A flexible z-Tree and oTree implementation of the Social Value Orientation Slider Measure

Authors

Paolo Crosetto, INRA-GAEL, Institut National pour la Recherche Agronomique, Grenoble Applied Economics Laboratory, 1241 Rue des Résidences -- Domaine Universitaire, 38400 Saint Martin d'Hères, France

Ori Weisel, Coller School of Management, Organizational Behavior Department, Tel Aviv Universityv Tel Aviv, 6997801, Israel

Fabian Winter, Max Planck Institute for Research on Collective Goods, MPRG "Mechanisms of Normative Change". Kurt Schumacher Strasse 10, 53113 Bonn, Germany

Keywords

z-Tree, oTree, experimental software, Social Value Orientation, Slider Measure

JEL-Classification: C91, D03, D64

Abstract

This manual describes implementations of the SVO Slider Measure [@murphy2011measuring SVOSM hereafter] in z-Tree [@fischbacher2007z] and oTree [@chen2016otree], two of the most commonly used software packages for computerized experiments in the social sciences. The SVOSM is designed to measure social preferences on a continuous scale, and is frequently used throughout the literature. Please refer to @murphy2011measuring for further details. Both implementations can be accessed via our github repository at https://github.com/drfwint/svo.

Citation

Paolo Crosetto, Ori Weisel, Fabian Winter (2019) A flexible z-Tree and oTree Implementation of the Social Value Orientation Slider Measure, Journal of Behavioral and Experimental Finance, Volume 23, Pages 46-53,

General remarks

@murphy2011measuring present two different versions of the SVOSM: a quasi-continuous web-based measure using sliders, and a discrete paper-based version capturing the important features of the web-based version, which some researchers find easier to administer. For performance reasons, the z-Tree implementation is based on the discrete version. The oTree implementation is based on the continuous version.

Our implementation is customizable to allow the user to select different matching protocols, multiple languages, scaling payoffs, and ordering and subsets of items. Table [tab:parameters]{reference-type="ref" reference="tab:parameters"} gives an overview of the important parameters of the implementation.

Matching

A major addition to the original publication is that we add a RANDOM_DICTATOR matching protocol to the RING matching protocol used by @murphy2011measuring. Assuming four decision makers, A, B, C, and D, in the RING protocol A gives to B, B gives to C, C gives to D and D gives to A (see left side of Figure 1{reference-type="ref" reference="matching"}). Each subject is both a sender and a receiver, but in each role is matched with a different person. The RING matching works with any number of subjects greater than 1.

In the RANDOM_DICTATOR matching (right side of Figure 1{reference-type="ref" reference="matching"}), subjects A,B,C,D are matched in groups of 2 (say (A; B) and (C;D)). Each subject submits his/her choices in the SVOSM. Later, one member of each group (say A and C) is randomly selected to be the sender, and the other (B and D, respectively) is the receiver. In this case, B receives from A and D receives from C. The choices by B and D are recorded but not implemented. RANDOM_DICTATOR matching thus limits the interaction to two subjects, where only one choice is finally implemented. This may reduce chains of reciprocity, where A might give to B because she expects to be compensated by D. The RANDOM_DICTATOR matching works with any number of Subjects that is a multiple of 2.

The RANDOM_DICTATOR matching uses the strategy method; decisions are only implemented with probability 1/2. It has been argued that such decision making may be perceived as "colder", and thus may lead to more strategic [@brandts2000hot] or normative choices [@rauhut2010sociological].

width=".7\textwidth"}\

Main measures

 [[subsec:main_measures]] label="subsec:main_measures"} The svo_angle is the core measure of the SVOSM. It is calculated from the primary six items as $$\arctan\left(\frac{\texttt{mean_to_other} - 50*\texttt{scale}}{\texttt{mean_to_self} - 50*\texttt{scale}}\right)$$

[see @murphy2011measuring for a detailed discussion].

The nine secondary items are used to calculate the inequality_aversion_score, which distinguishes, for subjects classified as prosocial according to the svo_angle, between efficiency maximization and material equality motives. It is calculated as

$$\frac{\texttt{avg_dist_to_equality}}{(\texttt{avg_dist_to_equality+avg_dist_to_joint})};$$

[see @murphy2011measuring for a detailed discussion].

z-Tree implementation

The z-Tree implementation of the SVOSM is a stand-alone treatment which is easy to integrate in existing z-Tree treatments. It was created in z-Tree 3.36 and can be used as any other treatment.

After implementing a slider-based version with real-time display of the slider's position in z-Tree, we experienced serious time delays, causing the server to freeze for up to several minutes. Whether this was due to network traffic or to excessive read-and-write activities on the hard drive could not finally be settled, but slightly increased performance after switching to a solid state drive points towards the latter cause. These problems finally led us to implement a point-and-click version of the SVOSM, heavily reducing network traffic as well as hard drive usage.

Our implementation of the SVOSM provides all the features of the paper-based version of the SVOSM, except for interior decisions (i.e., choices which are in between the discrete values on the scale). Note that while such interior decisions are possible in principle, they are very rarely used in practice.¹

Parameters are set in PARAMETERS

This section describes parameters of the z-Tree treatment which can easily be changed by the experimenter to match specific needs. Standard parameters, e.g. number of subjects, are treated in the standard z-Tree way. It is not necessary nor recommended to change the number of rounds or groups. To change treatment specific parameters, open the globals program "////INSERT PARAMETERS HERE ////" (see Figure 2{reference-type="ref" reference="screenshot"}) and change the respective values described below.

language

Sets the language in which the SVOSM is displayed to the decision maker. Implemented languages are English, French, German, and Italian. To add additional languages, first declare the new language in the program INTERNAL CONSTANTS, add the translations to all items, and finally choose the language in the program PARAMETERS.

select_items

Allows you to choose whether only the six primary items or all fifteen items (six primary and nine secondary) are displayed. Using only the secondary items is not possible.

items_in_random_order

Determines whether items are displayed in the order presented in the paper-based SVOSM as in @murphy2011measuring, or in random order. If the RANDOM option is chosen, the order is randomized separately for each subject.

matching

Determines the matching procedure: RING or RANDOM_DICTATOR (see subsection [1.1] reference="subsec:matching"} and Figure 1{reference-type="ref" reference="matching"}). RING matching works with any number of subjects greater than 1. RANDOM_DICTATOR matching works with any even number of subjects.

scale

The scale of the circle underlying the SVOSM can be changed. The default is scale = 1, which results in a circle with a diameter of 100 centered at (50,50), as in the original publication. Inputs greater than 1 scale the circle up (e.g. 2 results in a circle with a diameter of 200 centered at (100,100)); inputs smaller than 1 scale the circle down (e.g. 0.5 results in a circle with a diameter of 50 centered at (25,25)). Scaling may be useful if you want to present the decisions in real monetary values and cannot afford to pay 100 €/$/CHF/$\dots$

precision

Determines the precision of the displayed values. This option is limited to two modes---integers and two decimal digits---because z-Tree does not allow for conditional layout of decimals. INTEGERS displays values as integers, which is sensible for larger numbers, e.g. a circle with a diameter of 100. For a circle with a small diameter, two decimals after the point may be more appropriate (choose option TWO_DIGITS_AFTER_POINT).

debug

Displays debugging information, including a kill-button. Useful for testing the treatment.

Output

The z-Tree treatment writes all the relevant output in the subjects-table and calculates the important measures straight away, making it possible to readily use the results in the experiment (e.g. for matching purposes) and makes data analysis more convenient, since all important measures are already in the data set. Most of the variables in the subjects table are explained in Table [[tab:output]]. Some of the output variables are only relevant under specific parameters, e.g. avg_dist_to_equality is calculated only if the secondary items are used. The rightmost column of table [[tab:output]] indicates if and when the variables are used.

Profit

The Profit-variable is built-in in z-Tree, and is automatically written to the TotalProfit-variable in the session table. Thus, you can use the profits earned in the SVOSM in later treatments. To determine the profits, one of the sliders is randomly selected and the corresponding earnings are written to the Profit-variable. If RING matching is used, Profit contains the sum of the amount received as a receiver and the amount kept as a sender. When the RANDOM_DICTATOR option is chosen, Profit depends on the subjects role: If she was chosen to be a sender, Profit returns the amount kept, and if she was chosen to be a receiver, it returns the amount sent to her by the sender.

svo_angle

The svo_angle is the core measure of the SVOSM. The svo_angle is stored in the subjects table and in the session table. Storing values in the session table is useful if some of the SVOSM information is required in other treatments later in the session.

svo_type

svo_type assigns SVO types to specific value ranges of the svo_angle. The following values are used: 1 = Altruist, 2 = Prosocial, 3 = Individualist, 4 = Competitive. Thresholds for assigining labers are taken from @murphy2011measuring. svo_type is stored in the subjects table and in the session table.

inequality_aversion_score

The inequality_aversion_score is calculated from the secondary items and is computed only if they are used. It is calculated only if the subject is classified as being "prosocial" (see subsection [subsec:main_measures] above). In all other cases it is set to -99. inequality_aversion_score is stored in the subjects table and in the session table.

oTree implementation

The oTree implementation of the SVOSM is implemented as a normal oTree-app [@chen2016otree] and can be easily integrated in larger experiments.² Different from the z-Tree version, the oTree version is based on the continuous version of the SVOSM and has a somewhat more up-to-date layout. The interface includes a bar chart to show the allocations to each player, but is fully customizable using HTML and JavaScript. When implementing the SVOSM, we tried to use as much of the standard oTree-dialect as possible, and only rely on common JavaScript libraries compatible with oTree such as HighCharts or jquery when necessary. These libraries are either already included or referenced in the source code, so there is no need to install them manually. Important parts of the code are implemented in JavaScript, so a fair amount of knowledge may be helpful to make fundamental changes. Debugging for both the oTree and JavaScript parts is possible with the internal oTree debugger (See the oTree documentation). The JavaScript parts use cookies to temporarily store information in the client's browser and issues a notification the first time the page is opened.

![Screenshot of the oTree decision screen](

[[fig:otree_screenshot]] label="fig:otree_screenshot"}

Setting the Parameters

All relevant parameters can be set either in the file settings.py or, more conveniently, directly in the oTree web interface when creating a new session. Simply click on "Sessions" in the top panel of the admin page and change the parameters described below as required. If you want to set specific defaults, change them in settings.py.

In most cases the options available to the user are the same in the oTree implementation as in the z-Tree one. In these cases we refer to the relevant sections above. Other cases are discussed below.

LANGUAGE_CODE

Sets the language in which the SVOSM is displayed to the decision maker. Implemented languages are English, German, Italian, and French. Set LANGUAGE_CODE to ’en’, ’de’, ’it’, or ’fr’. Please refer to the oTree documentation on localization to learn how to add further languages. LANGUAGE_CODE also sets the language for the cookie warning.

select_items

See section [2.1.2].

items_in_random_order

See section [2.1.3].

matching

See section [2.1.4].

scale

See section [2.1.5].

precision

See section [2.1.6].

Output

All relevant output, including the SVO-angle etc, can be downloaded in CSV or XLS format in the standard oTree way via the web browser. The data is stored internally, ready to be used within the experiment (e.g. for matching purposes). Other than these differences, all details regarding the output are the same as in the z-Tree implementation (see [2.2].

payoff

The payoff-variable is the standard oTree variable to calculate payoffs. It is highly suggested to use this variable name, because it makes e.g. the summing of payoffs over different apps easier. To determine the profits, one of the sliders is randomly selected and the corresponding earnings are written to the payoff-variable. If RING matching is used, payoff contains the sum of the amount received as a receiver and the amount kept as a sender. When the RANDOM_DICTATOR option is chosen, payoff depends on the subjects role: If she was chosen to be a sender, payoff returns the amount kept, and if she was chosen to be a receiver, it returns the amount sent to her by the sender.

svo_angle

See section [2.2.2].

svo_type

See section [2.2.3].

inequality_aversion_score

See section [2.2.4].

Disclaimer

The z-Tree treatment of the SVOSM is thoroughly tested and has been used in several labs and experiments over the years. The oTree implementation has also been tested thoroughly, but is much younger. As it has not been circulated widely so far, it should be used with somewhat greater care than the z-Tree version. Nevertheless, the authors disclaim all warranties, expressed or implied, regarding the Software, including any implied warranties of satisfactory quality, merchantability or fitness for a particular purpose. The authors shall have no liability whatsoever to the User of the Software for any direct, indirect, special or consequential loss and/or expense (including loss of profit) suffered by the User and arising out of a malfunctioning of the Software.

You can use, modify and distribute the corresponding treatments if you agree with the above points. If you use these implementations of the SVOSM, please cite the SVOSM as @murphy2011measuring and make sure to follow the license agreements associated with z-Tree or oTree (in particular to cite @fischbacher2007z, or @chen2016otree, respectively).

References

Appendix

Parameters in the z-Tree treatment

Parameter	Values	Description
language	ENGLISH	Language is English
	GERMAN	Language is German
	ITALIAN	Language is Italian
select_items	PRIMARY	only the primary items (items 1-6) are elicited.
	FULL	primary and secondary items (items 1-15) are elicited
items_in_random_order	ORDERED	items are presented according to the order in Murphy et al. (2011)
	RANDOM	items are presented in random order
matching	RING	Subject A,B,C,D are ordered on a ring-structure as in Murphy et a. (2011). In this case, A gives to B, B gives to C, C gives to D and D gives to A, which makes everyone a sender AND a receiver.
	RANDOM_DICTATOR	Subjects A,B,C,D are matched in groups of 2 (say (A; B) and (C;D). One member of each group (say A and C) is selected to be the sender, the other one as receiver. In this case, B receives from A and D receives from C.
precision	TWO_DIGITS_AFTER_POINT	values on sliders are rounded to two digits after decimal point
	INTEGERS	values on sliders are rounded to integers
scale	$(0, +\infty]$	Parameter to scale up $(>1$) or down $(<1)$ all the numbers on a slider Default is 1, resulting in a circle of diameter 100
debug	{1;0}	set to 1 to display some debug info; set to 0 while running actual sessions

Relevant Variables in the Subjects Table

Relevant Variables in the Subjects Table	Description	Relevant for Parameters
Subject	unique identifier for the Subject	always
Group	matching group of the Subject.	matching=RANDOM_DICTATOR
Profit	Profit of the Subject	always
input_self[i]	allocation to self in item i	always
input_other[i]	allocation to other in item i	always
random_order[i]	order of item i in RANDOM order	items_in_random_order = RANDOM
chosen_option[i]	chosen option on item i, counted from left to right (leftmost option = 1, rightmost option = 9)	always
mean_to_self	mean allocation to self in primary items	always
mean_to_other	mean allocation to other in primary items	always
svo_angle	svo angle calculated as $$\arctan\left(\frac{\texttt{mean_to_other} - 50*\texttt{scale}}{\texttt{\texttt{mean_to_self}} - 50*\texttt{scale}}\right)$$	always
svo_type	svo type,	always
	1 = Altruist $\Leftrightarrow$ svo_angle $>57.15$,
	2 = Prosocial $\Leftrightarrow$ $57.15 \geq$ svo_angle $>22.45$,
	3 = Individualist $\Leftrightarrow$ $22.45 \geq$ svo_angle $>-12.04$,
	4 = Competitive $\Leftrightarrow$ svo_angle $\geq-12.04$
avg_dist_to_equality	average standardized distance of the choice to the choice which would maximize equality	select_items = FULL
avg_dist_to_altruist	average standardized distance of the choice to the choice which would maximize altruism	select_items = FULL
avg_dist_to_joint	average standardized distance of the choice to the choice which would maximize joint earnings	select_items = FULL
avg_dist_to_indiv	average standardized distance of the choice to the choice which would maximize individual gains	select_items = FULL
not_altru_indiv	Dummy taking the value 1 if avg_dist_equality, avg_dist_joint $>$ avg_dist_altruist, avg_dist_indiv	select_items = FULL
inequality_aversion_score	Degree of inequality aversion calculated as $$\frac{\texttt{avg_dist_to_equality}}{(\texttt{avg_dist_to_equality+avg_dist_to_joint})}$$ if not_altru_indiv == 1, else -99)	select_items = FULL
paid_slider	slider selected for payment as sender	always
slider_as_receiver	slider selected for payment as receiver	matching=RING
kept_of_sender	amount kept by the sender	always
received_from_sender	amount received from the sender	always
kept_as_sender	amount kept as sender	matching=RING
sent_as_sender	amount sent as sender	matching=RING

Parameters in the oTree treatment

Parameter	Values	Description
language	EN	Language is English
	DE	Language is German
	IT	Language is Italian
select_items	PRIMARY	only the primary items (items 1-6) are elicited.
	FULL	primary and secondary items (items 1-15) are elicited
items_in_random_order	ORDERED	items are presented according to the order in Murphy et al. (2011)
	RANDOM	items are presented in random order
matching	RING	Subject A,B,C,D are ordered on a ring-structure as in Murphy et al. (2011). In this case, A gives to B, B gives to C, C gives to D and D gives to A, which makes everyone a sender AND a receiver.
	RANDOM_DICTATOR	Subjects A,B,C,D are matched in groups of 2 (say (A; B) and (C;D). One member of each group (say A and C) is selected to be the sender, the other one as receiver. In this case, B receives from A and D receives from C.
precision	TWO_DIGITS_AFTER_POINT	values on sliders are rounded to two digits after decimal point
	INTEGERS	values on sliders are rounded to
scale	$(0, +\infty]$	Parameter to scale up $(>1$) or down $(<1)$ all the numbers on a slider. Default is 1, resulting in a circle of diameter 100
random_payoff	RAND	the payoff will be calculated as a random choice
	SUM	the payoff is the sum of all choices of the player
item_initialization	RAND	initialize the items in random manner
	AVG	initialize the items using average between min and max

Relevant variables in the subjects table of the oTree

Relevant Variables	Description	Relevant for Parameters
payoff	Profit of the Subject	always
input_self_X	The amount of money that the user has chosen for himself for item X	always
input_other_X	The amount of money that the user has chosen for others for item X	always
random_orderX	order of item X in RANDOM order	items_in_random_order = RANDOM
mean_to_self	mean allocation to self in primary items	always
mean_to_other	mean allocation to other in primary items	always
svo_angle	svo angle calculated as $$\arctan\left(\frac{\texttt{mean_to_other} - 50*\texttt{scale}}{\texttt{\texttt{mean_to_self}} - 50*\texttt{scale}}\right)$$	always
svo_type	svo type,	always
	1 = Altruist $\Leftrightarrow$ svo_angle $>57.15$,
	2 = Prosocial $\Leftrightarrow$ $57.15 \geq$ svo_angle $>22.45$,
	3 = Individualist $\Leftrightarrow$ $22.45 \geq$ svo_angle $>-12.04$,
	4 = Competitive $\Leftrightarrow$ svo_angle $\geq-12.04$
avg_dist_to_equality	average standardized distance of the choice to the choice which would maximize equality	select_items = FULL
avg_dist_to_altruist	average standardized distance of the choice to the choice which would maximize altruism	select_items = FULL
avg_dist_to_joint	average standardized distance of the choice to the choice which would maximize joint earnings	select_items = FULL
avg_dist_to_indiv	average standardized distance of the choice to the choice which would maximize individual gains	select_items = FULL
altru_indiv	Dummy taking the value true if avg_dist_equality, avg_dist_joint $\leq$ avg_dist_altruist, avg_dist_indiv	select_items = FULL
inequality_aversion_score	Degree of inequality aversion calculated as $$\frac{\texttt{avg_dist_to_equality}}{(\texttt{avg_dist_to_equality+avg_dist_to_joint})}$$ if altru_indiv == true, else -99)	select_items = FULL
paid_slider	slider selected for payment as sender	always
slider_as_receiver	slider selected for payment as receiver	always
kept_of_sender	amount kept by the sender always
received_from_sender	amount received from the sender	always

Footnotes

1. Personal communication with R. Murphy. ↩

2. This version app was implemented in oTree version 3.0.6. Please update the app according to the documentation if a new version of oTree is released. ↩