Simulation Dickey-Fuller distribution

Question

I tried to simulate the Dickey- Fuller statistics with a short Matlab routine. I set Y0=0 and generated $$ Y_{t}=Y_{t-1} + \varepsilon_t, $$ where $\epsilon_t$ is $\mathcal{N}(0,1)$-distributed. Then I estimate the model $$ \Delta Y_t = \beta Y_{t-1} + \varepsilon_t $$ and computed the Dickey-Fuller statistic in order to estimate the quantile.

---

 function [Y] = genY(n)
 %generate random walk.
 Y=zeros(1,n);
 Y(1)=0;
 for i=2:n
     Y(i)=Y(i-1)+randn;
 end
 end

---

 function [dist] = estBeta(n)
 %estimate  Dickey-Fuller statistics in one step
 Y=genY(n);
 DeltaY=Y(2:n)-Y(1:(n-1));
 est=(sum(DeltaY(2:end).*DeltaY(1:(end-1))))/(sum((DeltaY(1:(end-1))).^2));
 dist=est/(std(DeltaY)/sqrt(n-1));
 end 

---

 function [quantil] = perc(m)
 %calculation of quantile
 z=zeros(1,m);
 for i=1:m
    z(i)=estBeta(1000);
 end
 quantil=quantile(z,0.01);
 end

If I execute the program I will always end up with -2.3 at p=0.01. Has anybody an idea how to fix this problem?