Linking Landlords: An Examination of the Methods and Impacts of Linking Administrative Ownership Data

Forrest Hangen

1966

 

"Defining the ownership of slum tenements is a far from easy task"

 

566 Parcels
"Defining the ownership of slum tenements is a far from easy task"

 

566 Parcels

Developments since 1966

Computational Advances

  • Computing Power
  • Text-matching algorithms
  • Digitized Open(ish) Data

Computational Advances

  • Computing Power
  • Text-matching algorithms
  • Digitized Open(ish) Data

"Defining the ownership of rental housing is still no easy task!"

-Forrest after reading Sternlieb

Developments since 1966

Computational Advances

  • Computing Power
  • Text-matching algorithms
  • Digitized Open(ish) Data

Ownership Obscurity

  • Rise of the LLC

Developments since 1966

AIMS

Accurately link landlords to their properties to Uncover and Quantify the effects of Ownership Obscurity

Data

Property Tax Assessment Records

  • Owner's Name
  • Owner's Mailing Address

 

Boston, MA

2004-2019

Corporation Filing Records

  • LLC Name
  • Individuals' Names
  • Individuals' Addresses

Challenges

Messy Data

Intentional Obscurity

Elmer G Gill

Elmr G Gill
Greystone LLC

Eagle LLC

Challenges

Messy Data

Intentional Obscurity

2 Different Solutions

Challenges

Messy Data

Intentional Obscurity

Deliberate Strategy

Limiting Liability

Messy Data

Solution:

1. Text Cleaning and Standardization

L.L.C., Lmtd liability co,llc           LLC
Fifty-eight Green Street L.L.C.


58 Green st llc
Fifty-eight Green Street L.L.C.


58 Green st llc

Solution:

1. Text Cleaning and Standardization

L.L.C., Lmtd liability co,llc           LLC
Fifty-eight Green Street L.L.C.


58 Green st llc
58 GREEN ST LLC


58 GREEN ST LLC

Messy Data

Solution:

2. Fuzzy Matching

Elmer G Gill          Elmr G Gill
Glynn Realty Assoc IV LLC                 Glynn Realty Association I LLC

1. Text Cleaning and Standardization

L.L.C., Lmtd liability co,llc           LLC

Messy Data

Fuzzy Matching

Glynn Realty Assoc IV LLC          Glynn Realty Association II LLC

(Probabilistic Matching)

By Hand

( small-scale)

Out-of-the-Box

( Dedupe, OpenRefine, matchit, etc.)

Custom Probabilistic Matching

651,612 owners
Elmer G Gill          Elmer H Gill

Custom Probabilistic Matching

Out-of-the-Box

( Dedupe)

Elmer G Gill          Elmer H Gill

Pairs of Names

Glenn Ross 87 Blue Ave

Glenn Ross 4 East St

[' Gl','Gly','lyn','ynn','nn ','n R',' Re','Rea','eal','alt','lty','ty ' ...]

3-gram

"  A"  "  Ad" " Al" " As" "El " "Bar" "Llc" . . .
Glynn Realty Assoc IV LLC 0 0 0 .178 0 0 .199  . . .
 . . .  . . .  . . .  . . .  . . .  . . .  . . .  . . .  . . .

tf-idf

Custom Probabilistic Matching

Glynn Realty Assoc IV LLC          Glynn Realty Association II LLC

Custom Probabilistic Matching

Glynn Realty Assoc IV LLC          Glynn Realty Association II LLC

Cosine Similarity

cos(\theta) = \frac{\sum_{i=1}^{n}{A_{i}B_{i}}}{\sqrt{\sum_{i=1}^{n}{A_{i}^{2}}}\sqrt{\sum_{i=1}^{n}{B_{i}^{2}}}} 
measures similarity between two entities using the cosine of the angle between their two vectors in a multidimensional space

Is all this work worth it?

Raw Tax Assessment

Cleaning & Standardization

Fuzzy-Matching

4.8%

7.1%

12.3%

Owners

Raw Tax Assessment

Cleaning & Standardization

Fuzzy-Matching

24.5%

27.8%

33.7%

Units

Greystone LLC

Eagle LLC
Not done yet...

Intentional Obscurity

Greystone LLC

Eagle LLC
Greystone LLC
Eagle LLC

Corporation Filing Records

Intentional Obscurity

Greystone LLC

Eagle LLC
Greystone LLC
Eagle LLC

Corporation Filing Records

Fuzzy Matching

Intentional Obscurity

Intentional Obscurity

Intentional Obscurity

Intentional Obscurity

Intentional Obscurity

55 GREYSTONE LLC
EAGLE LLC
12 GREEN ST LLC

Tax Assessment

55 GREYSTONE LLC
EAGLE LLC
12 GREEN ST LLC

Corporation Records

Tax Assessment

55 GREYSTONE LLC
EAGLE LLC
12 GREEN ST LLC
Greystone Realty Corporation
Unique ID

Tax Assessment

55 GREYSTONE LLC
EAGLE LLC
12 GREEN ST LLC
Unique ID

  • Landlord Size

  • Corporate Status

  • Obscurity

  • Geographic Scope

  • Property Types

  • Property-level Outcomes

Tax Assessment

Raw Tax Assessment

4.8%

Owners

Raw Tax Assessment

Fuzzy-Matching

24.5%

Units

Fuzzy-Matching

12.3%

33.7%

+11.9%

+17.5%

Corporate Data

16.7%

Corporate Data

42.0%

Is all this work worth it?

Is all this work really worth it?

Ok, but ...

YES!

Intentional Obscurity

Consolidation

Herfindahl-Hirschman Index (HHI)
HHI = \sum_{i=1}^{N}{S_i^{2}}
where S = a landlord's share of an outcome

Units

Eviction Filings

Herfindahl-Hirschman Index (HHI)
HHI = .16^{2}+.16^{2}+.16^{2}+.16^{2}+.16^{2}+.16^{2}
HHI =0.166
HHI = .5^{2}+.33^{2}+.16^{2}
HHI =0.388

Consolidation of Units

0-No consolidation to 1-One owner for all properties

Consolidation of Units

Consolidation of Evictions

Consolidation of Evictions

Largest Owners of 10% of Units

All other owners

506

Evictions

3,092

Evictions

/ 16k units

/ 151k units

+49%

+32%

Takeaways

  • Difficulty in revealing ownership has shifted from messy data to intentional obscurity
  • Using Corporate Data reveals ownership obscurity
  • Ownership Obscurity hides new scales of consolidation and responsibility for evictions & property conditions

Thank You

Forrest Hangen

 

 

hangen.f@northeastern.edu

@forresthangen@sciences.social