Use of Logsum

12 posts / 0 new
Last post
xuwyvytec
Use of Logsum

Hello All,How do analysts use logsum in travel demand modeling and as an evaluation measure in transportation planning/modeling? If anyone could share any document that details how it is used, it would be greatly appreciated.Thank you!Bob

jabraham

I was just looking at Tomas de la Barra’s book Integrated Land Use and Transport Modelling: Decision Chains and Hierarchies Cambridge University Press 1989 and liked the way the composite utility benefit measures were described in it. In fact, I just used it in a PECAS training session today.

Someone told me this book is available for free on Tomas’s website, but I can’t find it right now. The book reference is here:

https://www.cambridge.org/core/books/integrated-land-use-and-transport-m...

Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision Chains and Hierarchies (Cambridge Urban and Architectural Studies). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359

Composite utility logsum measures are under utilized in my opinion. One reason, perhaps, is because a lot of benefits get captured by landlords into increased rents in the medium term, and then developers may respond shifting land use in the long term. So, you need an integrated land use and transport model to start to understand who’s REALLY going to benefit through time. Without a representation of supply/demand spatial economic equilibrium, it could be misleading.

But, even in a demand-only model, a logsum measure is a lot better than travel time savings measures, which can have the wrong sign entirely if your transportation project opens up economic opportunities to people and they travel more to take advantage of them. So, even in a demand-only model, it’s a lot less misleading than some of the other benefit measures people commonly use.

Hope this helps,

--
John Abraham
jea@hbaspecto.com
403-232-1060

> On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
>
> Hello All,How do analysts use logsum in travel demand modeling and as an evaluation measure in transportation planning/modeling? If anyone could share any document that details how it is used, it would be greatly appreciated.Thank you!Bob
>
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470

SuzanneChildress

Hi Bob,
Here is a link to the activity-based model primer that discusses logsums
extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf

I personally think that logsums are not good performance measures because
they are not intuitive to planners, let alone the general public. Here's
when I was ruminating on that topic:
http://psrc.github.io/2015/accessibility But I'd be interested in hearing
other people's views.

Suzanne

On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:

> I was just looking at Tomas de la Barra’s book Integrated Land Use and
> Transport Modelling: Decision Chains and Hierarchies Cambridge University
> Press 1989 and liked the way the composite utility benefit measures were
> described in it. In fact, I just used it in a PECAS training session today.
>
> Someone told me this book is available for free on Tomas’s website, but I
> can’t find it right now. The book reference is here:
>
> https://www.cambridge.org/core/books/integrated-land-use-and-transport-m...
>
>
> Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
>
> Composite utility logsum measures are under utilized in my opinion. One
> reason, perhaps, is because a lot of benefits get captured by landlords
> into increased rents in the medium term, and then developers may respond
> shifting land use in the long term. So, you need an integrated land use and
> transport model to start to understand who’s REALLY going to benefit
> through time. Without a representation of supply/demand spatial economic
> equilibrium, it could be misleading.
>
> But, even in a demand-only model, a logsum measure is a lot better than
> travel time savings measures, which can have the wrong sign entirely if
> your transportation project opens up economic opportunities to people and
> they travel more to take advantage of them. So, even in a demand-only
> model, it’s a lot less misleading than some of the other benefit measures
> people commonly use.
>
> Hope this helps,
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> >
> > Hello All,How do analysts use logsum in travel demand modeling and as an
> evaluation measure in transportation planning/modeling? If anyone could
> share any document that details how it is used, it would be greatly
> appreciated.Thank you!Bob
> >
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>

kkockelm@mail.u...

Like John, I’m a big fan of logsum differences, normalized. A student & I will be providing NCTCOG (& our website) with a tool, for anyone to use in spring 2021. It’ll show all sorts of AIs, side by side, based on standard inputs. One can evaluate the value of added access to one neighborhood vs another, and compare such welfare impacts to project costs, as done in this new paper: https://www.caee.utexas.edu/prof/kockelman/public_html/TRB21NairobiAIs.pdf.

Here’s the comment I posted after Suzanne’s blog note, which refers to other papers, comparing AIs & endogenizing long-term location choice in the logsums, as well as a holistic toolkit (which uses logsums for welfare evaluations, that are monetized alongside emissions & crash impacts of MPO & DOT projects):

I recommend that we use both simple and complex access metrics (so stakeholders can see easy to interpret & more rigorous measures, side by side), and that we are sure to use normalized differences in (nested) logsums for welfare calculations (e.g., before vs after land use patterns or travel times & costs are changed). The normalization is simply dividing differences by the marginal utility of money (which, of course, varies by income, education, & other features of the traveler).

By the way, we find that logsums correlate well with land values and simple/naive AIs (see http://www.ce.utexas.edu/prof/kockelman/public_html/TRB02RentsAccess.pdf, for example). And perfectly correct logsums would probably include the location-choice utilities as well (& rents, etc.). But it's complex to do: https://www.caee.utexas.edu/prof/kockelman/public_html/TRB15homelocation....

Here's an open-source, well-documented Project Evaluation Toolkit or "PET" for more holistic evaluations of land use & transportation changes for planners & modelers to use: https://www.caee.utexas.edu/prof/kockelman/PET_Website/homepage.htm. PET delivers benefit-cost ratios that include such nested logsum calculations (for traveler welfare impacts, recognizing mode, destination, & route choices), crash and emissions impacts (as well as toll revenue returns of investment, etc.). Let us know if you need help in application!

A useful PET-based paper is this one: https://www.caee.utexas.edu/prof/kockelman/public_html/TRB11Toolkit.pdf.

PS Thanks to all those who participated in the Bridging Transportation Researchers conference on Tuesday & Weds! We had >800 registrants and >80 speakers. It went off without a hitch, thanks to stellar speakers, moderators, Tech Staff (aka grad students :-) ), & all our excellent audience members.
From: childresssuzanne=gmail.com@mg.tmip.org On Behalf Of SuzanneChildress
Sent: Thursday, August 13, 2020 2:07 PM
To: TMIP
Subject: Re: [TMIP] Use of Logsum

Hi Bob,
Here is a link to the activity-based model primer that discusses logsums
extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf

I personally think that logsums are not good performance measures because
they are not intuitive to planners, let alone the general public. Here's
when I was ruminating on that topic:
http://psrc.github.io/2015/accessibility But I'd be interested in hearing
other people's views.

Suzanne

On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:

> I was just looking at Tomas de la Barra’s book Integrated Land Use and
> Transport Modelling: Decision Chains and Hierarchies Cambridge University
> Press 1989 and liked the way the composite utility benefit measures were
> described in it. In fact, I just used it in a PECAS training session today.
>
> Someone told me this book is available for free on Tomas’s website, but I
> can’t find it right now. The book reference is here:
>
> https://www.cambridge.org/core/books/integrated-land-use-and-transport-m...
>
>
> Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
>
> Composite utility logsum measures are under utilized in my opinion. One
> reason, perhaps, is because a lot of benefits get captured by landlords
> into increased rents in the medium term, and then developers may respond
> shifting land use in the long term. So, you need an integrated land use and
> transport model to start to understand who’s REALLY going to benefit
> through time. Without a representation of supply/demand spatial economic
> equilibrium, it could be misleading.
>
> But, even in a demand-only model, a logsum measure is a lot better than
> travel time savings measures, which can have the wrong sign entirely if
> your transportation project opens up economic opportunities to people and
> they travel more to take advantage of them. So, even in a demand-only
> model, it’s a lot less misleading than some of the other benefit measures
> people commonly use.
>
> Hope this helps,
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> >
> > Hello All,How do analysts use logsum in travel demand modeling and as an
> evaluation measure in transportation planning/modeling? If anyone could
> share any document that details how it is used, it would be greatly
> appreciated.Thank you!Bob
> >
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>
--
Full post: https://tmip.org/content/use-logsum
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470

Mark Bradley

A few years ago, RSG and EcoNorthwest completed a project for FHWA researching the use of various types of measures, including logsums, for benefit cost analysis (BCA). As part of that project, we created a tool called BCA4ABM, which is open source software that we have used for both trip-based models and activity-based models. It is programmed using the same Python-based scripting approach that is being used to create the ActivitySim open source activity-based model platform.

The webpage at http://rsginc.github.io/bca4abm/resources.html has links to the project report and various slide/poster presentations, as well as the code repository and instructions and examples for use.

Mark
………………………………
Mark Bradley
Senior Director
RSG
mark.bradley@.rsginc.com

From: childresssuzanne=gmail.com@mg.tmip.org On Behalf Of SuzanneChildress
Sent: Thursday, August 13, 2020 12:08 PM
To: TMIP
Subject: Re: [TMIP] Use of Logsum

CAUTION - EXTERNAL EMAIL

Hi Bob,
Here is a link to the activity-based model primer that discusses logsums
extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf

I personally think that logsums are not good performance measures because
they are not intuitive to planners, let alone the general public. Here's
when I was ruminating on that topic:
http://psrc.github.io/2015/accessibility But I'd be interested in hearing
other people's views.

Suzanne

On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:

> I was just looking at Tomas de la Barra’s book Integrated Land Use and
> Transport Modelling: Decision Chains and Hierarchies Cambridge University
> Press 1989 and liked the way the composite utility benefit measures were
> described in it. In fact, I just used it in a PECAS training session today.
>
> Someone told me this book is available for free on Tomas’s website, but I
> can’t find it right now. The book reference is here:
>
> https://www.cambridge.org/core/books/integrated-land-use-and-transport-m...
>
>
> Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
>
> Composite utility logsum measures are under utilized in my opinion. One
> reason, perhaps, is because a lot of benefits get captured by landlords
> into increased rents in the medium term, and then developers may respond
> shifting land use in the long term. So, you need an integrated land use and
> transport model to start to understand who’s REALLY going to benefit
> through time. Without a representation of supply/demand spatial economic
> equilibrium, it could be misleading.
>
> But, even in a demand-only model, a logsum measure is a lot better than
> travel time savings measures, which can have the wrong sign entirely if
> your transportation project opens up economic opportunities to people and
> they travel more to take advantage of them. So, even in a demand-only
> model, it’s a lot less misleading than some of the other benefit measures
> people commonly use.
>
> Hope this helps,
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> >
> > Hello All,How do analysts use logsum in travel demand modeling and as an
> evaluation measure in transportation planning/modeling? If anyone could
> share any document that details how it is used, it would be greatly
> appreciated.Thank you!Bob
> >
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>
--
Full post: https://tmip.org/content/use-logsum
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470

winufuwub

Hi Mark and TMippers,

I had written to Bob, privately, about his potential interest in looking at
the original paper that proposes the log-sum (inclusive value, composite
cost) by Prof. Huw Williams, back in the 70s when I started studying my PhD
in Leeds:

Williams, H.C.W.L. (1977) On the formation of travel demand models and
economic evaluation measures of user benefit. *Environment and Planning*
*9A*, 285–344.

I also mentioned that we discuss the measure in "*Modelling Transport*"
(2011, Chapters 5, 6 and mainly 7 and 8), and also that Prof. Jara-Díaz
does it for the case of evaluation in his book "*Transport Economic Theory*"
of 2007.

However, I forgot to mention that it could also be very interesting to look
at Chapter 4 and mainly at the start of Chapter 5, of the wonderful
following book (that ought to be read by everybody interested in the
history of our field)

Boyce, D.E. & Williams, H.C.W.L. (2015) Forecasting Urban Travel: Past,
Present and Future. Edward Elgar, Cheltenham.

Cheers to all

Juan de Dios

J. de D. Ortúzar
Emeritus Professor
Department of Transport Engineering and Logistics
Pontificia Universidad Catolica de Chile
www.ing.puc.cl/jos

El vie., 14 ago. 2020 a las 9:08, Mark Bradley ()
escribió:

> A few years ago, RSG and EcoNorthwest completed a project for FHWA
> researching the use of various types of measures, including logsums, for
> benefit cost analysis (BCA). As part of that project, we created a tool
> called BCA4ABM, which is open source software that we have used for both
> trip-based models and activity-based models. It is programmed using the
> same Python-based scripting approach that is being used to create the
> ActivitySim open source activity-based model platform.
>
> The webpage at http://rsginc.github.io/bca4abm/resources.html has links
> to the project report and various slide/poster presentations, as well as
> the code repository and instructions and examples for use.
>
> Mark
> ………………………………
> Mark Bradley
> Senior Director
> RSG
> mark.bradley@.rsginc.com
>
> From: childresssuzanne=gmail.com@mg.tmip.org On Behalf Of SuzanneChildress
> Sent: Thursday, August 13, 2020 12:08 PM
> To: TMIP
> Subject: Re: [TMIP] Use of Logsum
>
> CAUTION - EXTERNAL EMAIL
>
> Hi Bob,
> Here is a link to the activity-based model primer that discusses logsums
> extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf
>
> I personally think that logsums are not good performance measures because
> they are not intuitive to planners, let alone the general public. Here's
> when I was ruminating on that topic:
> http://psrc.github.io/2015/accessibility But I'd be interested in hearing
> other people's views.
>
> Suzanne
>
> On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:
>
> > I was just looking at Tomas de la Barra’s book Integrated Land Use and
> > Transport Modelling: Decision Chains and Hierarchies Cambridge University
> > Press 1989 and liked the way the composite utility benefit measures were
> > described in it. In fact, I just used it in a PECAS training session
> today.
> >
> > Someone told me this book is available for free on Tomas’s website, but I
> > can’t find it right now. The book reference is here:
> >
> > https://www.cambridge.org/core/books/integrated-land-use-and-transport-m
> ...
> >
> >
> > Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> > Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> > Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
> >
> > Composite utility logsum measures are under utilized in my opinion. One
> > reason, perhaps, is because a lot of benefits get captured by landlords
> > into increased rents in the medium term, and then developers may respond
> > shifting land use in the long term. So, you need an integrated land use
> and
> > transport model to start to understand who’s REALLY going to benefit
> > through time. Without a representation of supply/demand spatial economic
> > equilibrium, it could be misleading.
> >
> > But, even in a demand-only model, a logsum measure is a lot better than
> > travel time savings measures, which can have the wrong sign entirely if
> > your transportation project opens up economic opportunities to people and
> > they travel more to take advantage of them. So, even in a demand-only
> > model, it’s a lot less misleading than some of the other benefit measures
> > people commonly use.
> >
> > Hope this helps,
> >
> > --
> > John Abraham
> > jea@hbaspecto.com
> > 403-232-1060
> >
> > > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> > >
> > > Hello All,How do analysts use logsum in travel demand modeling and as
> an
> > evaluation measure in transportation planning/modeling? If anyone could
> > share any document that details how it is used, it would be greatly
> > appreciated.Thank you!Bob
> > >
> > > --
> > > Full post: https://tmip.org/content/use-logsum
> > > Manage my subscriptions: https://tmip.org/mailinglist
> > > Stop emails for this post:
> > https://tmip.org/mailinglist/unsubscribe/13470
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> >
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>

pschwa

We at FRA have also looked at the use of logsum to support benefit-cost analysis, having encountered (as Mr. Abraham references below) situations where travel-time savings estimates of user benefits lead to counter-intuitive results – e.g. mode shift being associated with negative travel time savings benefits. This situation can arise not only where there is significant induced demand, but also where mode shift is being driven by lower monetary cost to travelers, rather than a decrease in travel time (e.g. through the introduction of a lower-cost, but potentially slower travel option).

To address this, we have been looking at adopting a willingness-to-pay oriented (i.e. surplus-based) approach to BCA, rather than one that uses the tallying of individual types of benefits (including travel time savings) which seems more prevalent at the moment. As part of this, the change in consumer surplus is calculated using a logsum transformation of the basic utility functions that underpin the demand forecasting model. Where things get a bit tricky is in the need to calculate corresponding changes to producer surplus, which involves making some brave assumptions regarding the marginal costs of the various transportation modes, but even this is not insurmountable.

Here are links to a couple of versions of a paper that covers the subject fairly well.

https://www.rand.org/content/dam/rand/pubs/working_papers/2005/RAND_WR27...
https://www.researchgate.net/publication/222404182_The_logsum_as_an_eval...

All the best,

Peter Schwartz

From: mark.bradley=rsginc.com@mg.tmip.org On Behalf Of Mark Bradley
Sent: Friday, August 14, 2020 9:06 AM
To: TMIP
Subject: Re: [TMIP] Use of Logsum

CAUTION: This email originated from outside of the Department of Transportation (DOT). Do not click on links or open attachments unless you recognize the sender and know the content is safe.

A few years ago, RSG and EcoNorthwest completed a project for FHWA researching the use of various types of measures, including logsums, for benefit cost analysis (BCA). As part of that project, we created a tool called BCA4ABM, which is open source software that we have used for both trip-based models and activity-based models. It is programmed using the same Python-based scripting approach that is being used to create the ActivitySim open source activity-based model platform.

The webpage at http://rsginc.github.io/bca4abm/resources.html has links to the project report and various slide/poster presentations, as well as the code repository and instructions and examples for use.

Mark
………………………………
Mark Bradley
Senior Director
RSG
mark.bradley@.rsginc.com

From: childresssuzanne=gmail.com@mg.tmip.org On Behalf Of SuzanneChildress
Sent: Thursday, August 13, 2020 12:08 PM
To: TMIP
Subject: Re: [TMIP] Use of Logsum

CAUTION - EXTERNAL EMAIL

Hi Bob,
Here is a link to the activity-based model primer that discusses logsums
extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf

I personally think that logsums are not good performance measures because
they are not intuitive to planners, let alone the general public. Here's
when I was ruminating on that topic:
http://psrc.github.io/2015/accessibility But I'd be interested in hearing
other people's views.

Suzanne

On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:

> I was just looking at Tomas de la Barra’s book Integrated Land Use and
> Transport Modelling: Decision Chains and Hierarchies Cambridge University
> Press 1989 and liked the way the composite utility benefit measures were
> described in it. In fact, I just used it in a PECAS training session today.
>
> Someone told me this book is available for free on Tomas’s website, but I
> can’t find it right now. The book reference is here:
>
> https://www.cambridge.org/core/books/integrated-land-use-and-transport-m...
>
>
> Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
>
> Composite utility logsum measures are under utilized in my opinion. One
> reason, perhaps, is because a lot of benefits get captured by landlords
> into increased rents in the medium term, and then developers may respond
> shifting land use in the long term. So, you need an integrated land use and
> transport model to start to understand who’s REALLY going to benefit
> through time. Without a representation of supply/demand spatial economic
> equilibrium, it could be misleading.
>
> But, even in a demand-only model, a logsum measure is a lot better than
> travel time savings measures, which can have the wrong sign entirely if
> your transportation project opens up economic opportunities to people and
> they travel more to take advantage of them. So, even in a demand-only
> model, it’s a lot less misleading than some of the other benefit measures
> people commonly use.
>
> Hope this helps,
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> >
> > Hello All,How do analysts use logsum in travel demand modeling and as an
> evaluation measure in transportation planning/modeling? If anyone could
> share any document that details how it is used, it would be greatly
> appreciated.Thank you!Bob
> >
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>
--
Full post: https://tmip.org/content/use-logsum
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
--
Full post: https://tmip.org/content/use-logsum
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470

jabraham

I have a new-found confidence in the public's ability to interpret model results. We just convinced most people to stay home for months based on a simple model of disease transmission! (Granted, maybe not as well in the USA as elsewhere, but still!) I also have a new-found confidence in the public's ability to understand the notion of uncertainty in model results, and the ability to choose a policy direction incorporating uncertainty, since the early COVID models said "woah, we don't know what's going to happen, but it looks like it could be VERY VERY BAD" and the majority of the public was saying "wow, that sounds scary, tell us what to do".

When we were doing the training session in Atlanta the other day, we (and by "we" I mean Doug Hunt) simply explained, with some examples, that the model is based on observed behaviour regarding willingness-to-pay. You could give someone $3 and make them three dollars happier, but, we've also observed the choices made by people to avoid a $3 parking charge, so, that behaviour must make them just as happy, if they're choosing rationally. We run through some observed examples in housing markets, economic flow location choices (destination choice) route choice, etc. We run through a couple of examples of how travel-time-savings can be completely wrong. Eventually, in group sessions with a planning audience, people seem to be just fine with dollar-equivalent composite utility measures based on top-level logsums, especially after they learn that other, simpler, measures can be completely wrong.

What we've been doing in PECAS work is provide a system for people to drill-down into the benefit measures. So, if something seems a little counterintuitive, people can drill down by zone, or by interaction, and figure out something like "oh, it's because housing prices have gone up", or, "people are saving travel time but it's because they have less jobs to choose from", or "wow, it's because we forgot to plan for a hospital up in this high-growth area". All benefit measures in PECAS are normalized to dollar-equivalents (cost coefficients are always 1.0, we use the McFadden style logit nesting, not the "Daly Style" nesting) so everything is comparable. As we drill down into details we have to rely more-and-more on rule of one-half and less-and-less on logsums, so things become less precise, but, we always have the top level logsums for comparison, so we get a sense of the imprecision.

Sure, Suzanne you are right, people would rather have a simpler measure than top-level logsum composite utility comparisons. People would also rather the world was simple enough to not even have to use models. The main reason we model at all is to represent things that are more complex than "common sense". Since simpler measures can be wrong, I think we should report them only after checking them against the top-level logsum benefit measures. Fortunately, the basis behind the way logsum composite utility measures work can be explained fairly easily, with a few trade-off examples, and without diving deep into the math, so personally, I only go into simpler measures if necessary after drilling down and exploring the more correct measures.

--
John Abraham
jea@hbaspecto.com
403-232-1060

> On Aug 13, 2020, at 1:07 PM, SuzanneChildress wrote:
>
> Hi Bob,
> Here is a link to the activity-based model primer that discusses logsums
> extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf
> I personally think that logsums are not good performance measures because
> they are not intuitive to planners, let alone the general public. Here's
> when I was ruminating on that topic:
> http://psrc.github.io/2015/accessibility But I'd be interested in hearing
> other people's views.
>
> Suzanne
>
> On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:
>
> > I was just looking at Tomas de la Barra’s book Integrated Land Use and
> > Transport Modelling: Decision Chains and Hierarchies Cambridge University
> > Press 1989 and liked the way the composite utility benefit measures were
> > described in it. In fact, I just used it in a PECAS training session today.
> >
> > Someone told me this book is available for free on Tomas’s website, but I
> > can’t find it right now. The book reference is here:
> >
> > https://www.cambridge.org/core/books/integrated-land-use-and-transport-m ...
> >
> >
> > Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> > Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> > Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
> >
> > Composite utility logsum measures are under utilized in my opinion. One
> > reason, perhaps, is because a lot of benefits get captured by landlords
> > into increased rents in the medium term, and then developers may respond
> > shifting land use in the long term. So, you need an integrated land use and
> > transport model to start to understand who’s REALLY going to benefit
> > through time. Without a representation of supply/demand spatial economic
> > equilibrium, it could be misleading.
> >
> > But, even in a demand-only model, a logsum measure is a lot better than
> > travel time savings measures, which can have the wrong sign entirely if
> > your transportation project opens up economic opportunities to people and
> > they travel more to take advantage of them. So, even in a demand-only
> > model, it’s a lot less misleading than some of the other benefit measures
> > people commonly use.
> >
> > Hope this helps,
> >
> > --
> > John Abraham
> > jea@hbaspecto.com
> > 403-232-1060
> >
> > > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> > >
> > > Hello All,How do analysts use logsum in travel demand modeling and as an
> > evaluation measure in transportation planning/modeling? If anyone could
> > share any document that details how it is used, it would be greatly
> > appreciated.Thank you!Bob
> > >
> > > --
> > > Full post: https://tmip.org/content/use-logsum
> > > Manage my subscriptions: https://tmip.org/mailinglist
> > > Stop emails for this post:
> > https://tmip.org/mailinglist/unsubscribe/13470
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
> >
>
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470

elizabethsall

The part I don't like about logsums is that they are very dependent on the
number of choices available. Even though you want this to be the case
generally (3 good options is better than 1 option), it can also cause
strange results where logsums show a much higher level of accessibility
than is warranted in some parts of a region as a function of the options.

Things that can make things look strange fast:

- A zillion options in your mode choice model for one section of your
region that don't apply to others (i.e. scooter-to-cable-car,
bike-to-cable-car, walk-to-cable-car, tnc-to-cable-car) .
- Does my accessibility really improve just because you put a toll on a
road that used to be free and now I have "another option, the option to pay
money"?
- Different strategies for how your zones are delineated for different
parts of your region

Careful selection of mode choice options, good size terms and nesting help
but do not eliminate these issues.

These issues are particularly acute in route choice models where options
are overlapping but one could argue that many of our mode-choice models
have overlapping options as well. For some counter-intuitve logsum work
from my SFCTA tenure completed by some of our excellent interns at the
time, see:

-
https://archive.sfcta.org/sites/default/files/content/IT/SFCHAMP/PDFs/20...
- https://journals.sagepub.com/doi/abs/10.3141/2430-18

On Fri, Aug 14, 2020 at 12:09 PM jabraham wrote:

> I have a new-found confidence in the public's ability to interpret model
> results. We just convinced most people to stay home for months based on a
> simple model of disease transmission! (Granted, maybe not as well in the
> USA as elsewhere, but still!) I also have a new-found confidence in the
> public's ability to understand the notion of uncertainty in model results,
> and the ability to choose a policy direction incorporating uncertainty,
> since the early COVID models said "woah, we don't know what's going to
> happen, but it looks like it could be VERY VERY BAD" and the majority of
> the public was saying "wow, that sounds scary, tell us what to do".
>
> When we were doing the training session in Atlanta the other day, we (and
> by "we" I mean Doug Hunt) simply explained, with some examples, that the
> model is based on observed behaviour regarding willingness-to-pay. You
> could give someone $3 and make them three dollars happier, but, we've also
> observed the choices made by people to avoid a $3 parking charge, so, that
> behaviour must make them just as happy, if they're choosing rationally. We
> run through some observed examples in housing markets, economic flow
> location choices (destination choice) route choice, etc. We run through a
> couple of examples of how travel-time-savings can be completely wrong.
> Eventually, in group sessions with a planning audience, people seem to be
> just fine with dollar-equivalent composite utility measures based on
> top-level logsums, especially after they learn that other, simpler,
> measures can be completely wrong.
>
> What we've been doing in PECAS work is provide a system for people to
> drill-down into the benefit measures. So, if something seems a little
> counterintuitive, people can drill down by zone, or by interaction, and
> figure out something like "oh, it's because housing prices have gone up",
> or, "people are saving travel time but it's because they have less jobs to
> choose from", or "wow, it's because we forgot to plan for a hospital up in
> this high-growth area". All benefit measures in PECAS are normalized to
> dollar-equivalents (cost coefficients are always 1.0, we use the McFadden
> style logit nesting, not the "Daly Style" nesting) so everything is
> comparable. As we drill down into details we have to rely more-and-more on
> rule of one-half and less-and-less on logsums, so things become less
> precise, but, we always have the top level logsums for comparison, so we
> get a sense of the imprecision.
>
> Sure, Suzanne you are right, people would rather have a simpler measure
> than top-level logsum composite utility comparisons. People would also
> rather the world was simple enough to not even have to use models. The main
> reason we model at all is to represent things that are more complex than
> "common sense". Since simpler measures can be wrong, I think we should
> report them only after checking them against the top-level logsum benefit
> measures. Fortunately, the basis behind the way logsum composite utility
> measures work can be explained fairly easily, with a few trade-off
> examples, and without diving deep into the math, so personally, I only go
> into simpler measures if necessary after drilling down and exploring the
> more correct measures.
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> > On Aug 13, 2020, at 1:07 PM, SuzanneChildress wrote:
> >
> > Hi Bob,
> > Here is a link to the activity-based model primer that discusses logsums
> > extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf
> > I personally think that logsums are not good performance measures because
> > they are not intuitive to planners, let alone the general public. Here's
> > when I was ruminating on that topic:
> > http://psrc.github.io/2015/accessibility But I'd be interested in
> hearing
> > other people's views.
> >
> > Suzanne
> >
> > On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:
> >
> > > I was just looking at Tomas de la Barra’s book Integrated Land Use and
> > > Transport Modelling: Decision Chains and Hierarchies Cambridge
> University
> > > Press 1989 and liked the way the composite utility benefit measures
> were
> > > described in it. In fact, I just used it in a PECAS training session
> today.
> > >
> > > Someone told me this book is available for free on Tomas’s website,
> but I
> > > can’t find it right now. The book reference is here:
> > >
> > >
> https://www.cambridge.org/core/books/integrated-land-use-and-transport-m
> ...
> > >
> > >
> > > Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> > > Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> > > Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
> > >
> > > Composite utility logsum measures are under utilized in my opinion. One
> > > reason, perhaps, is because a lot of benefits get captured by landlords
> > > into increased rents in the medium term, and then developers may
> respond
> > > shifting land use in the long term. So, you need an integrated land
> use and
> > > transport model to start to understand who’s REALLY going to benefit
> > > through time. Without a representation of supply/demand spatial
> economic
> > > equilibrium, it could be misleading.
> > >
> > > But, even in a demand-only model, a logsum measure is a lot better than
> > > travel time savings measures, which can have the wrong sign entirely if
> > > your transportation project opens up economic opportunities to people
> and
> > > they travel more to take advantage of them. So, even in a demand-only
> > > model, it’s a lot less misleading than some of the other benefit
> measures
> > > people commonly use.
> > >
> > > Hope this helps,
> > >
> > > --
> > > John Abraham
> > > jea@hbaspecto.com
> > > 403-232-1060
> > >
> > > > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> > > >
> > > > Hello All,How do analysts use logsum in travel demand modeling and
> as an
> > > evaluation measure in transportation planning/modeling? If anyone could
> > > share any document that details how it is used, it would be greatly
> > > appreciated.Thank you!Bob
> > > >
> > > > --
> > > > Full post: https://tmip.org/content/use-logsum
> > > > Manage my subscriptions: https://tmip.org/mailinglist
> > > > Stop emails for this post:
> > > https://tmip.org/mailinglist/unsubscribe/13470
> > > --
> > > Full post: https://tmip.org/content/use-logsum
> > > Manage my subscriptions: https://tmip.org/mailinglist
> > > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> > >
> >
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>

winufuwub

Dear Elizabeth, if I understand correctly what you are saying, most of the
problems you mention are associated with an erroneous use of the logsum
measure. If you go to the principles of the nested logit model (see
Carrasco and Ortúzar, 2002 for a paper that discusses a series of
misconceptions about the model), you will see that the measure originates
as composite cost or inclusive value, when you wish to represent correlated
alternatives (i.e. alternatives that are more similar in comparison with
others) in the upper level of the hierarch, but you have to be careful with
the nesting structure. Things indeed may look strange rather fast if the
measure is applied without a correct comprehension of its origins.

In fact, it is correct that if you actually have more alternatives the
logsum increases, because in the nested logit model you assume extreme
value type I (formerly referred to as Gumbel or Weibull) distributions of
the errors, and the logsum is just the expected value of the distribution
of the maximum - and it can be shown that this value has to be greater than
or equal (if we are talking about utilities) than the means of the
distributions of each alternative. But again, you have to be very careful
with the nesting structure.

With best wishes

Juan de Dios

Carrasco, J.A. & Ortúzar, J. de D. (2002) Review and assessment of the
nested logit model.* Transport Reviews **22*, 197-218.

J. de D. Ortúzar
Emeritus Professor
Department of Transport Engineering and Logistics
Pontificia Universidad Catolica de Chile
www.ing.puc.cl/jos

El vie., 14 ago. 2020 a las 16:25, elizabethsall ()
escribió:

> The part I don't like about logsums is that they are very dependent on the
> number of choices available. Even though you want this to be the case
> generally (3 good options is better than 1 option), it can also cause
> strange results where logsums show a much higher level of accessibility
> than is warranted in some parts of a region as a function of the options.
>
> Things that can make things look strange fast:
>
> - A zillion options in your mode choice model for one section of your
> region that don't apply to others (i.e. scooter-to-cable-car,
> bike-to-cable-car, walk-to-cable-car, tnc-to-cable-car) .
> - Does my accessibility really improve just because you put a toll on a
> road that used to be free and now I have "another option, the option to pay
> money"?
> - Different strategies for how your zones are delineated for different
> parts of your region
>
> Careful selection of mode choice options, good size terms and nesting help
> but do not eliminate these issues.
>
> These issues are particularly acute in route choice models where options
> are overlapping but one could argue that many of our mode-choice models
> have overlapping options as well. For some counter-intuitve logsum work
> from my SFCTA tenure completed by some of our excellent interns at the
> time, see:
>
> -
> https://archive.sfcta.org/sites/default/files/content/IT/SFCHAMP/PDFs/20...
>
> - https://journals.sagepub.com/doi/abs/10.3141/2430-18
>
> On Fri, Aug 14, 2020 at 12:09 PM jabraham wrote:
>
> > I have a new-found confidence in the public's ability to interpret model
> > results. We just convinced most people to stay home for months based on a
> > simple model of disease transmission! (Granted, maybe not as well in the
> > USA as elsewhere, but still!) I also have a new-found confidence in the
> > public's ability to understand the notion of uncertainty in model
> results,
> > and the ability to choose a policy direction incorporating uncertainty,
> > since the early COVID models said "woah, we don't know what's going to
> > happen, but it looks like it could be VERY VERY BAD" and the majority of
> > the public was saying "wow, that sounds scary, tell us what to do".
> >
> > When we were doing the training session in Atlanta the other day, we (and
> > by "we" I mean Doug Hunt) simply explained, with some examples, that the
> > model is based on observed behaviour regarding willingness-to-pay. You
> > could give someone $3 and make them three dollars happier, but, we've
> also
> > observed the choices made by people to avoid a $3 parking charge, so,
> that
> > behaviour must make them just as happy, if they're choosing rationally.
> We
> > run through some observed examples in housing markets, economic flow
> > location choices (destination choice) route choice, etc. We run through a
> > couple of examples of how travel-time-savings can be completely wrong.
> > Eventually, in group sessions with a planning audience, people seem to be
> > just fine with dollar-equivalent composite utility measures based on
> > top-level logsums, especially after they learn that other, simpler,
> > measures can be completely wrong.
> >
> > What we've been doing in PECAS work is provide a system for people to
> > drill-down into the benefit measures. So, if something seems a little
> > counterintuitive, people can drill down by zone, or by interaction, and
> > figure out something like "oh, it's because housing prices have gone up",
> > or, "people are saving travel time but it's because they have less jobs
> to
> > choose from", or "wow, it's because we forgot to plan for a hospital up
> in
> > this high-growth area". All benefit measures in PECAS are normalized to
> > dollar-equivalents (cost coefficients are always 1.0, we use the McFadden
> > style logit nesting, not the "Daly Style" nesting) so everything is
> > comparable. As we drill down into details we have to rely more-and-more
> on
> > rule of one-half and less-and-less on logsums, so things become less
> > precise, but, we always have the top level logsums for comparison, so we
> > get a sense of the imprecision.
> >
> > Sure, Suzanne you are right, people would rather have a simpler measure
> > than top-level logsum composite utility comparisons. People would also
> > rather the world was simple enough to not even have to use models. The
> main
> > reason we model at all is to represent things that are more complex than
> > "common sense". Since simpler measures can be wrong, I think we should
> > report them only after checking them against the top-level logsum benefit
> > measures. Fortunately, the basis behind the way logsum composite utility
> > measures work can be explained fairly easily, with a few trade-off
> > examples, and without diving deep into the math, so personally, I only go
> > into simpler measures if necessary after drilling down and exploring the
> > more correct measures.
> >
> > --
> > John Abraham
> > jea@hbaspecto.com
> > 403-232-1060
> >
> > > On Aug 13, 2020, at 1:07 PM, SuzanneChildress wrote:
> > >
> > > Hi Bob,
> > > Here is a link to the activity-based model primer that discusses
> logsums
> > > extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf
> > > I personally think that logsums are not good performance measures
> because
> > > they are not intuitive to planners, let alone the general public.
> Here's
> > > when I was ruminating on that topic:
> > > http://psrc.github.io/2015/accessibility But I'd be interested in
> > hearing
> > > other people's views.
> > >
> > > Suzanne
> > >
> > > On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:
> > >
> > > > I was just looking at Tomas de la Barra’s book Integrated Land Use
> and
> > > > Transport Modelling: Decision Chains and Hierarchies Cambridge
> > University
> > > > Press 1989 and liked the way the composite utility benefit measures
> > were
> > > > described in it. In fact, I just used it in a PECAS training session
> > today.
> > > >
> > > > Someone told me this book is available for free on Tomas’s website,
> > but I
> > > > can’t find it right now. The book reference is here:
> > > >
> > > >
> > https://www.cambridge.org/core/books/integrated-land-use-and-transport-m
> > ...
> > > >
> > > >
> > > > Barra, T. (1989). Integrated Land Use and Transport Modelling:
> Decision
> > > > Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> > > > Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
> > > >
> > > > Composite utility logsum measures are under utilized in my opinion.
> One
> > > > reason, perhaps, is because a lot of benefits get captured by
> landlords
> > > > into increased rents in the medium term, and then developers may
> > respond
> > > > shifting land use in the long term. So, you need an integrated land
> > use and
> > > > transport model to start to understand who’s REALLY going to benefit
> > > > through time. Without a representation of supply/demand spatial
> > economic
> > > > equilibrium, it could be misleading.
> > > >
> > > > But, even in a demand-only model, a logsum measure is a lot better
> than
> > > > travel time savings measures, which can have the wrong sign entirely
> if
> > > > your transportation project opens up economic opportunities to people
> > and
> > > > they travel more to take advantage of them. So, even in a demand-only
> > > > model, it’s a lot less misleading than some of the other benefit
> > measures
> > > > people commonly use.
> > > >
> > > > Hope this helps,
> > > >
> > > > --
> > > > John Abraham
> > > > jea@hbaspecto.com
> > > > 403-232-1060
> > > >
> > > > > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> > > > >
> > > > > Hello All,How do analysts use logsum in travel demand modeling and
> > as an
> > > > evaluation measure in transportation planning/modeling? If anyone
> could
> > > > share any document that details how it is used, it would be greatly
> > > > appreciated.Thank you!Bob
> > > > >
> > > > > --
> > > > > Full post: https://tmip.org/content/use-logsum
> > > > > Manage my subscriptions: https://tmip.org/mailinglist
> > > > > Stop emails for this post:
> > > > https://tmip.org/mailinglist/unsubscribe/13470
> > > > --
> > > > Full post: https://tmip.org/content/use-logsum
> > > > Manage my subscriptions: https://tmip.org/mailinglist
> > > > Stop emails for this post:
> > https://tmip.org/mailinglist/unsubscribe/13470
> > > >
> > >
> > > --
> > > Full post: https://tmip.org/content/use-logsum
> > > Manage my subscriptions: https://tmip.org/mailinglist
> > > Stop emails for this post:
> > https://tmip.org/mailinglist/unsubscribe/13470
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> >
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>

SuzanneChildress

Hi John and Community,
I like your thoughts about how COVID modeling shows the capacity for the
public to understand uncertainty in modeling, while also trusting model
results to help guide decisions (at times).

Thanks for sharing your recent experience using logsums successfully to
explain holistic model outcomes. It sounds like it worked well as a measure
when you needed to zoom out, and get the big picture, and then you could
zoom back in to details. Also, it sounds like you had the time to work
through the details with the audience so they could understand nuances.
Using time-savings by itself, is indeed problematic; for example, if you
add high quality transit, people switch to it, and travel times go up.

I get worried if we try to use one measure like logsums or benefit-cost
scores by themselves because they bury our assumptions inside - such as
value-of-time for low income people or the value of a human life- and they
are limited by what we can measure and what we may not able to measure well
because of the lack of research and data.

Our team at PSRC shared some recent work about metrics at TRB and the
Transportation Planning Applications Conference, and I thought I'd share
below, since it explains where I'm coming from more. I've also attached the
poster from TRB.

This conversation has actually made me change my mind a bit (which is rare
in humans, right)!
I now think that logsums, BCA, and indices have their place, within a set
of metrics.

One final note: in the COVID modeling, honestly, I was interested in seeing
the use of quality adjusted life years (QALYs)
https://en.wikipedia.org/wiki/Quality-adjusted_life_year in more analysis
about the benefits and costs of various interventions. In much of the
popular media they were talking about a single measure like deaths or jobs
lost. So, yes, I did want a holistic measure; many people seem to get lost
in one part of the picture, and not see the full picture.

___________________________________________________________________________

PSRC has recently defined a set of travel model performance metrics that
were based on people’s lived experiences as opposed to traditional system
metrics. These metrics were defined in the last Regional Transportation
Plan through extensive conversation with the planners. The model metrics
will evolve based on shifting planning needs, including their upcoming use
in land use growth planning.

Examples of human-scale metrics are commute travel time, time spent in
congestion, miles a person walks each day, and annual household
transportation costs. PSRC modelers have found several principles that
allow for better model metrics through conversations with planners and
decision makers. These principles for better metrics include:

1. Selection of a few measures that tell a story, instead of every
possible measure

2. Use of measures that correlate with human happiness and health
such as short commutes

3. Avoidance of indices or measures that have several interlocking
layers and weights

4. Aggregation of geographic metrics to home locations as opposed to
trip ends

5. Ensuring the timescale makes the measure meaningful, such as daily
or annually

6. Analysis of how low-income people are impacted differently

7. Avoidance of translation steps such as converting time into money

8. Use of simple bar charts where appropriate, as opposed to complex
animations

The modeling team confronted a few challenges in trying to use the new
metrics. The most difficult aspect is the legacy of old metrics. Since
many measures have been used for decades, for consistency with previous
work, they are still required. The modeling team also found that we need
to improve our ability to model the locations of people with low income and
people of color. Other future work on the metrics may include improvements
better depiction of transportation costs and air quality exposure.

On Fri, Aug 14, 2020 at 9:29 AM John Abraham wrote:

> I have a new-found confidence in the public's ability to interpret model
> results. We just convinced most people to stay home for months based on a
> simple model of disease transmission! (Granted, maybe not as well in the
> USA as elsewhere, but still!) I also have a new-found confidence in the
> public's ability to understand the notion of uncertainty in model results,
> and the ability to choose a policy direction incorporating uncertainty,
> since the early COVID models said "woah, we don't know what's going to
> happen, but it looks like it could be VERY VERY BAD" and the majority of
> the public was saying "wow, that sounds scary, tell us what to do".
>
> When we were doing the training session in Atlanta the other day, we (and
> by "we" I mean Doug Hunt) simply explained, with some examples, that the
> model is based on observed behaviour regarding willingness-to-pay. You
> could give someone $3 and make them three dollars happier, but, we've also
> observed the choices made by people to avoid a $3 parking charge, so, that
> behaviour must make them just as happy, if they're choosing rationally. We
> run through some observed examples in housing markets, economic flow
> location choices (destination choice) route choice, etc. We run through a
> couple of examples of how travel-time-savings can be completely wrong.
> Eventually, in group sessions with a planning audience, people seem to be
> just fine with dollar-equivalent composite utility measures based on
> top-level logsums, especially after they learn that other, simpler,
> measures can be completely wrong.
>
> What we've been doing in PECAS work is provide a system for people to
> drill-down into the benefit measures. So, if something seems a little
> counterintuitive, people can drill down by zone, or by interaction, and
> figure out something like "oh, it's because housing prices have gone up",
> or, "people are saving travel time but it's because they have less jobs to
> choose from", or "wow, it's because we forgot to plan for a hospital up in
> this high-growth area". All benefit measures in PECAS are normalized to
> dollar-equivalents (cost coefficients are always 1.0, we use the McFadden
> style logit nesting, not the "Daly Style" nesting) so everything is
> comparable. As we drill down into details we have to rely more-and-more on
> rule of one-half and less-and-less on logsums, so things become less
> precise, but, we always have the top level logsums for comparison, so we
> get a sense of the imprecision.
>
> Sure, Suzanne you are right, people would rather have a simpler measure
> than top-level logsum composite utility comparisons. People would also
> rather the world was simple enough to not even have to use models. The main
> reason we model at all is to represent things that are more complex than
> "common sense". Since simpler measures can be wrong, I think we should
> report them only after checking them against the top-level logsum benefit
> measures. Fortunately, the basis behind the way logsum composite utility
> measures work can be explained fairly easily, with a few trade-off
> examples, and without diving deep into the math, so personally, I only go
> into simpler measures if necessary after drilling down and exploring the
> more correct measures.
>
> --
> John Abraham
> jea@hbaspecto.com
> 403-232-1060
>
> On Aug 13, 2020, at 1:07 PM, SuzanneChildress
> wrote:
>
> Hi Bob,
> Here is a link to the activity-based model primer that discusses logsums
> extensively: https://www.psrc.org/sites/default/files/shrp2_c46.pdf
>
> I personally think that logsums are not good performance measures because
> they are not intuitive to planners, let alone the general public. Here's
> when I was ruminating on that topic:
> http://psrc.github.io/2015/accessibility But I'd be interested in hearing
> other people's views.
>
> Suzanne
>
> On Thu, Aug 13, 2020 at 4:49 AM jabraham wrote:
>
> > I was just looking at Tomas de la Barra’s book Integrated Land Use and
> > Transport Modelling: Decision Chains and Hierarchies Cambridge University
> > Press 1989 and liked the way the composite utility benefit measures were
> > described in it. In fact, I just used it in a PECAS training session
> today.
> >
> > Someone told me this book is available for free on Tomas’s website, but I
> > can’t find it right now. The book reference is here:
> >
> > https://www.cambridge.org/core/books/integrated-land-use-and-transport-m
> ...
> >
> >
> > Barra, T. (1989). Integrated Land Use and Transport Modelling: Decision
> > Chains and Hierarchies (Cambridge Urban and Architectural Studies).
> > Cambridge: Cambridge University Press. doi:10.1017/CBO9780511552359
> >
> > Composite utility logsum measures are under utilized in my opinion. One
> > reason, perhaps, is because a lot of benefits get captured by landlords
> > into increased rents in the medium term, and then developers may respond
> > shifting land use in the long term. So, you need an integrated land use
> and
> > transport model to start to understand who’s REALLY going to benefit
> > through time. Without a representation of supply/demand spatial economic
> > equilibrium, it could be misleading.
> >
> > But, even in a demand-only model, a logsum measure is a lot better than
> > travel time savings measures, which can have the wrong sign entirely if
> > your transportation project opens up economic opportunities to people and
> > they travel more to take advantage of them. So, even in a demand-only
> > model, it’s a lot less misleading than some of the other benefit measures
> > people commonly use.
> >
> > Hope this helps,
> >
> > --
> > John Abraham
> > jea@hbaspecto.com
> > 403-232-1060
> >
> > > On Aug 6, 2020, at 3:39 PM, xuwyvytec wrote:
> > >
> > > Hello All,How do analysts use logsum in travel demand modeling and as
> an
> > evaluation measure in transportation planning/modeling? If anyone could
> > share any document that details how it is used, it would be greatly
> > appreciated.Thank you!Bob
> > >
> > > --
> > > Full post: https://tmip.org/content/use-logsum
> > > Manage my subscriptions: https://tmip.org/mailinglist
> > > Stop emails for this post:
> > https://tmip.org/mailinglist/unsubscribe/13470
> > --
> > Full post: https://tmip.org/content/use-logsum
> > Manage my subscriptions: https://tmip.org/mailinglist
> > Stop emails for this post:
> https://tmip.org/mailinglist/unsubscribe/13470
> >
> --
> Full post: https://tmip.org/content/use-logsum
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/13470
>
>
>

KenCervenka

Hello!
This was an interesting thread to read, but the only previous post I am copying to this email is the original one, from August 6.

For an example of MPOs turning “messy math” into something understandable to the local decision-makers, who are almost always faced with situations where the proposed investments in a region far outweigh the available funds, this document prepared by MTC is worth a full read through:
https://mtc.ca.gov/sites/default/files/ProjectPerformance_FinalFindings_...

Even though the document summarizes model-based predictions of dollar-value “benefits” for proposed projects, the actual word “logsum” is not used in the main document, instead this is mentioned in a technically-oriented document that is referenced within the main document, in case the public or decision-makers want to get a better understanding of what was actually done (which some might do only if their favored project did not rate well):
https://mtc.ca.gov/sites/default/files/ProjectPerformance_Methodology.pdf

As part of my work-related activities that date back to when I worked for an MPO (and even before that, when I was a multimodal transportation planning consultant), I have had opportunities to examine lots of documents that describe some type of “analysis of alternatives” process that ultimately results in selection of a locally-preferred alternative, and suffice to write in this post that these were often not particularly uplifting moments. It would be difficult figuring out an actually-doable work plan, but it could make for a really interesting graduate class project, Master’s Thesis, or maybe even a Dissertation (or two), to focus on how and why the “relative rankings” for different road/transit/other projects can differ, depending on the actual methodology and evaluation criteria that is used to conduct the “analysis of alternatives” process.

Ken C.

***********************************
From: robertpartick
Sent: Thursday, August 6, 2020 5:40 PM

Hello All,
How do analysts use logsum in travel demand modeling and as an evaluation measure in transportation planning/modeling? If anyone could share any document that details how it is used, it would be greatly appreciated.
Thank you!
Bob

***********************************