Affine Wealth Model (AWM) Overview

• Affine Wealth Model (AWM) is a stochastic agent-based asset-exchange model that extends earlier frameworks to include agents with negative wealth via additive shifts.
• It integrates redistribution and wealth-attained advantage within a nonlinear Fokker-Planck formulation, achieving high empirical accuracy with Lorenz curve fits.
• The model provides insights into partial wealth condensation and sets dynamical constraints on runaway concentration by linking labor income growth with invested wealth.

to=arxiv_search.search ગઈనం 天天中彩票中 2 天天中彩票上 json {"query":"Affine Wealth Model wealth distribution (Li et al., 2016, Polk et al., 2020, Astuti, 2022)","max_results":5,"sort_by":"relevance"} RTLRക്കം to=arxiv_search.search code 大发快三走势图 皇轩្មី to=arxiv_search.search 全民彩票 的天天彩票 微信的天天中彩票 彩神争霸 北京赛车女 五分彩 _植物百科通 天天购彩票 六和彩 632 日日啪 to=arxiv_search.search code _日本毛片免费视频观看 平台直属 to=arxiv_search.search 买天天中彩票 天天中彩票微信 to=arxiv_search.search code 天天中彩票买 by tool? to=search_arxiv 天天中彩票人工 彩神争霸是 天天爱彩票app 天天彩票app 大发时时彩开奖 天天中彩票软件 无码av 大发扑克 玩彩神争霸 手机天天中彩票 天天中彩票这个 彩彩票娱乐 json {"query":"(Li et al., 2016) Affine Wealth Model (Polk et al., 2020, Astuti, 2022)","max_results":10} The Affine Wealth Model (AWM) is a stochastic, agent-based, binary-transaction Asset-Exchange Model (AEM) for wealth distribution that allows for agents with negative wealth while retaining redistribution and wealth-attained advantage (WAA). It extends earlier asset-exchange models by imposing not only scaling invariance but also shift invariance, so that the agent density function admits both scalings and additive shifts. In the published formulation, the AWM is paired with a nonlinear Fokker-Planck description, a numerical inverse-problem methodology for fitting model parameters to empirical Lorenz curves, and later generalizations in which redistribution becomes an arbitrary function of wealth. Within this line of work, the AWM is used to study negative wealth, partial wealth condensation, near-critical tail behavior, and dynamical constraints on runaway concentration in affine-type stochastic wealth processes (Li et al., 2016, Polk et al., 2020, Astuti, 2022).

1. Conceptual basis and model class

The AWM was introduced as the most advanced member of a class of stochastic, agent-based, binary-transaction Asset-Exchange Models. Its central motivation is the inability of prior models, especially the Extended Yard-Sale Model (EYSM), to handle agents with negative wealth. In the motivating discussion, this was tied to the empirical fact that over 10% of US households have negative net worth. To address this, the AWM requires invariance under additive shifts in wealth in addition to the scaling invariance already used in earlier models (Li et al., 2016).

This affine structure is the source of the model’s name. Scaling invariance refers to multiplicative changes in population size and total wealth, while shifting refers to additive transformations of wealth. In the AWM, this is not a cosmetic reformulation: it changes the support of the wealth distribution and permits the Lorenz curve to dip below zero in the left tail, which is necessary for representing net-negative wealth populations. AWM therefore occupies a distinct position within AEMs by combining three ingredients that were already present separately or partially in related work: redistribution, WAA, and a formal treatment of negative wealth (Li et al., 2016).

The model is also situated within a broader class of affine stochastic wealth processes, in which wealth growth contains multiplicative and additive components. Later work used this broader affine perspective to prove general necessary and sufficient conditions for avoiding infinite wealth concentration without assuming the existence of an equilibrium wealth distribution. That result links the AWM to a larger family of models studied in economics and econophysics (Astuti, 2022).

2. Transaction mechanism, affine shift, and parameters

In the AWM, wealth takes values on the interval $w \in [-\Delta,\infty)$. The mechanism for allowing negative wealth is a shift construction: each transaction is performed as if both agents had been given an additive boost $\Delta$, the underlying EYSM interaction is applied to the shifted wealths, and then the boost is subtracted. If $w$ and $x$ are the pre-transaction wealths, the shifted variables are

$\bar w = w + \Delta, \qquad \bar x = x + \Delta.$

The transaction rule summarized for the model is $\eta$9 where $\gamma$ is a transactional rate constant, $\chi$ is the redistribution coefficient per transaction, and $\eta$ is the random variable whose expectation encodes WAA. The parameter $\zeta$ controls the magnitude of WAA, and the shift is parametrized by $\kappa$ through

$\Delta$0

with the convenient notation

$\Delta$1

These parameters are interpreted in the source material as follows: $\Delta$2 tends to equalize wealth by taxing and redistributing, $\Delta$3 gives systematic transactional advantage to wealthier agents, and $\Delta$4 sets the extent of the negative-wealth region relative to the average shifted wealth (Li et al., 2016).

A recurring substantive feature of the AWM is that WAA and redistribution compete. Higher $\Delta$5 means greater stabilization and less inequality, whereas higher $\Delta$6 amplifies inequality and, for $\Delta$7, leads to a phase in which a finite fraction of wealth is condensed in an oligarchy. The model therefore encodes inequality not only through random exchange, but through a systematic asymmetry favoring wealthier agents and an explicit countervailing redistribution mechanism (Li et al., 2016).

3. Fokker-Planck formulation and affine symmetries

The agent density function $\Delta$8 in the AWM evolves according to a generalized nonlinear Fokker-Planck equation. In the formulation summarized in the source, the equation is

$\Delta$9

with $w$0. Here $w$1 is the mean agent wealth, and the Pareto-Lorenz potentials are

$w$2

$w$3

$w$4

In steady state, the time derivative vanishes, leaving a nonlinear ordinary integrodifferential equation for $w$5 (Li et al., 2016).

Several structural consequences are explicitly identified. The redistribution term’s strength is reduced by an amount proportional to $w$6, the WAA effect is scaled by $w$7, and the diffusion term contains shift-dependent contributions that vanish at $w$8, where the model reduces to the EYSM. The source also emphasizes a transformation and duality: the AWM solution can be constructed from the EYSM’s by shifting and rescaling, and in supercritical cases the Lorenz curve can be obtained by swapping $w$9 and $x$0 in the subcritical EYSM solution and rescaling the outcome (Li et al., 2016).

The Lorenz representation is central both analytically and empirically. The model relation stated for the Lorenz curve is

$x$1

This affine transformation explains why the AWM can represent negative-wealth populations and why the Lorenz curve can dip below zero. It also provides a direct route for fitting the model to data and for interpreting supercritical states in terms of an oligarchical wealth fraction (Li et al., 2016).

4. Redistribution, criticality, and nonuniversal wealth tails

Later work modified the AWM by replacing flat redistribution with an arbitrary nonconstant function of wealth, $x$2. In that generalization, $x$3 gives the rate at which an agent with wealth $x$4 contributes to redistribution, subject only to mild regularity and integrability conditions at infinity. This change was introduced specifically to examine the effects of nonconstant redistribution on the very wealthy and to study how policy structure affects the tail of the wealth distribution (Polk et al., 2020).

For the generalized steady state, the wealth density $x$5 is governed by

$x$6

Assuming at high wealth that

$x$7

the asymptotic relation reported in the source is

$x$8

The first term, the integral of $x$9, is identified as governing the dominant behavior of the tail for general redistribution (Polk et al., 2020).

The resulting classification sharply modifies earlier expectations of universality. In the flat-redistribution case, previous studies had shown a phase transition to a partially wealth-condensed state, or partial oligarchy, at a critical value of an order parameter; they had also indicated an exponential tail precisely at criticality and a Gaussian tail away from criticality. In the generalized model, the critical exponential tail is recovered only as a special case. When $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$0 only asymptotically, the tail can instead be any slower-than-Gaussian subquadratic decay determined by the detailed asymptotics of $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$1 (Polk et al., 2020).

The source gives concrete examples. If $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$2 approaches $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$3 so that $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$4, the tail is exponential. If $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$5, the tail is lognormal. If $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$6, the tail is Pareto. Other subquadratic forms are also possible. By contrast, if the large-$\bar w = w + \Delta, \qquad \bar x = x + \Delta.$7 limit satisfies $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$8, partial oligarchy forms and the tail is Gaussian, while if $\bar w = w + \Delta, \qquad \bar x = x + \Delta.$9, there is no oligarchy and the tail remains Gaussian (Polk et al., 2020).

Regime	Condition	Consequence
Critical flat case	$\gamma$0 as a constant limit	Exponential tail
Supercritical	$\gamma$1	Partial oligarchy; Gaussian tail
Subcritical	$\gamma$2	No oligarchy; Gaussian tail

A central implication is that the functional form of the tail near criticality is not universal in nature but rather entirely determined by the specifics of public policy decisions. The paper’s inverse problem makes that dependence explicit:

$\gamma$3

This formula states, asymptotically, which redistribution profile generates a prescribed tail form. The associated empirical observation is that all 14 European countries fitted with constant $\gamma$4 were found to be near-critical, and the fitting errors were largest in the tail, suggesting that nonconstant redistribution may play a major role in shaping observed high-wealth behavior (Polk et al., 2020).

5. Empirical validation and inverse estimation

The original empirical validation of the AWM used United States Survey of Consumer Finances data from 1989–2016, merged with Forbes 400 data to restore the missing upper tail. Lorenz ordinates were constructed via weighted cumulative sums on sorted wealth data, and the inverse problem was defined as finding the parameter set $\gamma$5 minimizing the $\gamma$6 distance between the empirical and model Lorenz curves:

$\gamma$7

The optimization was described as a global numerical search, with a line search for $\gamma$8 seeded by a closed-form $\gamma$9 minimizer (Li et al., 2016).

Four models were tested: the Single Agent Model (SAM), EYSM without WAA, EYSM with WAA, and the full AWM. The AWM was reported to yield dramatically lower errors. In the abstract, the steady-state solutions of the Fokker-Planck equation were said to agree with empirical data of an average error less than 0.16\% over a time period of 27 years. The detailed summary further states that average local errors were typically less than 0.16\%, and that the fits were visually indistinguishable except at the highest magnification (Li et al., 2016).

The empirical fits also generated substantive diagnostics. For all years in the sample, the United States wealth distribution was found to be supercritical, with $\chi$0, implying that a finite fraction of wealth was held by a vanishingly small fraction of agents. The reported oligarchical share was 20–32% of total wealth. The time series of fitted parameters changed relatively slowly, which was used to justify an adiabatic approximation in which each year’s observed distribution was treated as a steady-state response to slowly varying parameters. In that interpretation, the fitted values of $\chi$1, $\chi$2, and $\chi$3 constitute a diagnostic tool for analyzing the evolving balance between redistribution, WAA, and the depth of negative-wealth poverty (Li et al., 2016).

6. Dynamical constraints on wealth concentration in affine-type processes

A subsequent paper studied a large class of stochastic wealth growth models that includes affine processes with multiplicative and additive components. The generic update rule was written as

$\chi$4

where $\chi$5 is the multiplicative stochastic process and $\chi$6 is the additive stochastic process. Within this framework, the paper defined

$\chi$7

and the ratio

$\chi$8

This ratio was interpreted as the average labor income relative to invested wealth at time $\chi$9 (Astuti, 2022).

The Gini evolution was decomposed as

$\eta$0

and, under the assumptions stated in the source, as

$\eta$1

The main result was a necessary and sufficient condition for avoiding infinite wealth concentration: wealth concentration remains bounded, $\eta$2, if and only if there exists a positive lower bound on $\eta$3 over time. Formally, the concluding statement was

$\eta$4

If $\eta$5, then $\eta$6 and runaway concentration is inevitable; if $\eta$7 stays strictly positive, then $\eta$8 remains bounded below 1 (Astuti, 2022).

This result is notable for two reasons already emphasized in the source. First, it generalizes earlier model-dependent conditions, including the Kesten condition, without assuming the existence of a stationary wealth distribution. Second, it identifies the additive component of growth, usually representing labor incomes, as the quantity that must keep pace with average invested wealth in order to limit concentration. In the toy model labeled “No Consumption (Pure Affine Model),” the paper states that equilibrium in inequality holds if and only if labor income growth at least matches capital growth; otherwise concentration runs away. In this sense, the AWM sits inside a broader theoretical statement: stochastic multiplicative growth can be compatible with bounded inequality only when the additive component does not become negligible relative to average wealth (Astuti, 2022).