1

Web Appendix

The Role of Heritage Connection in Consumer Valuation

Web Appendix A
Study 1: Table W1 ................................................................................................................. 2
Study 1: Table W2 Non-Sellers Excluded ........................................................................... 3
Study 1: Table W3 Robustness Checks ............................................................................... 4
Study 1: Figure W1 ............................................................................................................... 5
Web Appendix B
Study 2 Pre-Test 1: Table W4 .............................................................................................. 6
Study 2 Pre-Test 2: Table W5 .............................................................................................. 7
Study 2: Table W6 ................................................................................................................. 8
Study 2 Follow-up: Table W7 ............................................................................................... 9
Web Appendix C Appendix Study 1: Description and Table W8 .......................................... 10
Web Appendix D
Study 3: Table W9 Consistent ............................................................................................ 12
Study 3: Table W10 All ....................................................................................................... 13
Study 3 Follow-up: Table W11 ........................................................................................... 14
Study 3: Figure W2 ............................................................................................................. 15
Web Appendix E Study 4: Table W12 ...................................................................................... 16
Web Appendix F
Appendix Study 2: Description and Table W13 ............................................................... 17
Appendix Study 2: Figure W3 ............................................................................................ 19
Appendix Study 3: Description and Table W14 ............................................................... 20
Appendix Study 4: Description and Table W15 ............................................................... 22
Appendix Study 4: Figure W4 ............................................................................................ 28
Web Appendix G Study 5
Study 5: Table W16 ............................................................................................................. 29
Study 5: Table W17 ............................................................................................................. 30

Katherine L. Christensen and Suzanne B. Shu. Katherine L. Christensen is the corresponding

author kachris@iu.edu. These materials have been supplied by the authors to aid in the
understanding of their paper. The AMA is sharing these materials at the request of the authors.
 2

Web Appendix A Study 1: Table W1 - All Participants with Consistent Preferences

Dependent variable:
% Discount
(1) (2) (3) (4)
In Group vs. -0.112*** -0.063* -0.057* -0.055*
Heritage (0.025) (0.026) (0.026) (0.026)
Sentimental vs. -0.169*** -0.133*** -0.060+ -0.057
Heritage (0.025) (0.026) (0.036) (0.036)
Control vs. Heritage -0.184*** -0.116*** -0.056 -0.053
(0.025) (0.028) (0.040) (0.040)
Heritage Loss -0.029*** -0.030*** -0.030***
(0.006) (0.006) (0.006)
In Group Measure 0.020** 0.021**
(0.007) (0.007)
Sentimental Measure -0.009** -0.010**
(0.003) (0.003)
Age 0.001
(0.001)
Gender -0.005
(0.009)
Income -0.005+
(0.003)
Original WTA 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)
Constant 0.308*** 0.269 ***
0.234 ***
0.232***
(0.018) (0.019) (0.024) (0.024)
Observations 475 475 475 475
R2 0.127 0.171 0.198 0.206
Adjusted R2 0.119 0.163 0.186 0.189
Residual Std. Error 0.193 (df = 470) 0.188 (df = 469) 0.185 (df = 467) 0.185 (df = 464)
F Statistic 17.03*** 19.41*** 16.44*** 12.04***
*** ** * +
Note: p< 0.001, p< 0.01, p< 0.05, p< 0.10. Participants indicated their WTA for a car from
a buyer (% Discount on Original WTA), who was connected to or disconnected from the good’s
heritage, part of the seller’s family (in group), or had a high sentimental connection to the car
(sentimental value). Heritage loss, In group, Sentimental, Age, Income are mean-centered.
Gender is coded such that -1 is male and 1 is female, all other answers and non-response
recoded as 0. Standard errors presented in parentheses.
 3

Web Appendix A Study 1: Table W2 - Non-Sellers Excluded (N = 465)

Dependent variable:
% Discount
(1) (2) (3) (4)
In Group vs. -0.065** -0.034 -0.031 -0.029
Heritage (0.022) (0.024) (0.024) (0.024)
Sentimental vs. -0.122*** -0.100*** -0.053+ -0.050
Heritage (0.023) (0.023) (0.032) (0.032)
Control vs. Heritage -0.134*** -0.091*** -0.054 -0.053
(0.022) (0.025) (0.035) (0.035)
Heritage Loss -0.019*** -0.021*** -0.020***
(0.005) (0.005) (0.005)
In Group Measure 0.013* 0.014*
(0.006) (0.006)
Sentimental Measure -0.007* -0.007*
(0.003) (0.003)
Age 0.000
(0.001)
Gender 0.005
(0.008)
Income -0.006*
(0.002)
Original WTA 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)
Constant 0.261*** 0.237 ***
0.215 ***
0.213***
(0.016) (0.017) (0.021) (0.021)
Observations 465 465 465 465
R2 0.092 0.119 0.137 0.149
Adjusted R2 0.084 0.110 0.124 0.130
Residual Std. Error 0.168 (df = 460) 0.166 (df = 459) 0.164 (df = 457) 0.164 (df = 454)
F Statistic 11.61*** 12.41*** 10.38*** 7.95***
*** ** * +
Note: p< 0.001, p< 0.01, p< 0.05, p< 0.10. Participants indicated their WTA for a car from
a buyer (% Original WTA), who was connected to or disconnected from the good’s heritage,
part of the seller’s family (in group), or had a high sentimental connection to the car
(sentimental value). Heritage loss, In group, Sentimental, Age, Income are mean-centered.
Gender is coded such that -1 is male and 1 is female, all other answers, non-response recoded as
0. Standard errors presented in parentheses.
 4

Web Appendix A Study 1: Table W3 - Additional Tests

Table of Means
Heritage In Group In Group Sentimental
Variable
Loss (General) (Specific) Value
3.55 5.34 5.97 5.47
Heritage Buyer
(1.70) (1.49) (.93) (1.92)
5.27*** 5.17 5.44** 4.68*
In Group Buyer
(1.55) (1.37) (1.19) (3.02)
4.82*** 5.29 2.34*** 5.34
Sentimental Buyer
(1.51) (1.47) (1.37) (2.45)
5.93*** 5.29 1.93*** 2.91***
Control Buyer
(1.31) (1.46) (1.36) (2.66)
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05, All comparisons vs. Heritage Buyer condition.
Heritage loss is, “Compared to me as the seller, the [buyer] will feel less connected to the car
and to my grandparents that drove it.” In-Group general is, “In general, being associated with
family is an important part of my self-image.” In-Group specific is, “I feel the [buyer] is a part
of my family.” Sentimental value is, “The vintage car has sentimental value to the [buyer].”
Heritage loss results in Table A1 held controlling for in-group and sentimental value measures.
 5

Web Appendix A Study 1: Figure W1 Car and Photos in Online Photo Marketplace

“In preparing to sell this vintage Lincoln Continental you explore Kelley Blue Book to see what
similar vintage cars are currently offered at. Below is the Kelley Blue Book Information for your
car.”
 6

Web Appendix B Study 2 Pre-Test 1: Table W4 - Exploratory Factor Analysis

Heritage vs. Control on Seller’s WTA (Scaled Variables / Value)

Buyer Test
Control Buyer 𝐹(1, 308) = 8.31, p = .004, 𝜂2 = .03
Heritage Buyer 𝐹(1, 308) = .01, p = .909, 𝜂2 = .00
Heritage – Control Buyer 𝐹(1, 308) = 12.07, p < .001, 𝜂2 = .04

Exploratory Items – Item Total Correlations (α = .855)

Variable Item Total Alpha Without N
I would be satisfied with replacement. -.633 .920 310
Connects me to people I care about. .722 .829 310
Mine and no one else’s. .322 .857 310
Helps me understand who I am. .691 .832 310
Remember where I come from .806 .821 310
About more than just me. .800 .822 310
Gives me a sense of history. .813 .820 310
Connects me to traditions. .825 .819 310
Belongs to my family and not only mine. .504 .846 310
Connects me to people no longer living. .702 .829 310
Not something I should sell. .669 .833 310

Heritage Connection Measures (α=.931)

Variable Alpha Without N
Connects me to people I care about. .927 310
Remember where I come from .913 310
About more than just me. .921 310
Gives me a sense of history. .912 310
Connects me to traditions. .913 310
Connects me to people no longer living. .925 310

Additional Tests
DV Test
Ownership – Heritage vs. Control (α = .842) 𝐹(1, 308) = 2.25, p = .134, 𝜂2 = .01
Ownership – Effect of Order 𝐹(1, 308) = 3.50, p = .063, 𝜂2 = .01
Satisfied Replacement 𝐹(1, 308) = 203.78, p < .001, 𝜂2 = .40
 7

Web Appendix B Study 2 Pre-Test 2: Table W5 - Confirmatory Factor Analysis

Factor Loadings
Factor 1 Factor 2
Heritage 1 .79
Heritage 2 .86
Heritage 3 .86
Heritage 4 .89
Heritage 5 .89
Heritage 6 .92
Ownership 1 .81
Ownership 2 .93
Ownership 3 .85
SS Loadings 4.68 2.52
Prop Variance .52 .28
Cum Variance .52 .80

Correlation Matrix
Herit1 Herit2 Herit3 Herit4 Herit5 Herit6 Own1 Own2 Own3
Heritage 1 1.00 .76 .76 .72 .78 .75 .44 .35 .27
Heritage 2 .76 1.00 .80 .82 .81 .80 .43 .33 .28
Heritage 3 .76 .80 1.00 .84 .80 .82 .42 .40 .33
Heritage 4 .72 .82 .84 1.00 .84 .88 .45 .39 .32
Heritage 5 .78 .81 .80 .84 1.00 .88 .41 .38 .30
Heritage 6 .75 .80 .82 .88 .88 1.00 .38 .33 .27
Ownership 1 .44 .43 .42 .45 .41 .38 1.00 .80 .73
Ownership 2 .35 .33 .40 .39 .38 .33 .80 1.00 .81
Ownership 3 .27 .28 .33 .32 .30 .27 .73 .81 1.00

Preregistered Confirmatory Factor Analysis. We validated the measures of heritage using a

Factor Analysis with a two factor model with Varimax rotation, and a Scree plot analyzing both
ownership and heritage found two optimal coordinates with two Eigenvalues greater than the
mean. We determined that the goodness-of-fit measures were acceptable in the two factor model,
CFI = 0.98, RMSEA =0.08, SRMSR = .02, and χ2 = 47.36, df=26, p=.006.
 8

Web Appendix B Study 2: Table W6 - WTA/WTP Asymmetry

Dependent variable:
WTA Value WTP
(1) (2) (3)
Heritage vs. -7.227** 1.149*** 13.948***
Control Buyer (2.705) (0.081) (3.643)
Age 0.373*** 0.003 0.071
(0.106) (0.003) (0.143)
Gender 0.345 0.028 4.736*
(1.368) (0.041) (1.849)
Income -0.614 0.014 1.200*
(0.391) (0.012) (0.530)
Constant 54.088*** 5.163 ***
87.076***
(1.910) (0.057) (2.577)
Observations 383 766 766
R2 0.064 0.212 0.033
Adjusted R2 0.054 0.208 0.028
Residual Std. Error 26.334 (df = 378) 1.118 (df = 761) 50.408 (df = 761)
F Statistic 6.431*** 51.263*** 6.537***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In study 2, participants indicated their WTA for a set of
chimes from a friend of a friend, who was connected to or disconnected from the good’s
heritage. All participants also assessed the valuation and WTP of both a heritage buyer and a
control buyer (order was randomized). Age is mean-centered (there was no significant
interaction between age and condition). Gender is coded such that -1 is male and 1 is female.
Heritage vs. control buyer is dummy coded (1, 0). Standard errors are presented in parentheses.
Column 2 and 3 results are presented in an OLS regression for simplicity, but results are similar
in a linear mixed model regression.
 9

Web Appendix B Study 2 Follow-up: Table W7 - Buyer Value and WTP

Dependent variable:
Value WTP
(1) (2) (3) (4)
Heritage vs. 0.175 0.164 1.734 1.580
Control Buyer (0.124) (0.121) (4.455) (4.471)
0.007 0.055
Age
(0.005) (0.193)
0.280*** 2.267
Gender
(0.062) (2.294)
0.003 0.286
Income
(0.018) (0.681)
5.423*** 5.483*** 81.045*** 81.564***
Constant
(0.087) (0.086) (3.142) (3.184)
Observations 400 400 400 400
R2 0.005 0.061 0.0004 0.004
Adjusted R2 0.002 0.052 -0.002 -0.006
Residual Std. Error 1.239 1.208 44.553 44.646
F Statistic 1.998 6.463*** 0.151 0.371
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In this follow-up to study 2, participants who were
connected or disconnected from the good’s heritage indicated their valuation of and WTP for a
set of chimes from a friend of a friend. Age and Income are mean-centered. Gender is coded
such that -1 is male and 1 is female, all other answers and non-response recoded as 0. Heritage
vs. control buyer is dummy coded (1, 0). Standard errors are presented in parentheses.
 10

Web Appendix C Appendix Study 1: Selling Previously Purchased Heritage Goods

Appendix study 1 tests an alternative explanation to study 2. While buyer WTP could not

explain our effect in study 2, it seemed possible that the effect of heritage connection might

depend on a kind of reciprocity across consumers, where one consumer who has inherited or

been given an item feels an obligation to reduce the price to the next consumer because of a

similar heritage connection.

To test this alternative explanation, we re-ran study 2 (N = 399, average age = 41.2 years;

53.4% female), and 390 participants with consistent preferences were included according to the

pre-registered plan. This time we stipulated that the purchase of the chimes had been a market

transaction, “You purchased this set of Shenandoah chimes a long time ago,” before describing

the chimes. We found that sellers who had purchased their heritage goods had a lower WTA for

heritage buyers (MHeritage = $37.44, SDHeritage = $24.04) than for control buyers (MControl = $47.66,

SDControl = $25.46, t(387) = 4.08, p < .001, η² = .04), and this replication of our finding in study 2

also held using a non-parametric Wilcoxon rank sum test (W = 23684, p < .001). As a robustness

check, we tested the strength of this finding by adding two additional exclusions. We excluded

both 16 participants who gave the item away for free (NControl = 5, NHeritage = 11) and 27

participants who would not sell at any price (NControl = 18, NHeritage = 9). After removing non-

sellers, these results hold in both a simple t-test (t(343) = 3.35, p < .001, η² = .03) and in a non-

parametric Wilcoxon rank sum test (W = 18342, p < .001). See preregistration for more details:

https://aspredicted.org/qj3z9.pdf.
 11

Web Appendix C Appendix Study 1: Table W8 - Previously Purchased Heritage Goods

Dependent variable:
WTA
(1) (2) (3) (4)
Heritage vs. Control -10.225*** -10.005 ***
-7.497*** -7.348***
(2.507) (2.514) (2.237) (2.240)
+
Age 0.190 0.142
(0.101) (0.092)
Gender 1.260 1.540
(1.269) (1.133)
Income -0.062 0.014
(0.371) (0.329)
Constant 47.663*** 47.441*** 44.532*** 44.348***
(1.769) (1.779) (1.584) (1.591)
Observations 390 390 347 347
R2 0.041 0.053 0.032 0.043
Adjusted R2 0.039 0.043 0.029 0.032
Residual Std. Error 24.759 24.707 20.832 20.794
F Statistic 16.629*** 5.338 ***
11.236*** 3.885***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In this follow-up to study 2, participants indicated their
WTA for a set of their own, purchased chimes from a friend of a friend, who was connected to
or disconnected from the good’s heritage. Columns 1, 2 include all participants (except 9
participants with inconsistent preferences) while columns 3, 4 include only participants who
agreed to sell at one of the sale prices provided. Age is mean-centered. Gender is coded such
that -1 is male and 1 is female, 0 is all other answers and non-answer. Heritage is dummy coded
(1,0). Standard errors are presented in parentheses.
 12

Web Appendix D Study 3: Table W9 - Mediation by Heritage Loss - Consistent

Dependent variable:
Heritage Loss Appropriate WTA
(1) (2) (3) (4)
Heritage vs. -2.054*** 0.001 -73.345*
Control Buyer (0.163) (0.143) (31.938)
68.829***
Heritage Loss
(7.662)
0.005 0.009 1.434 1.166
Age
(0.007) (0.006) (1.342) (1.230)
0.173* -0.036 18.686 5.279
Gender
(0.083) (0.073) (16.234) (14.959)
-0.015 0.030 -5.634 -4.857
Income
(0.024) (0.021) (4.659) (4.269)
2.587*** 5.855*** 849.784*** 813.938***
Constant
(0.115) (0.101) (22.572) (14.702)
Observations 396 396 396 396
R2 0.299 0.017 0.023 0.179
Adjusted R2 0.291 0.006 0.013 0.171
Residual Std. Error 1.619 1.42 317.4 290.9
F Statistic 41.61*** 1.643 2.298+ 21.34***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In study 3, heritage participants indicated their WTA
from a heritage vs. control buyer. All participants also assessed their own heritage connection to
the good and the buyer’s heritage connection (heritage loss is heritage buyer – heritage seller,
mean-centered). They also assessed the appropriateness of the usage plans of both the heritage
and control buyer (order was randomized). Age, Income are mean-centered. Gender is coded
such that -1 is male and 1 is female, all other answers and non-response recoded as 0. Heritage
vs. control buyer is dummy coded (1, 0). Standard errors are presented in parentheses. Four
participants with inconsistent preferences were not included, but effects of condition and heritage
loss hold with these participants included.

Pre-screening questions designed to detect bots were answered on a Likert scale:

“When someone is smiling which of these is most likely how they feel?” (1-Sad, 7-Happy).
“When someone is crying which of these is most likely how they feel?” (1-Sad, 7-Happy).
“If someone has a winning lottery ticket do they feel good or bad?” (1-Bad, 7-Good).
“On a very bright and completely sunny day how does the sky look?” (1-Not Cloudy, 7-Very
Cloudy). There was an additional question that was not used to filter participants, “Money is
important to me.” (1-Strongly disagree, 7-Strongly agree).
 13

Web Appendix D Study 3: Table W10 - Mediation by Heritage Loss - All

Dependent variable:
Heritage Loss Appropriate WTA
(1) (2) (3) (4)
Heritage vs. -2.050*** 0.006 -81.076*
Control Buyer (0.162) (0.142) (32.303)
71.303***
Heritage Loss
(7.744)
0.005 0.009 1.520 1.212
Age
(0.007) (0.006) (1.354) (1.239)
0.164* -0.037 12.196 -.719
Gender
(0.082) (0.072) (16.426) (15.094)
-0.015 0.030 -6.496 -5.617
Income
(0.024) (0.021) (4.713) (4.309)
2.575*** 5.851*** 847.718*** 808.868***
Constant
(0.115) (0.100) (22.902) (14.859)
Observations 400 400 400 400
R2 0.297 0.016 0.024 0.184
Adjusted R2 0.290 0.006 0.014 0.175
Residual Std. Error 1.617 1.416 322.7 295.1
F Statistic 41.77*** 1.614 2.411* 22.2***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In study 3, heritage participants indicated their WTA
from a heritage vs. control buyer. All participants also assessed their own heritage connection
to the good and the buyer’s heritage connection (heritage loss is heritage buyer – heritage
seller, mean-centered). They also assessed the appropriateness of the usage plans of both the
heritage and control buyer (order was randomized). Age, Income are mean-centered. Gender
is coded such that -1 is male and 1 is female, all other answers and non-response recoded as 0.
Heritage vs. control buyer is dummy coded (1, 0). Standard errors are presented in parentheses.
Four participants with inconsistent preferences were not included, but effects of condition and
heritage loss hold with these participants included.
 14

Web Appendix D Study 3 Follow-up: Table W11 - Control Sellers

Dependent variable:
WTA
(1) (2) (3) (4)
Heritage vs. -15.946 -20.261 -17.702 -22.996
Control Buyer (36.422) (36.577) (36.717) (36.846)
2.280 2.407
Age
(1.647) (1.655)
-37.581* -39.089*
Gender
(18.426) (18.555)
-7.412 -7.664
Income
(5.878) (5.895)
386.530*** 390.755*** 386.530*** 390.889***
Constant
(25.818) (25.798) (25.897) (25.848)
Observations 201 201 199 199
R2 0.001 0.031 0.001 0.033
Adjusted R2 -0.004 0.011 -0.004 0.013
Residual Std. Error 258.2 256.2 259.0 256.7
F Statistic .192 1.556 .232 1.673
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In this follow-up to study 3, control participants indicated
their WTA from a connected (heritage) vs. disconnected (control) buyer. Age, Income are mean-
centered. Gender is coded such that -1 is male and 1 is female. Heritage vs. control buyer is
dummy coded (1, 0). Standard errors are presented in parentheses. Columns 1-2 include all
participants, and 3-4 include only participants with consistent preferences.
 15

Web Appendix D Study 3: Figure W2 - Watch Stimulus

 16

Web Appendix E Study 4: Table W12 - Selling a Heritage or Control Good

Dependent variable:
WTA Heritage Loss WTA
(1) (2) (3) (4)
-2.046* 4.103 ***
3.829***
Seller (Heritage vs. Control)
(0.833) (0.829) (0.463)
-5.833*** 1.755 +
0.268
Buyer (Heritage vs. Control)
(0.821) (0.922) (0.515)
***
-12.898 -2.818***
Seller : Buyer
(1.202) (0.671)
0.758***
Heritage Loss
(0.204)
0.000 0.000
Age
(0.021) (0.034)
-0.005 0.043
Gender
(0.301) (0.489)
12.928*** 9.654*** 1.414*** 6.647***
Constant
(0.726) (0.607) (0.338) (0.763)
Observations 123 123 123 123
R2 0.337 0.667 0.415 0.105
Adjusted R2 0.326 0.653 0.400 0.082
Residual Std. Error 4.527 3.249 1.820 5.282
F Statistic 30.49*** 46.86*** 28.12*** 4.65**
*** ** * +
Note: p< 0.001, p< 0.01, p< 0.05, p< 0.10. In study 4, heritage vs. control sellers indicated
their WTA from a heritage vs. control buyer. All participants also revealed their own heritage
connection to the good, “Through this (these) [university] pennant (cookies), I feel connected to
[university alums] who came before me.” If indicated a heritage connection to the good (i.e.,
they somewhat agreed, agreed, or strongly agreed with that statement), their heritage loss was the
single item measure, “Compared to me as the seller, the [buyer] will feel less connected to prior
generations through the [item].” If they did not indicate a heritage connection to the good, their
heritage loss was coded as 0. The effect of heritage condition on both WTA and heritage loss is
significant, and heritage loss increases WTA. Age is mean-centered. Gender is coded such that -
1 is male and 1 is female, any other answer or non-answer is coded as 0. Standard errors are
presented in parentheses.
 17

Web Appendix F Appendix Study 2: Heritage Buyers of Heritage Goods

To conceptually replicate the study 4 results in the lab, we ran three additional studies

(Appendix studies 2-4). The first of these follow-up studies was a conceptual replication of the

finding in study 4 that heritage sellers had a lower WTA for heritage buyers than for control

buyers (a “heritage discount”). We endowed 400 sellers from a large West Coast University

(“heritage sellers”) with a good connected to the heritage of the university (a commemorative

heritage magnet that symbolized 100 years of the university’s history). After exclusions due to

inconsistent preferences and one participant who was not eligible due to age, the final sample

size included 387 heritage sellers. We calculated heritage sellers’ WTA from a heritage buyer (a

university student) or a control buyer (a stranger with no university connection) using a Becker

DeGroot Marschak task with offers between $0 and $6 in $.25 increments. If participants would

not sell at the maximum offer ($6.00), we assigned them a WTA of $6.25. Heritage goods (or

credit to a university account) were available at the conclusion of the study.

We found that heritage sellers had a lower WTA from a heritage buyer ($2.20) than from

a control buyer ($2.63, F(1,385) = 6.05, p = .014, η² = .02), and this held in a Wilcoxon rank sum

test (W = 21044, p = .034) and when excluding participants who did not accept the highest offer

(NHeritage = 6, NControl = 11) or gave the item away for free (NHeritage = 6, NControl = 11), F(1,351) =

7.15, p = .008, η² = .02, W = 17889, p = .015. Heritage loss mediated the effect of condition on

WTA (95% C.I. : -$1.08, -$.58). This holds after removing non-sellers (95% C.I. : -$.93, -$.49).

The direct effect of buyer’s identity on WTA was ns (95% C.I. = -.01, .78) and this holds with

non-sellers excluded (95% C.I. = -.09, .64), suggesting full mediation by heritage loss (see pre-

registration at https://aspredicted.org/mr43j.pdf).
 18

Web Appendix F Appendix Study 2: Table W13 - Heritage Buyers of Heritage Goods

Dependent variable:
WTA Heritage Loss WTA
(1) (2) (3)
Heritage vs. -0.432* -1.665***
Control Buyer (0.176) (0.130)
Heritage Loss 0.428***
(0.054)
Age 0.005 -0.016 0.013
(0.016) (0.012) (0.015)
Gender -0.209* 0.049 -0.227*
(0.102) (0.075) (0.095)
Constant 2.731*** 1.012*** 2.437***
(0.133) (0.099) (0.095)
Observations 387 387 387
R2 0.027 0.308 0.151
Adjusted R2 0.019 0.302 0.144
Residual Std. Error (df = 383) 1.723 1.274 1.609
F Statistic (df = 3; 383) 3.490** 56.773*** 22.631***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. In this follow-up to study 4, participants
indicated their WTA from a heritage vs. control buyer. All participants also revealed
their WTA and evaluated both their own heritage connection to the good and the
buyer’s heritage connection to the good. The effect of heritage condition on both
WTA and heritage loss is significant, and heritage loss increases WTA. Duplicates
(n = 4), one participant unable to give informed consent due to age, and participants
with inconsistent preferences (n = 12) were not included. Age is mean-centered.
Gender is coded such that -1 is male and 1 is female. Heritage is dummy coded (1,0).
Standard errors are presented in parentheses.
 19

Web Appendix F Appendix Study 2 Figure W3: Heritage Buyers of Heritage Goods
 20

Web Appendix F Appendix Study 3: Heritage Buyers of Control Goods

A second follow-up to study 4 (N = 135, average age = 20.7 years; 27.4% female) was a

conceptual replication of the finding in study 5 that control sellers do not provide a discount to

heritage buyers. We performed this second follow-up study at a large Midwest university. As

pre-registered, we removed participants who had inconsistent preferences (n = 14), leaving 121

participants, but results hold if we include all participants’ first agreed upon sale price. This

study of control sellers was similar to Appendix study 2 with one key difference: the endowed

good was not a heritage good that connected sellers across time, but a university-branded

chocolate bar that could be eaten.

This follow-up study replicated our results with control sellers in study 4. As in study 4,

this study did not reveal any evidence of a discount for heritage buyers of a control good.

Instead, we found that control sellers had a directionally higher WTA from a heritage buyer

($2.40) than from a control buyer ($2.12), although this result was not significant in either a

standard regression (F(1,119) = 1.00, p = .319) or a non-parametric Wilcoxon rank sum test (W

= 1588, p = .219), and this null result is robust to participants who would not sell at the highest

available offer price (2 in each condition) or who gave the item away for free (3 in each

condition) in a linear regression (F(1,109) = 1.47, p = .228) or a non-parametric Wilcoxon rank

sum test (W = 1294.5, p = .155). Please see: https://aspredicted.org/qq6mr.pdf.

 21

Web Appendix F Appendix Study 3: Table W14 - Heritage Buyers of Control Goods

Dependent variable:
WTA Heritage WTA
(1) (2) (3)
Heritage vs. 0.282 -0.492+
Control Buyer (0.285) (0.256)
Buyer’s Heritage 0.137
(0.101)
Age 0.123 0.019 0.110
(0.094) (0.084) (0.093)
Gender -0.128 -0.089 -0.142
(0.159) (0.143) (0.157)
Constant 2.072*** 3.409*** 1.758***
(0.200) (0.181) (0.358)
Observations 121 121 121
2
R 0.029 0.033 0.036
2
Adjusted R 0.005 0.008 0.012
Residual Std. Error (df = 117) 1.543 1.390 1.537
F Statistic (df = 3; 117) 1.182 1.329 1.475
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. Participants indicated WTA from heritage vs.
control buyer. In a deviation from the preregistration, we measured buyer’s heritage
connection but not seller’s heritage connection for this control good. The effect of
heritage condition on WTA is ns, effect of buyer’s heritage connection on WTA is ns.
Control buyers seemed marginally more connected to the heritage of a control good.
Age mean-centered, Gender (-1 male, +1 female). Heritage vs. control dummy coded
(1,0). Standard errors presented in parentheses. As noted in pre-registration participants
with inconsistent preferences (n = 14) were not included in the study.
 22

Web Appendix F Appendix Study 4: Testing for Usage, Similarity, Favoritism

To investigate the effect of a buyer’s perceived heritage connection on the seller’s

willingness to accept for a heritage good, uses a Becker-DeGroot-Marschak elicitation method

similar to appendix studies 2 and 3, and we randomly selected one participant to receive the

results of their choice. Our goal in this study was to replicate the heritage discount and examine

if this reduced WTA would hold after controlling for buyer usage.

All participants included in the study were residents of the state of California, and the

endowed heritage good represented California’s history. We offered participants the opportunity

to sell their heritage good, and we manipulated the connection of the buyer to the good’s heritage

(high vs. low). Finally, we assessed the role of buyers’ valuation of heritage goods in

participants’ trading decisions.

Method

In Appendix study 4, we recruited 400 California residents on Amazon’s Mechanical

Turk (48.3% female, average age = 36.2 years, 366 participants after inconsistent preferences

were removed) to participate in an incentive compatible experiment related to California’s

history. The study ran online during the 250th anniversary of a city in California that is home of

California's first constitution, California's first scenic highway, and the first place where

Californians raised the American flag. The city of Monterey also has a famous aquarium, and

visitors come to see it for reasons unrelated to heritage.

We showed California sellers a heritage good (a 250th anniversary commemorative hat

that celebrated the history of California), and we told them a heritage (control) buyer wanted to

buy their heritage item. In the heritage condition, the heritage buyer was connected to the

heritage of California (resident of Monterey, CA). In the control condition, the control buyer was
 23

disconnected from that history (visitor to Monterey, CA). To elicit sellers’ WTA, we used the

Becker-DeGroot-Marschak task (prices from $0-$10, $.50 increments), and informed all

participants that a price would be chosen randomly and if they were selected (and had indicated a

willingness to sell at that price), they would give up the heritage good and receive the money

instead. On the other hand, if they had indicated that they were not willing to sell at that price,

they would keep the heritage good. After the study finished, Qualtrics randomization software

generated a participant number and an offer price between $0 and $10, and we presented that

offer to the winning participant. If the winning participant had chosen to keep the heritage good

at the offer price, they received the heritage good (the good could be sent to them from the

anniversary website without experimenters collecting personal data from subjects). If the

participant had chosen to sell at the offer price, they received the offer as a deposit in their

Amazon Mechanical Turk account.

After eliciting participants’ WTA for their heritage good, we asked participants how

much value they thought the buyer would get from having the hat (Likert Scale: 1- Low Value,

7-High Value), their likelihood to wear the hat in public, wear the hat in private, collect the hat,

their likelihood to sell the hat, and give the hat away (all measured on a Likert Scale 1-Not

Likely, 7-Very Likely). We control for buyer usage intent in our analysis.

Results

WTA. We predicted that sellers’ WTA from the control buyer would be higher than their

WTA from the heritage buyer, and this is what we found (MControl = $5.67, SDControl = $3.68 vs.

MHeritage = $4.79, SDHeritage = $3.55, F(1,364) = 5.4, p = .021, η² = .01), and this held using the

non-parametric Wilcoxon rank sum test (W = 19046, p = .022). In this model, participants who

would not sell at any price up to $10.00 were assigned a WTA of $10.50, but results are robust to
 24

excluding these participants (NControl = 30, NHeritage = 22) from the sample in a linear regression

(F(1,312) = 4.1, p = .045, η² = .01) and Wilcoxon rank sum test (W = 13916, p = .046).

High Perceived Valuation of Heritage Buyers and Lower WTA. As expected in traditional

economic theory, when sellers’ WTA increased as perceived buyer value increased (B = .45, p <

.001, η² = .03). Sellers perceived that heritage buyers valued the good more than control buyers

(MHeritage = 5.22, SDHeritage = 1.36 vs. MControl = 4.71, SDControl = 1.54, F(1,364) = 11.14, p < .001,

η² = .03). Nonetheless, heritage buyers received a lower WTA (p = .021) than control buyers.

Exploratory Mediation. While the total effect of buyer type (heritage vs. control) on

WTA is negative and reduces WTA (95% C.I. = -1.62, -.15), we found a positive, indirect effect

of buyer type (heritage vs. control) on WTA through higher perceived value (.26, 95% C.I. = .08,

.49, Hayes 2012). This indicates competitive mediation through value, and this result is robust to

removing participants who would not sell at the highest price (.20, 95% C.I. = .03, .42), and

removing participants who gave the item away for free (.24, 95% C.I. = .06, .48, Hayes 2012).

Buyer Usage. We found support for differences in usage between buyers, but these

differences in usage could not fully differences in WTA. Heritage buyers were perceived to be

more likely to wear the good in public (B = .56, p < .001), more likely to wear it in private (B =

.38, p = .032), more likely to collect the hat (B = .35, p = .023), and less likely to sell it (B = -.53,

p = .004). There was no difference in buyers’ likelihood to give the hat away (p = .277). Of these

differences, only likelihood to collect and likelihood to sell affected WTA (B = .28, p = .033, B =

.26, p = .015). Since heritage buyers were higher in likelihood to collect, and lower in likelihood

to sell, buyer usage does not offer a clear prediction on the price charged to heritage buyers, and

the effect of heritage connection is robust when controls for buyer usage and demographics are

added to the model (B = -.89, p = .019, η² = .02). Please see: https://aspredicted.org/3n4wk.pdf.

 25

Post Test of Similarity of Buyer and Seller. Participants might reveal a willingness to give

discounts to people who seem similar to them, so we ran a post-test to examine perceived

similarity between California overall and Monterey residents who also live in California

(Winterich, Mittal, and Ross Jr. 2011). California residents indicated lower levels of similarity

with a Monterey resident (M = 3.72, SD = 1.61) than with a Monterey visitor (M= 4.69, SD =

1.60, t(210) = 7.12, p < .001). Thus, similarity between California sellers and Monterey residents

(vs. visitors) could not explain sellers’ lower WTA for heritage buyers.

Post Test of In-Group Favoritism. Although similarity between buyer and seller could not

explain our effect, there are other valid alternative explanations for our results. Perhaps the

sellers simply liked the heritage buyer, the Monterey resident, better than the control buyer, and

felt that they belonged to the same group. To test this alternative explanation, we ran a post-test

with these participants, and assessed sellers’ willingness to do a favor for “a person who has

lived in (visited) Monterey, CA.” We told them that participants would “like to hear about any

recommendations you have of books to read about California. As a favor, please list any

recommendations you have for them.” Participants provided a similar level of recommendations

to Monterey residents (M= 1.45, SD = 1.19) as to Monterey visitors (M = 1.35, SD = 1.02, t(131)

= .53, p = .596). There was no difference in participants’ likelihood to do a favor across

conditions, χ2(1, N = 133) = .26, p = .609, https://aspredicted.org/9wt8e.pdf.

Discussion

The results of this follow-up to Appendix 4 suggest that shared heritage connection

between buyers and sellers impacts market transactions for goods that connect consumers to a

shared past. When sellers connected to California’s heritage indicated their WTA for a 250 year

anniversary heritage good from a control buyer less connected to the heritage of the good, sellers
 26

charged more. Crucially, we found that this price hike occurred despite the fact that sellers

expected that connected buyers received greater value from the object and thus might be

objectively willing to pay higher prices. This effect persisted after controlling for buyer usage.

Further, in follow-up studies with participants who completed Appendix study 4, we found no

evidence of higher levels of perceived similarity between California sellers and heritage buyers.

We also found no difference in sellers’ willingness to do a favor for residents vs. visitors. In sum,

using real goods and real transactions, we found that sellers discounted their WTA for heritage

buyers, and that prior theories of buyer usage, buyer value, buyer similarity, and in-group

favoritism could not fully explain sellers’ lower WTA.

 27

Web Appendix F Appendix Study 4: Table W15 - Heritage Good, Usage, Value

Dependent variable:
WTA
(1) (2) (3) (4) (5) (6)
Heritage vs. -0.877* -1.135** -0.986* -0.890*
Control Buyer (0.378) (0.376) (0.381) (0.379)
***
Value 0.441 0.507*** 0.466** 0.474***
(0.127) (0.128) (0.142) (0.140)
Wear in Public -0.011 -0.067 -0.070
(0.152) (0.152) (0.151)
Wear in Private -0.135 -0.140 -0.150
(0.124) (0.122) (0.122)
Sell 0.292* 0.228* 0.226*
(0.113) (0.112) (0.111)
Collect 0.329* 0.195 0.190
(0.141) (0.145) (0.144)
Give Away -0.093 -0.082 -0.081
(0.130) (0.128) (0.126)
Age -0.034*
(0.015)
Gender -0.321
(0.186)
*** *** *** *** ***
Constant 5.671 5.235 5.799 5.235 5.725 5.675***
(0.266) (0.187) (0.263) (0.188) (0.264) (0.262
Observations 366 366 366 366 366 366
2
R 0.015 0.032 0.056 0.036 0.077 0.100
2
Adjusted R 0.012 0.029 0.051 0.023 0.059 0.078
3.613 3.581 3.542 3.593 3.526 3.491
Residual Std. Error
(df = 364) (df = 364) (df = 363) (df = 360) (df = 358) (df = 356)
F Statistic 5.394** 12.044*** 10.714*** 2.722** 4.266*** 4.415***
*** ** *
Note: p< 0.001, p< 0.01, p< 0.05. Study 1 is a between-participant study with
heritage and control conditions. Value, age, and all usage variables are mean-centered.
Gender is coded such that -1 is male and 1 is female. Heritage is dummy coded (1,0).
Standard errors are presented in parentheses. As noted in pre-registration participants
with inconsistent preferences (n = 34) were not included in the study.
 28

Web Appendix F Appendix Study 4: Figure W4 - Heritage Good, WTA, Perceived Value

Heritage Good Evaluated by Participants

 29

Web Appendix G Study 5: Table W16 - Buying a Heritage Good on Facebook Marketplace

The table below provides the table for sellers sent a message on Facebook Marketplace by an RA

blind to study hypothesis.

Sales by Condition and Region

Seller Location Message Sale No Sale
New York Heritage 42% 58%
New York Control 29% 71%
Midwest Heritage 47% 53%
Midwest Control 34% 66%
 30

Web Appendix G Study 5: Table W17 - Effect of Messaging Type on Likelihood to Sell

The table below provides the table for sellers sent a message on Facebook Marketplace by an RA

blind to study hypothesis.

Effect of Heritage vs. Control on Seller’s Likelihood to Sell

Buyer Test
Heritage vs. Control χ2 (1) = 7.17, p = .007, φ = .13
Heritage vs. Control (New York) χ2 (1) = 3.69, p = .055, φ = .14
Heritage vs. Control (Midwest) χ2 (1) = 3.51, p = .061, φ = .13