Why is this Python program running 3 times faster than the similar Julia program

Question

I wanted to try the Julia programming language since I heard that it should be faster than Python. I decided to try the dynamic program for Project Euler # 14 , as it concerned a large number of calculations (finding the longest sequence of Collatz).

I wrote a program for it in Julia, and then decided to try to make a similar program in Python, so see how fast each one is. However, on my machine, the Python program runs in about 2 seconds, while Julia takes about 7. This surprised me because, as I said, Julia claims to be faster than Python, so I was wondering if something that is inherent in the scripts that make Julia so slower. I tried to make them both work out the same way.

Here are the scenarios, I would appreciate any insight!

Julia:

global max_val = 0
global max_item = 0
global d = Dict{Integer,Integer}()
function coll_length(n)
    #println(n)
    global max_val
    global max_item
    global d
    if haskey(d,n)
        return d[n]
    end
    if n==1
        answer = 1
    elseif n%2==1
        answer = 1+coll_length(3n+1)
    else
        answer = 1+coll_length(n/2)
    end

    d[n]=answer
    if max_val<answer
        max_val=answer
        max_item=n
    end
    return answer
end


for i = 1:1000000
    coll_length(i)
end
println(max_item)

Python:

d = {}
max_val = 0
max_item = 0

def coll_length(n):
    global d 
    global max_val 
    global max_item 

    if n in d:
        return d[n]
    if n==1:
        answer= 1
    elif n%2==0:
        answer= 1+coll_length(n/2)
    else:
        answer= 1+coll_length(3*n+1)

    d[n]=answer
    if max_val<answer:
        max_val=answer
        max_item=n
    return answer

for n in range(1,1000000+1):
    coll_length(n)
print max_item