed...I'd really like to hear any and all thoughts on this subject.

2. This has been addressed in several other threads...so either look through the economic development threads or wait 'til Cardinal posts and go with his sage like answer.

3. I completely disagree. I have run several searches (using words like population, demographic, projection, etc.) and reviewed over 3 years of posts and haven't found any thread like this.

There are threads on how to do such and such calculation...that's not what I am asking, I know how to do the calculations. I want to know what people that use demographic estimates and forecasts regularly actually do.

I look forward, though, to reading Cardinal's response.

4. Aw shucks...

I usually cheat. Depending on the needs of the project I will use state estimates, the American Community Survey, or a source such as ESRI. On rare occassions I will construct my own estimates (usually for small areas) which might use a cohort model, building permit data, or straight line methods.

5. Just like Cardinal -
I have used state estimates, the American Community Survey, building permit data, and straight line methods.

how do you incorporate real-world knowledge into estimates?
My favorite example is where the Census Bureau's county estimates are plausible ? but their county subdivision (township) estimates are way off even a joke - they have our riverbottom (floodplain and floodway) with no water and sewer service, no housing permits are issued township doubling in population. So allocation models are troublesome.

6. DemographicsNow is a great resource. It requires a paid subscription. I typically make my own projections, then compare them against DemographicsNow's estimates and projections. I like using my own when I can because I'm better able to explain the methodology I use. It's difficult to explain someone else's methodology when that methodology is rather complex.

7. We update our population estimates on a monthy basis for ovver 300 communities. This first document is our yearly estimate. The methodology is found at the end of the Document: http://library.semcog.org/InmagicGen...higan_2008.pdf

As you can see by the map, shrinking populations are having a dramatic effect on our core areas. The map can be somewhat misleading because the region still has quite a few rural areas. The shape of the local economy is forcing us into a time when folks are leaving in droves. This allows a quicker exit from the central cores to second tier suburbs where many of the jobs are now. This complicates forecasting population as we have a huge number of vacant homes since the foreclosure crisis started a few years back. Number of housing units is not a good source because so many years are under transition and have homes that are unoccupied for long periods of time.

The second document is our October figures:

We use our data primarily to run our models and to share with local agencies in our region. Therefore we do forecast future numbers on a regular basis, but I don't think we do a full forecast every month (I am not a forecaster.)

8. Estimates:

We were not included in the ACS, so I'm stuck with Census 2000 for some baseline numbers like pph, etc. As you will see below, we take a VERY conservative approach to estimation, mainly because missing your estimate at a critical threshold point in Texas (like all of the funding changes, ETJ issues, etc. at 50,000) can be quite controversial.

The City uses the 2000 Census information as a base point for estimating population. Estimates for subsequent years are prepared using the housing unit method, which considers building permit data, occupancy rates, and persons-per-household. The building permit is added to the population estimate when it is ready for occupancy. For single-family, a six month delay is used for occupancy from the month the permit is issued. For example, a house permitted in January would not be considered for population estimates until June. Multi-family uses a similar process, but with a 10 month delay.

Single-family population = total units x occupancy rate x Census 2000 persons per household (owner-occupied)

Multi-family population = total units x occupancy rate x Census 2000 persons per household (renter-occupied)

Population from Annexation = based on Department of Justice population assessment required for each annexation; new units after annexation use the two methods above

Other population = the On-campus University population. 4,608 were in dorms at the time of the 2000 Census, establishing a baseline to adjust from in subsequent years. I update this annually with the __________ University Office of Institutional Research. The on-campus University population was likely undercounted in the 2000 Census, but the City conservatively assumes all were counted in 2000.

The occupancy rate for single-family is updated annually based on the number of ___________ MLS listings for “single-family”, “manufactured real property” and “townhouse” with “San Marcos” identified as the City. I typically take the total number of listings and double it to account for any for sale by owner homes not listed on the MLS. I assume 100% of the units listed on the MLS, doubled, are vacant. Some of these listings are technically outside of the city limits, but I include them anyway to assure a conservative estimate. Likewise, assuming 100% are vacant and doubling the number of units listed in the MLS contributes to a conservative estimate.

The occupancy rate for multi-family is updated annually based on the Fourth Quarter __________ Multi-family Trend Report prepared by <Local Proprietary Multifamily Research Group with a good track record>.

Projections:

I hate doing this voodoo. I basically use the average annual growth rate over the last 10 years, with it plugged into a decay model to inject some realism 5-20 years out. I also show the numbers for a 1.5%, 3.0% and 5.0% growth rate with the same decay. When asked for a projection, I show all four estimate models. I've tried using the 1.0/1.5/2.0 model, which is used by our state demographer, but it hasn't performed well for our city over the last 10 years or so (they suck at dealing with explosive growth patterns in metro areas).

9. Originally posted by Cardinal
Aw shucks...

I usually cheat. Depending on the needs of the project I will use state estimates, the American Community Survey, or a source such as ESRI. On rare occassions I will construct my own estimates (usually for small areas) which might use a cohort model, building permit data, or straight line methods.
Well, I looked forward to your response because I figured my situation would be similar to yours (private sector, having to come up with data for many different geographies and rather quickly). Correct me if I'm wrong, but isn't the ESRI data just ACS estimates GIS-ified? (I have the white paper, but I haven't read it yet)

Originally posted by DetroitPlanner
We update our population estimates on a monthy basis for ovver 300 communities. This first document is our yearly estimate. The methodology is found at the end of the Document: http://library.semcog.org/InmagicGen...higan_2008.pdf

As you can see by the map, shrinking populations are having a dramatic effect on our core areas. The map can be somewhat misleading because the region still has quite a few rural areas. The shape of the local economy is forcing us into a time when folks are leaving in droves. This allows a quicker exit from the central cores to second tier suburbs where many of the jobs are now. This complicates forecasting population as we have a huge number of vacant homes since the foreclosure crisis started a few years back. Number of housing units is not a good source because so many years are under transition and have homes that are unoccupied for long periods of time.

The second document is our October figures:

We use our data primarily to run our models and to share with local agencies in our region. Therefore we do forecast future numbers on a regular basis, but I don't think we do a full forecast every month (I am not a forecaster.)
DetroitPlanner, thanks for the great response! Its amazing (but not suprising) that Detroit has lost 90k people in the last 8 years...

Just to summarize (and correct me if I'm wrong), you use Census 2000 data as a baseline, updated with building permits and demolitions, and apply vacancy rates and average houshold sizes (by housing type and extrapolated from the 1990-2000 %age change) to gain the total population estimate. Right? I like this method, especially with your ammendments to the Census Bureau method.

How long, would you estimate, does it take you to produce the numbers for your monthly updates?

Originally posted by Suburb Repairman
Estimates:

We were not included in the ACS, so I'm stuck with Census 2000 for some baseline numbers like pph, etc. As you will see below, we take a VERY conservative approach to estimation, mainly because missing your estimate at a critical threshold point in Texas (like all of the funding changes, ETJ issues, etc. at 50,000) can be quite controversial.

The City uses the 2000 Census information as a base point for estimating population. Estimates for subsequent years are prepared using the housing unit method, which considers building permit data, occupancy rates, and persons-per-household. The building permit is added to the population estimate when it is ready for occupancy. For single-family, a six month delay is used for occupancy from the month the permit is issued. For example, a house permitted in January would not be considered for population estimates until June. Multi-family uses a similar process, but with a 10 month delay.

Single-family population = total units x occupancy rate x Census 2000 persons per household (owner-occupied)

Multi-family population = total units x occupancy rate x Census 2000 persons per household (renter-occupied)

Population from Annexation = based on Department of Justice population assessment required for each annexation; new units after annexation use the two methods above

Other population = the On-campus University population. 4,608 were in dorms at the time of the 2000 Census, establishing a baseline to adjust from in subsequent years. I update this annually with the __________ University Office of Institutional Research. The on-campus University population was likely undercounted in the 2000 Census, but the City conservatively assumes all were counted in 2000.

The occupancy rate for single-family is updated annually based on the number of ___________ MLS listings for “single-family”, “manufactured real property” and “townhouse” with “San Marcos” identified as the City. I typically take the total number of listings and double it to account for any for sale by owner homes not listed on the MLS. I assume 100% of the units listed on the MLS, doubled, are vacant. Some of these listings are technically outside of the city limits, but I include them anyway to assure a conservative estimate. Likewise, assuming 100% are vacant and doubling the number of units listed in the MLS contributes to a conservative estimate.

The occupancy rate for multi-family is updated annually based on the Fourth Quarter __________ Multi-family Trend Report prepared by <Local Proprietary Multifamily Research Group with a good track record>.

Projections:

I hate doing this voodoo. I basically use the average annual growth rate over the last 10 years, with it plugged into a decay model to inject some realism 5-20 years out. I also show the numbers for a 1.5%, 3.0% and 5.0% growth rate with the same decay. When asked for a projection, I show all four estimate models. I've tried using the 1.0/1.5/2.0 model, which is used by our state demographer, but it hasn't performed well for our city over the last 10 years or so (they suck at dealing with explosive growth patterns in metro areas).
SR, I'm suprised San Marcos isn't in the ACS. And wow, that is a pretty conservative method...avoiding 50k, are we? I wouldve assumed you'd have passed that already with all the growth happening in Kyle and Buda.

I have a queston about your estimates. Do you think that your city's avg persons per household remains the same as it was in 2000? What about the city's ethnic breakdown? I would assume San Marcos is experiencing an increase in hispanic population (like most of our state) and would therefore have an increased persons per household figure. This is an issue that I have been thinking about quite a bit and can't really figure out how to address in a politically-correct or statistically-correct manner.

Finally, as for your projections...what is your reasoning for using the exponential decay on your annual percent increase (I'm assuming that's what you mean, you decay the percentage for each year into the future)? Do you foresee a slowing in growth in San Marcos over the next few years?

Again, thanks everyone for your responses!

10. Originally posted by FueledByRamen
DetroitPlanner, thanks for the great response! Its amazing (but not suprising) that Detroit has lost 90k people in the last 8 years...

Just to summarize (and correct me if I'm wrong), you use Census 2000 data as a baseline, updated with building permits and demolitions, and apply vacancy rates and average houshold sizes (by housing type and extrapolated from the 1990-2000 %age change) to gain the total population estimate. Right? I like this method, especially with your ammendments to the Census Bureau method.

How long, would you estimate, does it take you to produce the numbers for your monthly updates?

Unfortunately, I can't answer these things. I work in a totally different area of planning. As I mentioned, there are cases where using permits and demolitions do not work (this is particularly true in quickly depopulating areas or in areas with lots of speculative building this makes vacancy rates difficult to calculate.)

From what I understand through casual conversation, you are spot on.

11. Originally posted by FueledByRamen

SR, I'm suprised San Marcos isn't in the ACS. And wow, that is a pretty conservative method...avoiding 50k, are we? I wouldve assumed you'd have passed that already with all the growth happening in Kyle and Buda.

I have a queston about your estimates. Do you think that your city's avg persons per household remains the same as it was in 2000? What about the city's ethnic breakdown? I would assume San Marcos is experiencing an increase in hispanic population (like most of our state) and would therefore have an increased persons per household figure. This is an issue that I have been thinking about quite a bit and can't really figure out how to address in a politically-correct or statistically-correct manner.

Finally, as for your projections...what is your reasoning for using the exponential decay on your annual percent increase (I'm assuming that's what you mean, you decay the percentage for each year into the future)? Do you foresee a slowing in growth in San Marcos over the next few years?

Again, thanks everyone for your responses!

We hadn't *officially* crested 50,000 in Census pop estimates in 2005, so we won't be in the ACS until after 2010.

We weren't avoiding 50,000--we just knew that certain neighboring cities would challenge our estimates and we wanted to be extra safe should we land in court. I think the last time I ran the estimate it was probably around 3% undercount in reality.

PPH I think has remained pretty stable (we're low due to young college population). I can't see it going lower than it is. The age has far more to do with our PPH than ethnicity. Hispanic ethnicity was already high in 2000, so I don't suspect hispanic growth will affect it that much, particularly with the university growing so quickly with young, childless students.

Projections... The exponential decay is applied to the annual percentage for each successive year. I should have clarified that I don't begin applying the decay until after the five year projection. This is mainly because once you get about 10 years out, the numbers start looking really crazy (hard population projection numbers become a parabolic curve going up). I don't see the growth in population going down (it will probably continue to increase some), but I don't think a nearly 5% annual growth rate is sustainable for a long period of time. Think about it: 10% annual growth might be normal for a city like San Marcos (add about 5,000 people--not that scary). Applying a 10% annual growth rate to a city like Austin in one year would give the City Council a nosebleed, as they'd be adding about 90,000 folks. Growth rates tend downward as cities increase in size, so that is why I add a decay (it's not huge though).

San Marcos is an interesting city. It's far enough from both Austin & San Antonio, and has enough key employers like the university, that it functions pretty clearly as its own economy. It makes doing demography and economic stuff here very interesting.

12. Originally posted by FueledByRamen
Well, I looked forward to your response because I figured my situation would be similar to yours (private sector, having to come up with data for many different geographies and rather quickly). Correct me if I'm wrong, but isn't the ESRI data just ACS estimates GIS-ified? (I have the white paper, but I haven't read it yet)
I have not looked into the ESRI sources either, but I always thought that they had their own methodology. If not, then there is still the benefit of covering levels of geography not found in the ACS.

Many of the clients I work with prefer that we use "objective" sources like the state or federal estimates. These sometimes fail to capture information known locally, but then they are also more difficult to challenge.

13. ## VILLAGE

HOW CAN I HAVE GOOD FORECAST FOR POPULATION OF A VILLAGE AND ITS EMPLOUMENT? IS THERE A SCIENTIFIC METHOD?