Skip to main content

Asymmetric social protection systems with migration


We study the consequences of the coexistence of different social protection systems on contribution rate levels and migration in a two-country model. Before any migration takes place, the levels of contribution rates are chosen by a representative elected in each country. The migration of each agent depends on her anticipation of other agents’ migrations. We show that the richest agents are attracted to the Bismarckian country. The poorest agents tend to migrate toward one country or the other depending on the Beveridgean country contribution rate. The Beveridgean country can set a higher contribution rate to limit the departures of rich agents.

This is a preview of subscription content, access via your institution.


  1. The USA, Japan, and Australia, for example, have adopted a liberal regime.

  2. See Besley and Coate (1997) for a complete analysis of the citizen–candidate notion.

  3. Empirical studies show that a value of \(\lambda >1\) is realistic (see Friend and Blume 1975; Grossman and Shiller 1982; Brown and Gibbons 1985).

  4. If the relative risk aversion \(\lambda\) is lower than \(1,\) then the opposite effect occurs: the richer the agents are, the lower their preferred contribution rate.

  5. We assume that there is a unique median income \(w_{m}\).

  6. Note that the opposite is true for \(\lambda<1.\)

  7. \(f\) is defined on whole \(R\), but \(f(x)\) is the expatriation cost probability density only for \(x>0\).

  8. \(F\) is an increasing function from \(R\) to \([0;2]\), with \(F(0)=1\).

  9. Boerner and Uebelmesser consider a median voter model. Migration leads to a reduction of the median income, but since the average income decreases as well, the effect on the tax rate is ambiguous.

  10. \( w_{s} = t^{A} \overline{w} ^{p} {\left( {1 - p} \right)}^{{\frac{\lambda } {{1 - \lambda }}}} {\left[ {1 - p^{\lambda } {\left( {1 - t^{A} } \right)}^{{1 - \lambda }} } \right]}^{{\frac{1} {{\lambda - 1}}}} \) if \(t^{A}\leq 1-p^{\lambda /(\lambda -1)}\) and \(w_{s}=0\) if \( t^{A}\geq 1-p^{\lambda /(\lambda -1)}\).


  • Besley T, Coate S (1997) An economic model of representative democracy. Q J Econ 112:85–114

    Article  Google Scholar 

  • Boerner K, Uebelmesser S (2005) Migration and the welfare state: the economic power of the non-voter? CESifo Working Paper No. 1517

  • Brown D, Gibbons M (1985) A simple econometric approach for utility-based asset pricing models. J Finance 40:359–381

    Article  Google Scholar 

  • Cremer H, Pestieau P (1998) Social insurance, majority voting and labor mobility. J Public Econ 68:397–420

    Article  Google Scholar 

  • Cremer H, Pestieau P (2003) Social insurance competition between Bismarck and Beveridge. J Urban Econ 54:181–196

    Article  Google Scholar 

  • Epple D, Romer T (1991) Mobility and redistribution. J Polit Econ 99:828–858

    Article  Google Scholar 

  • Esping-Andersen G (1990) The three worlds of welfare capitalism. Princeton Univ. Press, Princeton, NJ

    Google Scholar 

  • Friend I, Blume M (1975) The demand for risky assets. Am Econ Rev 65:900–922

    Google Scholar 

  • Grossman S, Shiller J (1982) Consumption correlatedness and risk measurement in economics with non-traded asset and heterogeneous information. J Financ Econ 10:195–210

    Article  Google Scholar 

  • Hansen N, Kessler A (2001) The political geography of tax h(e)avens and tax hells. Am Econ Rev 91:1103–1115

    Article  Google Scholar 

  • Lejour A, Verbon H (1994) Labour mobility and decision making on social insurance in an integrated market. Public Choice 79:161–185

    Article  Google Scholar 

  • Meier V (2000) Time preference, international migration, and social security. J Popul Econ 13:127–146

    Article  Google Scholar 

Download references


We would like to give special thanks to Hubert Kempf, Patrick Toche, Etienne Lehmann, and EUREQua’s “political economy” group for their comments. We also thank the Journal’s anonymous referees. We take full responsibility for any remaining errors.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Stéphane Rossignol.

Additional information

Responsible editor: Alessandro Cigno



The first-order condition in country \(B\) gives:

$$-pw_{j}U^{\prime }\left( w_{j}\left( 1-t^{B}\right) \right) +pw_{j}U^{\prime }\left( t^{B}w_{j}\frac{p}{1-p}\right) =0$$

the second-order condition is satisfied because we have, by concavity of \(U\!\!\):

$$pw_{j}^{2}U^{''}\left( w_{j}\left( 1-t^{B}\right) \right) +\frac{ p^{2}}{1-p}w_{j}^{2}U^{''}\left( t^{B}w_{j}\frac{p}{1-p}\right)<0 $$

and in country \(A\!\!\):

$$ -pw_{j}U^{\prime }\left( w_{j}\left( 1-t^{A}\right) \right) +p\overline{w} U^{\prime }\left( t^{A}\overline{w}\frac{p}{1-p}\right) =0$$

with a CRRA utility:

$$-p\left( w_{j}^{1-\lambda }\left( 1-t^{A}\right) ^{-\lambda }-\overline{w} ^{1-\lambda }\left( \frac{t^{A}p}{1-p}\right) ^{-\lambda }\right) =0 $$

the second-order condition is satisfied since we have:

$$ pw_{j}^{2}U^{''}\left( w_{j}\left( 1-t^{A}\right) \right) +\frac{ p^{2}}{1-p}\overline{w}^{2}U^{''}\left( t^{A}\overline{w}\frac{p }{1-p}\right)<0$$

Solving Eqs.12 and 13, we have for \(\lambda >1\!\!\):

$$ t^{B*}=1-p\rm{ }$$


$$ t_{Cj}^{A*}=\frac{1}{1+\left( \frac{\overline{w}}{w_{j}}\right) ^{1-\frac{1}{ \lambda }}\frac{p}{1-p}}\rm{}$$

Existence of a unique migration equilibrium for uniform costs probability density

We want to prove that there exists a unique equilibrium satisfying (i), (ii), and (iii) in Eq. 5, if the costs density function is uniform on \([0;a]\) (for \(a\) large enough).

We can rewrite Eq. 5 (iii) for any \(j\) with \(t^{B}=1-p\) as:

$$ x_{j}=\frac{D}{1-\lambda }(\overline{w}^{a})^{1-\lambda }+w_{j}^{1-\lambda }K\left( t^{A}\right)$$

where \(K\left( t^{A},t^{B}\right) =pU\left( 1-t^{A}\right) -U\left( p\right)\) and \(D=(1-p)\left( t^{A}\frac{p}{1-p}\right) ^{1-\lambda }\!\!\). We set \(\phi (\overline{w}^{a})=\frac{\sum_{j}N_{j}F(x_{j})w_{j}}{\sum_{j}N_{j}F(x_{j})}\)

We assume that the cost density function is uniform in \([0;a]\!\!\). This means that \(f(x)=\frac{1}{a}\) for \(x\in [-a;+a]\!\!\), and otherwise \(f(x)=0\!\!\). When \(a>0\) is sufficiently large, we obtain \(\left| x_{j}\right|<a\) for any \(j\!\!\), which means that no income group in any country can be totally emptied by the migration of its agents. We then have \(\phi (\overline{w}^{a})=\frac{ \sum_{j}N_{j}(1+\frac{x_{j}}{a})w_{j}}{\sum_{j}N_{j}(1+\frac{x_{j}}{a})}\!\!\).

Our aim is thus to prove that a unique migration equilibrium, i.e. a unique couple \((\overline{w}^{a},\overline{w}^{p})\) exists, such that \(\overline{w} ^{a}=\overline{w}^{p}=\phi (\overline{w}^{a})\!\!\). Clearly, we have \(\frac{dx_{j}}{d\overline{w}^{a}}=D(\overline{w} ^{a})^{-\lambda }\!\!\!\), so we can write:

$$ \begin{array}{*{20}c} {\phi \prime {\left( {\overline{w} ^{a} } \right)} = }{\frac{{\frac{D} {a}{\left( {\overline{w} ^{a} } \right)}}} {{{\left[ {{\sum\nolimits_j {N_{j} w_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right]}^{2} }}{\left[ {{\left( {{\sum\nolimits_j {N_{j} w_{j} } }} \right)}{\left( {{\sum\nolimits_j {N_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right)} - {\left( {{\sum\nolimits_j {N_{j} w_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right)}{\left( {{\sum\nolimits_j {N_{j} } }} \right)}} \right]}} \\ = {\frac{{\frac{D} {a}{\left( {\overline{w} ^{a} } \right)}^{{ - \lambda }} }} {{{\left[ {{\sum\nolimits_j {N_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right]}}}{\left[ {{\left( {{\sum\nolimits_j {N_{j} w_{j} } }} \right)} - \phi {\left( {\overline{w} ^{a} } \right)}{\left( {{\sum\nolimits_j {N_{j} } }} \right)}} \right]}} \\ \end{array} $$

and \(\phi ^{\prime }(\overline{w}^{a})\geq 0\Leftrightarrow \phi (\overline{w }^{a})\leq \frac{\sum_{j}N_{j}w_{j}}{\sum_{j}N_{j}}=\bar{w},\) where \(\bar{w}\) is the average ex ante income.

Let us compute the second derivative:

$$ \phi {\left( {\overline{{\text{w}}} ^{a} } \right)} = \frac{{\frac{D} {a}}} {{{\left[ {{\sum\nolimits_j {N_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right]}^{2} }}\left[ {{\left( {{\sum\nolimits_j {N_{j} {\left( {1 + \frac{{x_{j} }} {a}} \right)}} }} \right)}} \right.{\left[ {{\left( { - \lambda {\left( {\overline{w} ^{a} } \right)}^{{ - 1 - \lambda }} } \right)}{\left( {{\sum\nolimits_j {N_{j} w_{j} - \phi {\left( {\overline{w} ^{a} } \right)}{\sum\nolimits_j {N_{j} } }} }} \right)} + {\left( {\overline{w} ^{a} } \right)}^{{ - \lambda }} {\left( { - \phi \prime {\left( {\overline{w} ^{a} } \right)}{\sum\nolimits_j {N_{j} } }} \right)}} \right]} - \frac{D} {a}{\left( {\overline{w} ^{a} } \right)}^{{ - 2\lambda }} {\left( {{\sum\nolimits_j {N_{j} w_{j} - \phi {\left( {\overline{w} ^{a} } \right)}} }{\sum\nolimits_j {N_{j} } }} \right)}{\sum\nolimits_j {N_{j} } } $$

We can see that this expression is negative for \(\phi (\overline{w}^{a})\leq \bar{w}\). We have thus shown that \(\phi \) is a decreasing function for \(\phi ( \overline{w}^{a})\geq \bar{w}\) and an increasing and concave function for \(\phi (\overline{w}^{a})\leq \bar{w}\!\!\). In the plane of coordinates \(( \overline{w}^{a},\overline{w}^{p})\!\!\!\), we denote the representative curve of \(\phi\) in \([w_{1},w_{M}]\) by \(C\phi ,\) and the line \(\overline{w}^{a}= \overline{w}^{p}\) by \(\Delta\!\).

Two cases must be considered:

  • If \(K(t^{A})<0\), \(x_{j}=w_{j}^{1-\lambda }K(t^{A})+\frac{D}{1-\lambda }( \overline{w}^{a})^{1-\lambda }\) is an increasing function of \(w_{j}\!\!\). Setting \(N_{j}^{\prime }=N_{j}(1+\frac{x_{j}}{a})\!\!\), the ratio \(\frac{ N_{j}^{\prime }}{N_{j}}\) increases with \(w_{j}\!\), which implies, according to Lemma 2 (stated further), that \(\phi (\overline{w}^{a})=\frac{ \sum_{j}N_{j}^{\prime }w_{j}}{\sum_{j}N_{j}^{\prime }}\geq \bar{w}\!\!\). The curve \(C_{\phi }\) is thus decreasing and intersects \(\Delta\) at a unique point \(w^{\ast }\) (see Fig. 1).

  • If \(K(t^{A})>0\!\), then \(x_{j}\) is a decreasing function of \(w_{j}\!\). The ratio \(\frac{N_{j}^{\prime }}{N_{j}}\) decreases with \(w_{j}\!\), which implies, according to Lemma 2, that \(\phi (\overline{w}^{a})\leq \bar{w}\!\). The curve \(C_{\phi }\) is then increasing and concave: it thus intersects \(\Delta\) at a unique point \(w^{\ast }\) (see Fig. 2).

Fig. 1
figure 1

Unique equilibrium with \(K\left(t^{A}\right)<0\)

Fig. 2
figure 2

Unique equilibrium with \(K\left(t^{A}\right)>0\)

In both cases, we have proved the existence of a unique migration equilibrium.

Proof of Propositions 2 and 3

Propositions 2 and 3 can be rewritten using the \(x_{j}\). They assert that:

  • If \(t^{A}<t_{s}\!\), then:

    \( x_{j} < 0 \Leftrightarrow w_{j} > w_{s} \)

    \(x_{j}\) is an increasing function of \(w_{j}\!\!\).


  • If \(t^{A}>t_{s}\!\), then:

    \(x_{j}<0\) for every \(j\)

    \(x_{j}\) is a decreasing function of \(w_{j}\)


We first prove that \(x_{j}\) is a monotonous function of \(w_{j}\!\!\). Secondly, we study its sign and variations (according to \(t^{A}<t_{s}\) and \(t^{A}>t_{s}\) ) and finally compare \(\overline{w}^{p}\) and \(\overline{w}\!\).

  1. 1.

    Monotonicity of \(x_{j}\) with respect to \(w_{j}\)

    $$ x_{j} = {\left( {1 - p} \right)}U{\left( {t^{A} \overline{w} ^{p} \frac{p} {{1 - p}}} \right)} + pU{\left( {w_{j} {\left( {1 - t^{A} } \right)}} \right)} - {\left( {1 - p} \right)}U{\left( {t^{B} w_{j} \frac{p} {{1 - p}}} \right)} - pU{\left( {w_{j} {\left( {1 - t^{B} } \right)}} \right)}$$

    Taking \(t^{B}=1-p\!\!\), we can write \(x_{j}\) in the following way:

    $$ x_{j}=\left( 1-p\right) U\left( t^{A}\overline{w}\,^{p}\frac{p}{1-p}\right) +w_{j}^{1-\lambda }K\left( t^{A}\right)$$


    $$ K\left( t^{A}\right) =pU\left( 1-t^{A}\right) -U\left( p\right)$$

    \(x_{j}\) is thus a monotonous function of \(w_{j}\!\!\).

  2. 2.

    Sign and variations of \(x_{j}\!\!\) We can rewrite \(x_{j}\) as:

    $$ x_{j}=w_{j}^{1-\lambda }H\left( z\right) {\rm\text for }\; z={\frac{\overline{w}^{p} }{w_{j}}}$$


    $$ H\left( z\right) =\left( 1-p\right) U\left( t^{A}z\frac{p}{1-p}\right) +K\left( t^{A}\right)$$

    \(H\) is increasing on \(]0;+\infty [\), with \({\mathop {\lim H{\left( z \right)}}\limits_{z \to 0} } = - \infty \) and \( {\mathop {\lim }\limits_{z \to + \infty } }H{\left( z \right)} = K{\left( {t^{A} } \right)} \)

    If this last limit is negative, then the equation \(H\left( z\right) =0\) has no solution. Observe that

    $$ K{\left( {t^{A} } \right)} < 0 \Leftrightarrow t^{A} > t_{s} \;{\text{where}}\;t_{s} = 1 - \frac{\lambda } {{p^{{\lambda - 1}} }}$$

    We have thus the two following cases:

    • For every \(t^{A}>t_{s}\!\), \(H\left( z\right)<0\) for every \(z\) i.e. \( x_{j}<0\) for any \(j\!\!\).

      This is case (b) of Proposition 3: all agents are incited to migrate to country \(B.\) They migrate if their expatriation costs are lower than \(-x_{j}\!\!\).

      Since \(K\left( t^{A}\right)<0\!\!\), \(x_{j}\) is an increasing function of \( w_{j}, \) which means that the proportion of migrants per group is a decreasing function of income.

    • If \(t^{A}<t_{s},\) we have \(K\left( t^{A}\right) >0\!\!\). Thus, there exists \(z^{*},\) such that \(H\left( z^{*}\right) =0\!\!\). Solving this equation, we obtain:

      $$ z^{*}=\frac{1}{t^{A}}\left( 1-p\right) ^{\frac{\lambda }{\lambda -1}}\left[ 1-p^{\lambda }\left( 1-t^{A}\right) ^{1-\lambda }\right] ^{\frac{1}{ 1-\lambda }}$$

      We can set

      $$ w_{s} = \frac{{\overline{w} ^{p} }} {{z*}} = t^{A} \overline{w} ^{p} {\left( {1 - p} \right)}^{{\frac{\lambda } {{1 - \lambda }}}} {\left[ {1 - p^{\lambda } {\left( {1 - t^{A} } \right)}^{{1 - \lambda }} } \right]}^{{\frac{\lambda } {{\lambda - 1}}}}$$

      then, since \(H\left( z\right) \) is increasing, for \(w_{j}>w_{s}\) we have \( z = \frac{{\overline{w} P}} {{wj}} < z*, \) thus \(x_{j}<0\!\!\). Symmetrically, for \( w_{j}<w_{s,}\) we have \(x_{j}>0\!\!\).

    This means that agents who are richer than \(w_{s}\) are prompted to migrate to \(B\!\!\), and those who are poorer than \(w_{s}\) are prompted to migrate to \(A\!\). Here, \( x_{j} \) is necessarily decreasing. Thus, the proportion of migrants to \(B\) is an increasing function of income and that of migrants to \(A\) a decreasing function of income.

  3. 3.

    Variation of \(\bar{w}\!\!\). To compare \(\bar{w}\,^{p}\) and \(\bar{w}\!\!\), we first need the following lemma.

Lemma 2

Let \(N_{j}^{p}\)be the post-migration population of income group\(w_{j}\)and thus

$$ \overline{w} ^{p} = \frac{{{\sum {N^{p}_{j} w_{j} } }}} {{{\sum {N^{p}_{j} } }}}$$


  1. (a)

    If \(\left( \frac{N_{j}^{p}}{N_{j}}\right)\)is a decreasing function of \(j\!\!\), then \(\overline{w}\,^{p}\leq \bar{w}\!\!\).

  2. (b)

    If \(\left( \frac{N_{j}^{p}}{N_{j}}\right)\)is an increasing function of \(j\!\!\), then \(\overline{w}\,^{p}\geq \bar{w}\!\!\).


  1. (a)

    Let \(\overline{w}^{\prime }=\frac{\sum N_{j}^{\prime }w_{j}}{\sum N_{j}^{\prime }}\!\!\). It follows that if \(N_{j}^{\prime }\geq N_{j}\) for \( w_{j}\leq \overline{w}^{\prime }\) and \(N_{j}^{\prime }\leq N_{j}\) for \( w_{j}\geq \overline{w}^{\prime }\!\!\), then \(\overline{w}^{\prime }\leq \overline{w}\!\), since the \(w_{j}\) that are lower than \(\overline{w}^{\prime }\) get a higher weight in \(\overline{w}^{\prime }\) than in \(\overline{w}\!\!\). Now let \(d>0\!\!\), such that:

    $$ \frac{d.N_{1}^{p}}{N_{1}}\geq ...\geq \frac{d.N_{l}^{p}}{N_{l}}\geq 1\geq \frac{d.N_{l+1}^{p}}{N_{l+1}}\geq ...\geq \frac{d.N_{M}^{p}}{N_{M}}$$

    where \(l\) satisfies \(w_{l}\leq \overline{w}^{p}<w_{l+1}\!\!\). Then, setting \(N_{j}^{\prime }=d.N_{j}^{p}\!\!\), we have \(\overline{w}^{\prime }=\overline{w} ^{p}\leq \overline{w}\!\!\).

  2. (b)

    Similarly, inverting \(\bar{w}\) and \(\overline{w}^{p}\!\!\).

We can now achieve the proof of Proposition 3. We have:

$$ \bar{w}=\frac{\sum N_{j}w_{j}}{\sum N_{j}}{\rm\text \;and }{{\;\overline{w}\,^{p}=\frac{\sum N_{j}F(x_{j})w_{j}}{\sum N_{j}F(x_{j} ) }}}$$

Adopting notations of Lemma 2, we set \(N_{j}^{\prime }=N_{j}F(x_{j})\!\!\), where we know that \(F\) is an increasing function.

  • For \(t^{A}<t_{s}\!\), \(x_{j}\) is decreasing; thus, \(\frac{N_{j}^{\prime }}{ N_{j}}=F(x_{j})\) is decreasing with \(j\!\!\). We then obtain \(\overline{w}^{p}\leq \bar{w}\) by applying Lemma 2.

  • For \(t^{A}>t_{s}\!\), \(x_{j}\) is increasing; thus \(\frac{N_{j}^{\prime }}{ N_{j}}=F(x_{j})\) is increasing with \(j\!\). We then obtain \(\overline{w}^{p}\geq \bar{w}\!\!\).

Proof of Lemma 1

  • Let us assume \(w_{k}\geq \overline{w}^{p}\)

    Since \(\overline{w}^{p}>w_{s}\) agents with income \(w_{k}\) migrate from \(A\) to \(B\!\!\).

  • Let us assume \(w_{k}<\overline{w}^{p}\!\!\!\). We can then set:

    $$ \begin{array}{*{20}c} {\Phi _{j} {\left( {t^{A} ,\overline{w} ^{p} } \right)} = x_{j} } \\ { = {\left( {1 - p} \right)}U{\left( {t^{A} \overline{w} ^{p} \frac{p} {{1 - p}}} \right)} + pU{\left( {w_{j} {\left( {1 - t^{A} } \right)}} \right)} - {\left( {1 - p} \right)}U{\left( {t^{B} w_{j} \frac{p} {{1 - p}}} \right)}} \\ \end{array} $$

    for all given \(\overline{w}^{p}\) and \(t^{A}\!\!\).

If we take \(t^{A}=t^{B*}=1-p\!\), we can then write:

$$ \Phi _{j} {\left( {t^{B} ,\overline{w} ^{p} } \right)} = {\left( {1 - p} \right)}{\left( {U{\left( {t^{{B*}} \overline{w} ^{p} \frac{p} {{1 - p}}} \right)} - U{\left( {t^{{B*}} w_{j} \frac{p} {{1 - p}}} \right)}} \right)}$$

When \(j=k\!\!\), it leads to:

$$ {Φ} _{k}\left( t^{B},\overline{w}\,^{p}\right) >0\, {\rm{since}} \;{{w_{k} < \overline{w}\,^{p}}}\rm{ and }{\it{U}}\;\rm{ is \,\,an \,\,increasing \,\,function}$$

i.e., \(x_{k}>0\) for \(t^{A}=t^{B*}\) and for any \(\overline{w}\,^{p}.\)

The level of taxation is chosen by the representative of income \(w_{k}.\) The program can be written as:

$$ {\mathop {\arg \;\max \Phi _{k} {\left( {t^{A} ,\overline{w} ^{p} {\left( {x_{1} {\left( {t^{A} } \right)},...,x_{M} {\left( {t^{A} } \right)}} \right)}} \right)}}\limits_{t^{A} } } $$ $$ {\mathop {\arg \;\max \Phi _{k} {\left( {t^{A} ,\overline{w} ^{p} {\left( {x_{1} {\left( {t^{A} } \right)},...,x_{M} {\left( {t^{A} } \right)}} \right)}} \right)}}\limits_{t^{A} } } $$

the maximum is obtained at \(t_{O}^{A\ast }.\) At this point, we obtain:

$$ x_{k}={Φ} _{k}\left( t_{O}^{A\ast },\overline{w}\,^{p}\left( x_{1}(t_{O}^{A\ast }),...,x_{M}(t_{O}^{A\ast })\right) \right) \geq {Φ} _{k}\left( t^{B\ast },\overline{w}^{p}\left( x_{1}(t^{B\ast }),...,x_{M}(t^{B\ast })\right) \right) >0$$

We then obtain that \(x_{k}>0\) at the optimum. Thus, \(t_{O}^{A\ast }<t_{s}\) according to Propositions 2 and 3.

Sign of \(\protect\eta\) terms

Indeed, for a CRRA utility function with \(\lambda >1\), the sign of \(\frac{ dx_{j}}{dt^{A}}\) at the optimum is given by (substituting Eq. 9 in Eq. 10):

$$\matrix { \frac{dx_{j}}{dt^{A}} &=&\!\!\!\!\!\!-pw_{j}U^{\prime }\left( w_{j}\left( 1-t^{A}\right) \right) +pw_{k}U^{\prime }\left( w_{k}\left( 1-t^{A}\right) \right) \notag \cr &=& {\kern-27pt {-p.w_{j}^{1-\lambda }\left( 1-t^{A}\right) ^{-\lambda }+p.w_{k}^{1-\lambda }\left( 1-t^{A}\right) ^{-\lambda } } }}$$

thus, for \(\lambda >1\)

$$ \frac{dx_{j}}{dt^{A}}>0\Longleftrightarrow w_{k}<w_{j}$$

Moreover, recall that

$$ \frac{\partial \overline{w}^{p}}{\partial x_{j}}>0\Longleftrightarrow j{\rm \text \,\,such \;\; that \;\; }{{w_{j}>\overline{w}\,^{p}}}$$

\(\eta\) is positive if the sum of the absolute values of the terms for which \(w_{j}\) is between \(w_{k}\) and \(\overline{w}^{p}\) is smaller than the sum of the other terms. Moreover, Eq. 16 allows us to state that:

$$ \frac{{dx_{j} }} {{dt^{A} }} > \frac{{dx\prime _{j} }} {{dt^{A} }} > \;{\text{for}}\;w_{k} < w\prime _{j} < w_{j}$$

Sensitivity to migrate \(\frac{dx_{j}}{dt^{A}}\) is thus lower for agents with incomes that are close to the policy-maker’s. Lastly,

$$ \frac{\partial \overline{w}\,^{p}}{\partial x_{j}}<\frac{\partial \overline{w} \,^{p}}{\partial x_{j^{\prime }}}<0{\rm\text \;\; for \;\; }w_{j}< {{w_{j^{\prime }}<\overline{ w}\,^{p}}}$$

which means that the reaction \(\frac{\partial \overline{w}^{p}}{\partial x_{j}}\) of the average income of country \(A\) to the net flow of migration of group \(j\) has an absolute value that is lower for agents with an income that is close to the ex post average income.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Rossignol, S., Taugourdeau, E. Asymmetric social protection systems with migration. J Popul Econ 19, 481–505 (2006).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Social insurance
  • Migration

JEL Classification

  • D7
  • H5
  • F22