# Proving the existence of a steady state

I am trying to prove the existence of at least one stable steady state in a situation where I am given properties of the production function but not an explicit functional form. I have the following (implicit) first order difference equation for the evolution of capital (comes from an overlapping generations model with no uncertainty):

$$(1+n)k_{t+1}(k_t)=s(w_t,R_{t+1})$$ Where $$k$$ is per capita capital, $$n$$ is population growth, $$s$$ is savings, $$w$$ is wages (the problem assumes that labor is inelastically supplied), and $$R$$ is rental rate on capital, with the appropriate time subscripts. From the firm's problem, and the assumption that the production function is CRS, this can be re-written as: $$(1+n)k_{t+1}(k_t)=s(f(k_t)-k_tf'(k_t), f'(k_t)-\delta)$$ Where $$\delta$$ is depreciation. We are given standard assumptions on $$f(k)$$: $$f(0)=0, f'(k)>0, f''(k)<0$$, Inada conditions: $$\lim\limits_{k \to \infty}f'(k)=0,\lim\limits_{k \to 0}f'(k)=\infty$$

I have been able to prove that

1) $$k_{t+1}(0)=0$$

2) $$\lim\limits_{k \to \infty} \frac{k_{t+1}}{k_t}=0$$, implying that at some point, $$k_{t+1}$$ lies below the $$45^\text{o}$$ line

3) Solving for the derivative of $$k_{t+1}$$ with respect to $$k_t$$: $$k_{t+1}^{'}(k_t)= \frac{-\frac{\partial s}{\partial w_t}k_tf''(k_t)}{1+n-\frac{\partial s}{\partial R_{t+1}}f''(k_{t+1})}$$

Making some additional assumptions on the derivatives of savings with respect to wages and interest (that aren't relevant to my question), I have that this is positive.

However, in order to show that there exists at least one stable steady state, I need to show that $$k_{t+1}$$ crosses the $$45^\text{o}$$ line from above. This is where I am stuck. My professor says that this follows from the Inada conditions on $$f(k)$$, but I'm not seeing it. If I can show that $$\lim\limits_{k \to 0}k_{t+1}^{'}(k_t)=\infty$$, then I think I'm set. And this will hold if $$\lim\limits_{k \to 0}k_tf''(k_t)=\infty$$ (I think). But the Inada conditions are on $$f'(k_t)$$, not $$f''(k_t)$$, so I don't know how to get what I need.

Thanks very much in advance for the help!!

• Hi Julia, wouldn't your postulation (2) that $\lim\limits_{k \to \infty} \frac{k_{t+1}}{k_t}=0$ imply that it reaches a steady state? Isn't $\frac{k_{t+1}}{k_{t}}=0$ the growth rate of capital being zero and hence stable? I am not familiar with OLG as I solved a similar problem but in terms of differentials. I interpret that as the growth rate approaching a steady state. You could probably back that argument up using the Inada conditions to state that the graph of $k_{t}$ with respect to time starts vertically and becomes horizontal, where it necessarily starts above in the first instance. – Brennan Jul 17 '19 at 16:58
• Hi @Brennan 2, Thanks for responding. The interpretation of $\frac{k_t+1}{k_t}$ is a bit different than how you've written it. Suppose t=3 and $k_3$ is very large (approaching the limit). Then $k_4$ is quite small in relation to $k_3$ (hence the fraction approaches zero) so the difference between the two is large, and the growth rate = $\frac {k_4-k_3} {k_3}$ is large (and negative) - sort of the opposite of a steady state! – Julia B Jul 19 '19 at 19:59
• Postulation (2) implies that, if $k_{t+1}$ starts out steep enough (i.e. with a slope greater than one), then at some point it will cross the 45 degree line, i.e. where $k_{t+1} = k_t$, implying a stable steady state, but I don't have (yet) that that slope is high enough. I hope that helps. Thank you again! – Julia B Jul 19 '19 at 20:00